• 专利附录
  • 专利说明
  • 权利要求
  • 类似技术
  • 同族专利
  • 引用文献
  • 法律状态
  • 优先权
    • 专利摘要:
    A method of geolocation of the environment of a carrier comprising the following steps: - Acquiring (201) a plurality of shots of the environment of the carrier, - Determining (202) at least one information relating to the position of the carrier. carrier for a set of shots, - generating (203) a virtual reconstruction of the environment from the shots of said set, position information and a measurement bias parameter (204) for each position, said virtual reconstruction being parameterized according to the measurement bias (204) - Modify (206) the measurement bias parameter (204) to be applied to each position to obtain a virtual reconstruction parameterized by a modified measurement bias, the modification (206) ) being performed so as to minimize a distance between said modified virtual reconstruction and a prior virtual representation (205) of the environment.
    公开号:FR3019310A1
    申请号:FR1452786
    申请日:2014-03-31
    公开日:2015-10-02
    发明作者:Dorra Larnaout；Steve Bourgeois；Vincent Gay-Bellile；Michel Dhome
    申请人:Commissariat a lEnergie Atomique CEA；Commissariat a lEnergie Atomique et aux Energies Alternatives CEA；
    IPC主号:
    专利说明:

    [0001] The invention relates to the field of geo-location systems in an external environment, for example a road, or interior, for example the interior of a building. The invention relates more specifically to a method and a system for geo-location of a carrier, in particular a mobile carrier, which allows the precise positioning of the carrier in a virtual reconstruction of his environment obtained from shots via a system. observation of the environment. Triangulation localization systems, for example satellite positioning systems, localized Wifi terminals or cellular telephony networks, are generally subject to location errors which can be modeled by measurement noise and bias. . The noise is usually random and of small magnitude. The measurement bias is generally more important and substantially stable for a more or less limited period of time as well as over a spatial area. The existence of measurement bias directly impacts geo-location systems that require high precision in the required positioning. A technical problem to solve is to find a solution to compensate or correct this measurement bias to make more accurate positioning. The estimation of the measurement bias nevertheless remains complex since it can be linked to the combination of phenomena as different as the nature of the materials traversed by the triangulation signal, the phenomenon of multiple reflection of the signal emitted during its course of travel. to the receiver, the use of a small number of visible transmitters, positioning errors on the transmitters or poor synchronization of the clocks between the different transmitters.
    [0002] To estimate the bias affecting navigation data from a triangulation location system, methods are known based on the measurement of the vector connecting the current position provided by the triangulation location system with that provided by a perception system. of the environment. Such methods are notably described in documents [1] and [2] which concern a camera-based perception system and in document [3] which describes a solution based on a camera pair or stereo head. Other environmental perception systems, such as lidar or radar systems, are also contemplated in the prior art. According to the aforementioned methods of the prior art, the measurement of said vector generally involves uncertainties associated with each measurement of position, that is to say both the measurements provided by the triangulation location system such as those provided by the perception system. This measurement is supplemented by a prediction of the measurement bias and its associated uncertainty which is based on a bias evolution model and which is estimated using filtering methods of the Kalman filter type. ([1]) or Particle Filter ([2]). These filtering estimation methods thus make it possible to take into account the fact that the bias can sometimes be stable from one measurement to another or evolve slowly. The methods described in the three aforementioned documents, however, have several disadvantages as to their robustness. First, the methods used to predict and update the bias estimate are not robust to the presence of erroneous estimates of bias. 30 Indeed, these methods do not make it possible to call into question the estimates, made at an earlier time, of the value and the uncertainty of the bias because they are not reevaluated. On the contrary, in the case of a bad estimate, the mechanism for predicting bias and propagating the associated uncertainty will tend to propagate the error committed. In the case where the predicted uncertainty is incompatible with the uncertainties estimated by the different location systems, some systems propose to restart the estimate of the bias at zero (in other words to consider that the bias is zero with great uncertainty). . If this is generally preferable to the propagation of a bad estimate, this solution leads to a loss of precision.
    [0003] A second problem of robustness of the methods known from the prior art is related to the fact that the position provided by the perception system is based on the matching of the current observations (for example the current image of the camera in the document [1 ], or the current reconstruction of a stereo head in the document [3]) provided by the perception sensor with a map known to the environment. An example of mapping is mapping the position of a pedestrian crossing in the image of a camera to the position of that pedestrian in a road map. In order to perform this mapping, the known solutions exploit the estimated position at the current time of the system. This approach has two disadvantages. Firstly, since the navigation data provided by the location system is used as a priori to guide the mapping, the lack of estimation of the measurement bias on these data or an inaccurate estimate or an erroneous estimate of that one. this can disturb this matching step. Secondly, even in the presence of a correct estimate of the measurement bias on the navigation data, it is unrealistic to consider the location provided by the environmental perception system as completely reliable. Indeed, using the observations provided only at the current time by the perception system may prove too discriminating to allow to establish a precise correspondence with the map, due to the existence of localization ambiguities. For example, a junction with a high concentration of pedestrian crossings may lead to an association error. An association error can occur and result in a location whose uncertainty is very undervalued or even inconsistent with the local location. As noted above, the bias estimation process is not robust to outliers.
    [0004] In the document [3], the authors propose to locate a system carrying a GPS receiver, an inertial unit and a couple of cameras. The system uses the data provided by all of these sensors to estimate the position of the system using a particle filter algorithm. To estimate the bias affecting the data provided by the GPS receiver, this solution proposes to align the reconstruction of the environment provided at the current time by the pair of cameras with a three-dimensional map of the environment. This approach is limited in terms of robustness and accuracy. Indeed, the alignment of a single stereo reconstruction with a 3D model of the environment can be ambiguous when the geometry of the observed scene has repetitive structures or a very simple structure. For example, the observation of a plane wall does not allow to constrain the position of the system vis-à-vis a translation parallel to the wall. The environmental perception system may therefore provide an erroneous localization that will disturb the bias estimate. In such a case, the GPS bias risks either being badly estimated (mis-match) or being infrequently estimated (the system identifies the matching problem and decides not to estimate the bias).
    [0005] The present invention aims at remedying the drawbacks of the solutions of the prior art by proposing a method of geo-localization of the environment of a carrier who exploits both navigation data provided by a triangulation location system and a virtual reconstruction of the environment provided by a perception system and which makes it possible to correct the measurement bias on the navigation data. The approach proposed by the invention makes it possible to offer a more accurate and more robust estimate of the bias of the location system. In the nonlimiting case where the perception system is a stereo head, instead of considering only a stereo reconstruction provided by a stereo head at the current time, the invention allows to exploit the N last observations of this stereo head so to create a more complete reconstruction of the environment respecting the constraints of the multi-view geometry induced by all the pairs of stereo images. The reconstruction obtained, although imperfect, offers a better discriminating power during the step of mapping to a 3D map of the environment since this reconstruction covers a larger area of the scene, which reduces the probability that this it contains only repetitive structures or a structure that is too simple. The invention therefore has the advantage of allowing a more robust matching, which limits the risk of outliers. In addition, the bias estimated by the method according to the invention is that for which a reconstruction from the last N observations of the stereo head is aligned with the best 3D map of the environment while checking, in this case, the constraints of multi-view geometry. The presence of these multi-view constraints and the fact that the estimation of biases affecting the data is done in a global manner reduces the probability that a local error of reconstruction by the stereo head could disturb the estimate of the bias. The fact that the bias of each navigation data is re-estimated at each instant makes it possible to call into question past estimates, and thus favors the detection and correction of erroneous bias estimates.
    [0006] The subject of the invention is a method of geolocation of the environment of a wearer comprising the following steps: - Acquiring a plurality of shots of the wearer's environment, - Determining at least one information relating to the position the carrier for a set of shots, - Generate a virtual reconstruction of the environment from the shots of said set, position information and a measurement bias parameter for each position, said virtual reconstruction being parameterized according to the measurement bias, - Modifying the measurement bias parameter to be applied to each position to obtain a virtual reconstruction parameterized by a modified measurement bias, the modification being performed so as to minimize a distance between said modified virtual reconstruction and a virtual representation a priori of the environment. According to a particular embodiment of the invention, the generation of a virtual reconstruction of the environment comprises: - The correction of each navigation data by a bias parameter, - The generation of a virtual reconstruction of the environment from the plurality of shots and corrected position information.
    [0007] According to a particular embodiment of the invention, the generation of a virtual reconstruction of the environment comprises: - The generation of a virtual reconstruction of the environment from the plurality of shots, - The application at said virtual reconstruction, a deformation calculated from position information and bias parameters.
    [0008] According to a particular embodiment of the invention, the generation of a virtual reconstruction of the environment comprises the generation of a plurality of geometrical elements parameterized by said measurement bias and defining said reconstruction and the modification of the virtual reconstruction. of the environment comprises the following sub-steps, executed iteratively: a step of mapping between a plurality of parametric geometrical elements of the virtual reconstruction and a plurality of fixed geometric elements of the virtual representation a priori, step of calculating at least one distance between a plurality of parameterized geometrical elements of said virtual reconstruction and a plurality of corresponding fixed geometrical elements in said a priori representation, - a step of modifying said measurement bias parameter so as to minimize said distance. According to a particular embodiment of the invention, the same bias value is associated with a group of shots. According to a particular embodiment of the invention, the values of the biases associated with the different shots are linked together by a parametric or non-parametric model. According to a particular embodiment of the invention, said set of shots is taken in a sliding window in time or in space. According to one particular embodiment of the invention, the distance between said virtual reconstruction and the prior virtual representation is a distance in the point-to-point space or point-to-plane or plane-to-plane or a re-projection error in the plane or a combination of several of these distances. According to a particular embodiment of the invention, the prior virtual representation of the environment is a cartographic representation comprising at least one model from a terrain elevation model, a three-dimensional model of the buildings of an area geographic, a cloud of points in three dimensions, an architect's plan. According to a particular embodiment of the invention, the acquisition of a plurality of shots of the wearer's environment is carried out using a perception system of the environment of the wearer.
    [0009] According to a particular embodiment of the invention, the determination of at least one carrier navigation data for a set of shots is performed using a triangulation location system. The subject of the invention is also a computer program comprising instructions for carrying out the method of geolocation of the environment of a carrier according to the invention, when the program is executed by a processor and a support device. processor-readable record on which is recorded a program comprising instructions for executing the method geolocation of the environment of a carrier according to the invention, when the program is executed by a processor. The invention also relates to a geo-location system, intended to be embedded on a carrier, comprising a perception system of the environment of the carrier adapted to acquire a plurality of shots, a positioning system by suitable triangulation providing at least one navigation datum for each shot, a database containing a prior virtual representation of the environment and a processor configured to execute the method of geolocation of the environment of a carrier according to the invention to produce a geo-localized virtual reconstruction of the carrier environment.
    [0010] The carrier's environmental perception system may be a two-dimensional camera or a couple of two-dimensional cameras or a three-dimensional camera or a lidar system or system. radar.
    [0011] The triangulation localization system can be a satellite geo-location system or a Wi-Fi geo-location system or a mobile network network geo-location system. Other features and advantages of the present invention will appear better on reading the description which follows in relation to the appended drawings which represent: FIG. 1, a block diagram of a geolocation system according to the invention; FIG. 2 is a flowchart illustrating the steps for implementing the method according to the invention. FIG. 1 represents a block diagram of a geolocation system 100 according to the invention, intended to be embedded on a mobile carrier or not, for example a vehicle. Such a system includes an environmental perception system SPE which is capable of capturing images of the environment of the geolocation system 100. The perception system of the SPE environment may include but is not limited to a two-dimensional camera, a couple of two-dimensional cameras, a three-dimensional camera. , a lidar system or a radar system. Any other equivalent device for acquiring a plurality of images of the environment is compatible with the system 100 according to the invention. The system 100 according to the invention also comprises an SLT triangulation localization system which may include but is not limited to a GNSS receiver, for example a GPS receiver, a location receiver based on a WIFI network or on a network. cellular telephone system or any other system making it possible to obtain navigation data relating to the carrier of the system 100. The navigation data notably include any information concerning the position of the system 100 or any information making it possible indirectly to deduce information on the position of the system. 100, in particular speed or acceleration. The system 100 according to the invention also comprises a database containing an a priori representation of the environment RE which can take the form of a cartography of the environment in which the carrier of the system 100 is supposed to move in various forms. including, but not limited to, a terrain elevation model, a three-dimensional model of buildings in a geographic area, a three-dimensional point cloud, a building plan, but also data from a geographic information system (GIS) or visual representation and exploration software. The system 100 according to the invention also comprises a processor 101 comprising a first calculation module 110 configured to generate a virtual reconstruction of the environment from the captured shots, navigation data associated with these shots and a measurement bias. The processor 101 also comprises a second calculation module 120 configured to compare the virtual reconstruction determined by the first calculation module 110 with the prior representation of the environment stored in the database to deduce a precise geo-location of this reconstruction of the environment as well as an estimate of the measurement bias affecting the navigation data. The calculation steps implemented by the processor 101 are developed further upstream in the description. The processor 101 can be a generic processor, a specific processor, an application-specific integrated circuit (also known as ASIC for "Application-Specific Integrated Circuit") or a network of programmable gates in situ (also known by the English name of FPGA for "Field-Programmable Gate Array") or any other equivalent computing device. The geolocation method according to the invention described below can be implemented as a computer program comprising instructions for its execution. The computer program can be recorded on a recording medium readable by the processor 101. FIG. 2 represents the steps of implementing the geolocation method according to the invention in the form of a flowchart. In a first step 201, the perception system of the environment SPE acquires a plurality of shots of the environment. In other words, if the SPE perception system consists of a camera, it captures several images of the environment. This plurality of images can be composed of several successive images according to a temporal order or of several images of the same scene taken according to different spatial points of view. In a second step 202, the location system SLT determines, for each shot of the environment, information on the position of the system 100. The estimated positions are tainted by a measurement bias that one seeks to estimate precisely. The bias can be independent for each shot or can follow a variation verifying known properties, for example a variation pattern. The principle underlying the invention consists in estimating, for each shot, the bias 204 impacting the position estimated by the location system, which makes it possible to obtain a virtual reconstruction of the environment nearest to a priori representation 205 of the environment. In a third step 203, a geo-localized virtual reconstruction of the environment is generated from the different shots and the positions associated with each shot. In addition, a bias parameter is associated with each position so as to produce a multi-view parametric reconstruction as a function of the bias values associated with each shot. MW is the virtual reconstruction generated. This can consist of a set of points, segments, regions, plans or even a mesh, a map of occupation, or a volumetric probability representation. All these examples are given for illustrative and not limiting, any type of virtual representation of a scene can be envisaged by the skilled person.
    [0012] To generate a reconstruction of an environment from different shots, there are several state-of-the-art algorithms that can be used depending on the sensor envisioned for the perception system of the environment. For example, if the environmental perception sensor is a camera then "slam" type algorithms, as described in references [4] and [5] or "structure from motion" as described in reference [13]. ] make it possible to obtain a sparse or dense three-dimensional reconstruction of the environment. In the context of a lidar or laser type sensor, the various point reconstructions can be agglomerated into a single overall reconstruction consisting of approaches of ICP (iterative closest point) type as described in the publication [14]. The positions provided by the location sensor, denoted by Pw = i varying from 1 to N with N the number of shots, can be used directly by some of the abovementioned algorithms in order to obtain a localized three-dimensional reconstruction in a reference frame. W of the localization system. These positions are subject to a bias parameter noted b. We then denote by f (0, Pw, b) the localized virtual reconstruction function that makes it possible to obtain the MW reconstruction of the environment from observations O 30 = pl with j varying from 1 to N, resulting from the N acquisitions of the system. of perception.
    [0013] Any other known method equivalent to those described in references [4], [5], [13] and [14] which make it possible to obtain a geo-localized virtual representation of a visual environment from several photographic views. this environment and associated locations can be used instead of the referenced methods. We can also mention the methods described in documents [6] and [7]. According to a first variant embodiment of step 203 of the method according to the invention, the parametric virtual reconstruction of the environment can be generated by first correcting each position information of a bias parameter and then generating the virtual reconstruction to from shots and corrected positions set through. According to a second embodiment of step 203 of the method according to the invention, the virtual configuration of the environment can be generated directly from the shots. We denote g the reconstruction function applied to the observations O to obtain the reconstruction ML in a geometric reference L potentially different from the reference W attached to the location system SLT. In a second step, a strain A is calculated from the Pw data from the triangulation location system and the bias parameters b associated with each of the triangulation location data. Deformation A is applied to the reconstruction ML so that it is expressed in the W coordinate system of the location system and is consistent with the data from the triangulation location system. The deformation A used may be a similarity or a more evolved approach that will take into account local positioning errors such as the estimation of a non-rigid transformation, for example a B-spline or TPS (Thin Plate) type deformation. Spline) In a fourth step 206, the parameterized reconstruction obtained at the end of the third step 203 is compared with the prior representation of the environment 205 in order to find the optimal values of the bias parameters to be applied to each view to obtain a reconstruction of the environment most similar to the representation a priori. The main origin of the shift between the MW reconstruction and the prior representation of the Aw environment is the bias of the location sensor. Other sources of error are local inaccuracies due to the method of reconstructing the virtual environment or a priori about the imperfect environment. The optimal values of the bias parameters affecting the location sensor are those that generate the virtual reconstruction of the environment which is aligned at best with the environment a priori. Thus the bias parameters b are estimated by minimizing an error between the virtual reconstruction and the prior representation of the environment. This error can be calculated as a distance between the reconstruction and the representation a priori, the problem to be solved then consists in finding the values of the parameters of bias b = {b;}, for i varying from 1 to N, which minimize the distance d, which can be expressed as the minimization of the cost function F (b) = d (f (0, Pw, b), 4w). A modification of the bias causing a modification of the reconstruction via the function f (0, Pw, b), the virtual reconstruction of the environment is re-estimated during the minimization process thus making it possible to call into question the initial reconstruction that was tainted by bias. The distance d gives information on the differences between the reconstruction and the prior representation. Any geometric distance 30 to obtain information on these differences can be considered. In particular, the distance d may be a distance in the point-to-point or point-to-plane or plane-to-plane space or a re-projection error in the plane as introduced in documents [17] and [18] or a combination of several of these distances. More generally, the distance d is calculated between two corresponding geometric elements, for example between a point of the virtual reconstruction and the corresponding point in the prior representation of the environment. The minimization of the cost function F (b) can be carried out by any known linear or nonlinear optimization algorithm such as for example the Levenberg Marquardt algorithm described in document [15] or the solutions described in the articles. [20], [21], [22] and [23]. When the correspondences between the reconstruction and the prior representation are not known, we proceed in two stages. In a first step, a mapping is chosen between the geometric elements of the reconstruction and those of the prior representation and then in a second step the cost function F (b) is minimized. The two steps can be iterated several times in order to call into question the complete reconstruction process due to the variations of bias at each iteration of the optimization, which thus makes it possible to take into account the local inaccuracies of the reconstruction. Such an iterative approach is described in particular in document [16]. The result of the operation of minimizing the cost function F (b) makes it possible to obtain the optimal bias values b = {b;} for correcting the positioning information and also makes it possible to obtain a geo virtual reconstruction. -Located with a positioning whose accuracy is improved compared to the exploitation only of the data provided by a triangulation localization system. According to an alternative embodiment of the invention, the same bias parameter value can be applied to a group of shots according to information external to the system. For example, a hypothesis can be taken as to the temporal evolution of the horizon bias of the number of shots exploited. In other words, the bias can be assumed constant over a period equivalent to several shots.
    [0014] Taking into account hypotheses of variations in the bias makes it possible to reduce the number of parameters of the cost function to be minimized and thus to reduce the complexity. External information can be taken into account to refine the bias evolution model. For example, in the case of a satellite positioning system, information concerning the loss of signal received by a satellite can be exploited to refine the ranges of assumed variations of the bias. According to another variant embodiment of the invention, the various bias parameters can be linked by means of a parametric model comprising, for example, a translation, a rigid transformation, a similarity, a piecewise similarity, a deformation of the type B-Spline or TPS-type transformation (Thin Plate Spline). The bias parameters can also be linked using a non-parametric model, for example a model based on a regularized strain field. The use of a parametric or non-parametric model to link the bias parameters again makes it possible to limit the number of parameters of the cost function and thus reduce the complexity of resolution. According to another embodiment of the invention, the selected images are taken in a sliding time window, which makes it possible to reduce the number of data to be taken into account for the virtual reconstruction but also indirectly to limit the number of images. bias settings.
    [0015] In the case where a reconstruction method of the simultaneous localization and mapping type is used, the selected images can be selected as key images. The shots can also be selected within a spatial window, that is, several different viewpoints of the same scene can be selected. The invention makes it possible to provide precise geo-location of the wearer's environment by offsetting the measurement bias of the navigation data provided by a triangulation location system. Unlike the solutions of the prior art, the invention is not based on an estimation of measurement bias by a difference between the current position provided by the triangulation location system and the current location provided by a system. of perception. On the contrary, the invention uses an environment reconstruction method exploiting both the observations of the perception system and the location information provided by a triangulation system, and a search of the bias values for which the correction of the bias in the reconstruction process would better align the reconstruction of the environment with a geometric model of the environment known a priori. The criterion used to characterize the bias estimation quality is therefore different from the usual criteria based on an alignment of the location measurements of a triangulation system with those provided by a perception system. The criterion used by the invention instead consists in an alignment of the reconstruction with a model of the environment. Using a reconstruction exploiting several observations of the perception system, additional geometric constraints are introduced, which makes the system according to the invention more robust than conventional systems exploiting only the current observations.
    [0016] In addition, this reconstruction covering a larger area of the environment, it allows to establish an alignment with the geometric model known a priori with greater certainty and with greater frequency. By constantly re-estimating bias for a set of triangulation data, the invention offers greater robustness to erroneous estimates because past estimates are challenged by the new data provided by triangulation location systems. of perception.
    [0017] The method according to the invention allows a more precise, more frequent and more robust estimation of the bias affecting the data resulting from a triangulation localization system. This solution can be used with a large number of triangulation and perception localization systems as well as a large number of geometric models a priori on the environment. In particular, the invention can operate with low cost sensors, such as a consumer GPS receiver and a simple camera and a low-definition 3D model of the environment. The ratio between the cost and the ease of deployment and the quality and continuity of the localization service offered by this solution makes it possible to envisage exploitation in a large number of applications, whether in the field of navigation aids, for example augmented reality guidance or in the field of robots or autonomous vehicles.30 References [1] J. Laneurit, R. Chapuis and F. Chausse, "Accurate Vehicle Positioning on a Numerical Map, "INTERNATIONAL JOURNAL OF CONTROL AUTOMATION AND SYSTEMS, 2005. [2] KCMS Kichun Jo," GPS-bias bias correction for precise localization of autonomous vehicles, "at Intelligent Vehicles Symposium (IV), 2013 IEEE, 2013. [3] LDD Vernaza P., "Robust GPS / INS-aided localization and mapping via GPS bias estimation," at 10th International Symposium on Experimental Robotics, 2008. [4] LMDMDFSP Mouragnon E., "Real-Time Localization and 3D Reconstruction," at IEEE Conference on Computer Vision and Pattern Recognition, 2006. [5] K. Georg and M. David, "Parallel Tracking and Mapping for Small AR Workspaces," at International Symposium on Mixed and Augmented Reality, 2007. [6] L. S.,. DA Newcombe Richard A., "DTAM: Dense tracking and mapping in real-time," at IEEE International Conference onComputer Vision, 2011. [7] V. Pradeep, C. Rhemann, S. Izadi, C. Zach, M. Bleyer and S. Bathiche, "MonoFusion: Real-time 3D reconstruction of small scenes with a single web camera," at IEEE International Symposium onMixed and Augmented Reality, 2013. [8] RA Newcombe, S. Izadi, O. Hilliges, D. Molyneaux, D. Kim, AJ Davison, P. Kohli, J. Shotton, S. Hodges, and A. Fitzgibbon, "KinectFusion: Real-Time Dense Surface Mapping and Tracking," at International Symposium on Mixed and Augmented Reality, 2011. [ 9] M. Lhuillier, "Fusion of GPS and Structure-from-Motion using Constrained Bundle Adjustments," at IEEE Conference on Computer Vision and Pattern Recognition, 2011. [i0] C. Strecha, T. Pylvanainen and P. Fua, " Dynamic and Scalable Large Scale Image Image Reconstruction, "at 3D Data Processing, Visualization and Transmission, 2010. [11] RS Kaminsky, N. Sn avely, SM Seitz and R. Szeliski, "Alignment of 3D point clouds to overhead image," at Workshop on Internet Vision, 2009. [12] N. Sunderhauf, "Robust Optimization for Simultaneous Localization and Mapping," (PhD), 2012 [13] S. Agarwal, N. Snavely, I. Simon, S. Seitz, and R. Szeliski, "Building Rome in a Day," at ICCV, 2009. [14] WD Lui, TJJ Tang, T. Drummond, and W. Li, "Robust estimating estimation using ICP in inverse depth coordinates.," At ICRA, 2012. [15] K. Levenberg, "A method for the solution of some non-linear problems in least squares.," At Quarterly of Applied Mathematics, 1944. [16] RM Neal and GE Hinton, "A View of the EM Algorithm that Justifies Incremental, Sparse, and Other Variants.," At Learning in Graphical Models, 1998. [17] P. Lothe, S. Bourgeois, F. Dekeyser, E. Royer and M. Dhome, "Towards Geographical Referencing of Monocular SLAM Reconstruction Using 3D City Models: Application to Real-Time Accurate Vision-Based Localizati on, "at IEEE Conference on Computer Vision and Pattern Recognition, 2009. [18] M. Tamaazousti, V. Gay-Bellile, S. Naudet-Collette, S. Bourgeois and M. Dhome," NonLinear Refinement of Structure of Motion Reconstruction by taking advantage of a partial knowledge of the environment, "at CVPR, 2011. [19] C.-P. Wang, K. Wilson and N. Snavely, "Accurate georegistration of point clouds using geographic data," at 3DV, 2013. [20] http://en.wikipedia.org/wiki/Iteratively_reweighted_least_squares [21] http: // en .wikipedia.org / wiki / Conjugate_gradient_method [22] Lourakis, M. IA; Argyros, A.A., "Is Levenberg-Marquardt the most efficient optimization algorithm for implementation bundle adjustment ," Computer Vision, 2005. ICCV 2005. Tenth IEEE International Conference on, vol.2, no., Pp.1526, 1531 Vol. 2, 17-21 Oct. 2005 [23] http://en.wikipedia.org/wiki/Simulated_annealing
    权利要求:
    Claims (16)
    [0001]
    REVENDICATIONS1. A method of geolocation of the environment of a carrier comprising the following steps: - Acquiring (201) a plurality of shots of the environment of the carrier, - Determining (202) at least one information relating to the position of the carrier. carrier for a set of shots, - generating (203) a virtual reconstruction of the environment from the shots of said set, position information and a measurement bias parameter (204) for each position, said virtual reconstruction being parameterized according to the measurement bias (204), - Modifying (206) the measurement bias parameter (204) to be applied to each position to obtain a virtual reconstruction parameterized by a modified measurement bias, the modification ( 206) being performed so as to minimize a distance between said modified virtual reconstruction and a prior virtual representation (205) of the environment.
    [0002]
    2. A method of geolocation of the environment of a carrier according to claim 1 wherein the generation (203) of a virtual reconstruction of the environment comprises: - The correction of each navigation data by a bias parameter (204), - Generation of a virtual reconstruction of the environment from the plurality of shots and corrected position information.
    [0003]
    3. A method of geo-localization of the environment of a carrier according to claim 1 wherein the generation (203) of a virtual reconstruction of the environment comprises: - The generation of a virtual reconstruction of the environment to from the plurality of shots, - The application, to the said virtual reconstruction, of a deformation calculated from position information and bias parameters (204).
    [0004]
    4. A method for geolocation of the environment of a carrier according to claim 1 wherein the generation (203) of a virtual reconstruction of the environment comprises the generation of a plurality of geometric elements parameterized by said bias. method and defining said reconstruction and the modification (206) of the virtual reconstruction of the environment comprises the following sub-steps, performed iteratively: - A mapping step between a plurality of parameterized geometric elements of the virtual reconstruction and a plurality of fixed geometric elements of the prior virtual representation, - a step of calculating at least one distance between a plurality of parameterized geometric elements of said virtual reconstruction and a plurality of corresponding fixed geometric elements in said representation a priori, - A step of modifying said mesu bias parameter re so as to minimize said distance.
    [0005]
    5. A method of geolocation of the environment of a carrier according to one of the preceding claims wherein the same bias value is associated with a group of shots.
    [0006]
    6. A method of geo-localization of the environment of a carrier according to one of the preceding claims wherein the bias values associated with different shots are linked together by a parametric or non-parametric model.
    [0007]
    7. A method of geolocation of the environment of a carrier according to one of the preceding claims wherein said set of shots is taken in a sliding window in time or in space.
    [0008]
    8. A method of geo-localization of the environment of a carrier according to one of the preceding claims wherein the distance between said virtual reconstruction and the virtual representation a priori is a distance in the point-to-point space or point to plan or plane to plane or a re-projection error in the plane or a combination of several of these distances.
    [0009]
    9. A method of geolocation of the environment of a carrier according to one of the preceding claims wherein the prior virtual representation of the environment is a cartographic representation comprising at least one model from a terrain elevation model. , a three-dimensional model of the buildings of a geographical area, a cloud of points in three dimensions, an architect's plan.
    [0010]
    10. A method of geolocation of the environment of a carrier according to one of the preceding claims wherein the acquisition of a plurality of shots of the environment of the carrier is carried out using a environmental perception system of the wearer.
    [0011]
    11. A method of geolocation of the environment of a carrier according to one of the preceding claims wherein the determination of at least one navigation data of the carrier for a set of shots is performed using a triangulation localization system.
    [0012]
    12. A computer program comprising instructions for carrying out the method of geolocation of the environment of a carrier according to any one of claims 1 to 11, when the program is executed by a processor.
    [0013]
    A processor-readable recording medium on which is recorded a program comprising instructions for carrying out the method of geolocation of the environment of a wearer according to any one of claims 1 to 11, when the program is executed by a processor.
    [0014]
    14.System (100) of geo-location, intended to be embedded on a carrier, comprising a system (SPE) of perception of the environment of the carrier adapted to acquire a plurality of shots, a system (SLT) of location by triangulation capable of delivering at least one navigation datum for each shot, a database (RE) containing a prior virtual representation of the environment and a processor (101) configured to execute the geolocation method of the A carrier environment according to any one of claims 1 to 11 for producing a geolocated virtual reconstruction of the carrier environment.
    [0015]
    The geolocation system according to claim 14 wherein the carrier environment perception system (SPE) is a two-dimensional shooting camera or a pair of two-dimensional shooting cameras or a two-dimensional shooting camera. three-dimensional shooting camera or a lidar system or a radar system.
    [0016]
    16.System geolocation system according to one of claims 14 or 15 wherein the triangulation localization system (SLT) is a satellite geolocation system or a Wifi network geolocation system or a geo-localization system. location by mobile phone network.
    类似技术:
    公开号 | 公开日 | 专利标题
    US10134196B2|2018-11-20|Mobile augmented reality system
    EP2572319B1|2017-12-20|Method and system for fusing data arising from image sensors and from motion or position sensors
    EP3126864B1|2020-07-15|Method for geolocating the environment of a carrier
    Wang et al.2010|Automated road sign inventory system based on stereo vision and tracking
    Moreau et al.2017|Fisheye-based method for GPS localization improvement in unknown semi-obstructed areas
    Sermet et al.2020|Crowdsourced approaches for stage measurements at ungauged locations using smartphones
    KR102200299B1|2021-01-08|A system implementing management solution of road facility based on 3D-VR multi-sensor system and a method thereof
    EP3252615A1|2017-12-06|Method and system for determining cells crossed by a measuring or viewing axis
    FR2964774A1|2012-03-16|METHOD FOR ESTIMATING THE MOVEMENT OF A BEARER IN RELATION TO AN ENVIRONMENT AND CALCULATION DEVICE FOR A NAVIGATION SYSTEM
    EP2724203B1|2017-09-13|Generation of map data
    AU2018253281A1|2019-10-17|Vegetation detection and alert method and system for a railway vehicle
    Gaspar et al.2018|Urban@ CRAS dataset: Benchmarking of visual odometry and SLAM techniques
    AU2018253980A1|2019-09-12|Distributed device mapping
    US20190304161A1|2019-10-03|Dynamic real-time texture alignment for 3d models
    Wang et al.2021|Pointloc: Deep pose regressor for lidar point cloud localization
    FR3060115A1|2018-06-15|LOCATION OF A VEHICLE
    Lucks et al.2021|Improving trajectory estimation using 3D city models and kinematic point clouds
    Dabisias et al.2018|VALUE: Large Scale Voting-based Automatic Labelling for Urban Environments
    WO2019180160A1|2019-09-26|Method for geopositioning in a building, called ips, and device implementing said method
    CN111383337A|2020-07-07|Method and device for identifying objects
    CN112965076A|2021-06-15|Multi-radar positioning system and method for robot
    GB2572025A|2019-09-18|Urban environment labelling
    FR3079306A1|2019-09-27|METHOD FOR GEOLOCATION IN AN IPS BUILDING AND DEVICE IMPLEMENTING SAID METHOD
    FR3084864A1|2020-02-14|METHOD FOR DENSIFYING A DEPTH MAP
    FR3090148A1|2020-06-19|MULTIMODAL MAPPING RECALING METHOD BASED ON HUMAN REPETITIVE BEHAVIOR
    同族专利:
    公开号 | 公开日
    FR3019310B1|2016-04-01|
    WO2015150129A1|2015-10-08|
    US20170108338A1|2017-04-20|
    EP3126864A1|2017-02-08|
    EP3126864B1|2020-07-15|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    US20080033645A1|2006-08-03|2008-02-07|Jesse Sol Levinson|Pobabilistic methods for mapping and localization in arbitrary outdoor environments|
    US20100305854A1|2009-06-01|2010-12-02|Robert Bosch Gmbh|Method and apparatus for combining three-dimensional position and two-dimensional intensity mapping for localization|
    JP6322812B2|2014-08-21|2018-05-16|パナソニックＩｐマネジメント株式会社|Information management apparatus, vehicle, and information management method|
    US9600146B2|2015-08-17|2017-03-21|Palantir Technologies Inc.|Interactive geospatial map|
    US10368047B2|2017-02-15|2019-07-30|Adone Inc.|Six-degree of freedom video playback of a single monoscopic 360-degree video|
    WO2018201097A2|2017-04-28|2018-11-01|FLIR Belgium BVBA|Video and image chart fusion systems and methods|
    法律状态:
    2015-03-31| PLFP| Fee payment|Year of fee payment: 2 |
    2016-03-31| PLFP| Fee payment|Year of fee payment: 3 |
    2017-03-31| PLFP| Fee payment|Year of fee payment: 4 |
    2018-03-29| PLFP| Fee payment|Year of fee payment: 5 |
    2020-03-31| PLFP| Fee payment|Year of fee payment: 7 |
    2021-03-30| PLFP| Fee payment|Year of fee payment: 8 |
    优先权:
    申请号 | 申请日 | 专利标题
    FR1452786A|FR3019310B1|2014-03-31|2014-03-31|METHOD FOR GEO-LOCATION OF THE ENVIRONMENT OF A BEARER|FR1452786A| FR3019310B1|2014-03-31|2014-03-31|METHOD FOR GEO-LOCATION OF THE ENVIRONMENT OF A BEARER|
    PCT/EP2015/056037| WO2015150129A1|2014-03-31|2015-03-23|Method for geolocating the environment of a carrier|
    EP15711514.8A| EP3126864B1|2014-03-31|2015-03-23|Method for geolocating the environment of a carrier|
    US15/128,597| US20170108338A1|2014-03-31|2015-03-23|Method for geolocating a carrier based on its environment|
    [返回顶部]