• 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:
    Procedure for the punctual speed measurement of motor vehicles in short section with minimum error geometry. The invention relates to a method for the punctual measurement of the speed (2) of motor vehicles (3), by means of at least two cameras (4), (5) at a point (1), pointing to two regions (6).), (7) different from the track (8), calculating relative distances (10), (11) of the vehicle with respect to the cameras by detecting license plate (9) and its internal elements, storing timestamps, calculating the speed for what you give the possible combinations of distances (12) between cameras that are at the optimum distance that generates minimum error in the calculation of the speed, and calculating the average speed of all the velocity measurements obtained for optimal distances of minimum error. The minimum error geometry involves configuring the height, orientation and focal length of the cameras, to fix a first region, calculate the distance of the second region that minimize the error in speed. (Machine-translation by Google Translate, not legally binding)
    公开号:ES2665939A1
    申请号:ES201600904
    申请日:2016-10-28
    公开日:2018-04-30
    发明作者:David Fernández Llorca；Miguel Angel SOTELO VÁZQUEZ；Iván GARCÍA DAZA；Ignacio Parra Alonso；Carlota SALINAS MALDONADO
    申请人:Universidad de Alcala de Henares UAH；
    IPC主号:
    专利说明:

    PROCEDURE FOR THE SPECIFIC MEASUREMENT OF MOTOR VEHICLE SPEED IN SHORT SECTION WITH MINIMUM GEOMETRY 5 ERROR TECHNICAL SECTOR
    The present invention is framed within the speed measurement systems of 10 motor vehicles or cinemometers, and more specifically, within the point speed measurement systems based on optical or artificial vision systems. BACKGROUND OF THE INVENTION
    15 Depending on the methodology used to measure the speed and depending on the various commercial systems available, kineometers are usually categorized into two main groups: punctual and of average section or speed. On the one hand, point kineometers measure the speed of a motor vehicle from a fixed point. For this, radar-based technologies are used (from the Doppler effect),
    20 based on laser (from the measurement of flight time, for example US6160492) or based on sensors integrated in the road sign of inductive or piezoelectric type. These cinemometers usually carry one or several integrated cameras to be able to take a picture of the vehicle that commits an infraction and even automatically detect the license plate of said vehicle. But nevertheless
    25 These cameras are not used to obtain any type of measurement for vehicle speed detection. On the other hand, the section or medium speed kinemometers (for example, EP 2360647, EP 2220634, US 8189048) are based on the use of two or more cameras located at different points of the same road, with a known separation of the order several hundred meters or kilometers, so
    30 that by identifying the registration code in different locations, the overall capture time of both images and the known distance data, the average speed of the motor vehicle along the section is calculated. For these cases, the greater the length of the section, the smaller the error made in the calculation of the average speed, and the smaller the impact of the error in the distance measurement
    2

    relative of the vehicle to each of the cameras. In practice, for sections greater than
    one kilometer, this distance is not even estimated.
    For what concerns the present invention patent, the most relevant state of the art is directly related to systems based on the use of cameras for the detection of the speed of motor vehicles from a fixed point (not in separate sections by kilometers or hundreds of meters). This idea has been presented generically in, for example, US 20040054513, and WO / 2004/100102, but without taking into account any requirement of precision in the measurement of system speed, that is, without being directly applicable to its use as a precision kineometer. In almost all previous cases the systems have been designed from the use of a camera pointing to a specific region of one or several lanes, calculating the speed from measurements of distance between images without taking into account any requirement related to distances that produce a minimum error in the calculation of the speed. In US-8964031 some type of known feature of the vehicle is used to calculate the relative distance to the cameras. To do this, a calibration is initially carried out and the known distances are stored in a table according to the size in pixels of some characteristic of the vehicle. In normal mode, distances are calculated from the previously calibrated table. In no case is there any requirement among the distance measurements used to calculate the speed. In general, almost all systems require a prior calibration process to be able to calculate the homography from known elements on the road (cones, lines, etc.). Once the system is calibrated, the vehicles are detected in the camera images, usually with background subtraction or optical flow techniques, and their position in the metric space given by the homography is calculated (for example, US 8184863, US 2010 / 0158321). But again, there is no mechanism in place that guarantees that the distance measurements used to calculate the speed of the vehicles meet a minimum error requirement in the speed calculation. Of all existing patents to date, only one that has been generated in a commercial system presented as a precision cinemometer, US 7982634 (KRIA SRL T-EXSPEED system http://www.kria.biz/traffic/texspeed) /). It is a system that combines two infrared sensitive cameras and two RGB cameras, arranged to be used in stereoscopic mode, and in any case pointing to a single region that covers several

    lanes of the track without any restriction that filters the distance measurements between images that do not comply with a geometry of minimum error in the calculation of the speed. In this case, registration location techniques, stereoscopic techniques, structure based on movement, and optical flow are applied to form
    5 combined calculate if there has been a speeding over the measurement section. From the description and claims of US 7982634 it is impossible to draw clear conclusions about the specific method used to calculate the speed and the accuracy obtained.
    10 From the entire state of the known art it is perfectly clear that there is to date no point speed measurement system of motor vehicles from cameras that takes into account the optimum distance between measures of relative distance to the cameras to provide speed measurements with minimal error, which allows the use of technology as a high precision kineometer. EXPLANATION OF THE INVENTION
    The invention relates to a method for the timely measurement, that is, from a fixed point of the speed of motor vehicles that pass through the lanes of a 20-track, by using at least two cameras, located therein. point either on a pole or on a porch, each pointing to two different regions of the same lane. The distance ranges between both regions must meet a specific criterion of minimum speed error. For this, the height, orientation and focal length of the cameras must be configured in a specific way, being those suitable for said geometric restriction that guarantees minimum errors in the calculation of the speed. The geometry of minimum error in speed is achieved in the following way: suppose that one or more cameras take pictures of a vehicle at various moments of time TI and T2 and calculate the relative distance of the vehicle to the camera (s). For the calculation of the relative distance 30 different procedures can be followed (contact point of the wheels with the ground, width of the known car, etc.), but we will consider in this case the calculation of the distance from the location precise of the corners of the matrícula, with previous knowledge of the real dimensions of this. That is, for every moment of time the cameras are able to obtain relative distances
    Z I and Z2, which configure a section of dimension 5 = Z2 -ZI. Speed
    average of the vehicle along said section will be given by the expression:
    being i1T = T2 -TI 'Now, taking into account the location errors or errors in the estimation of the distance Zl err and Z2err, typical of a camera-based system, in the worst case there will be an erroneous estimate of the
    speed v '= (5 + Z2err + Zlerr) / (T2-TI), the error in speed being equal to: 10
    the relative speed error being:
    As an example, we can consider that for a stretch of 5 = lkm, even if there are errors in the estimation of distance of 15m at each of the points, the relative speed error would be 3%. This is one of the reasons why the 20-section kinemometers do not require great precision when estimating the relative distance of the vehicles to the cameras at each of the measurement points, provided that the sections are of the order of several kilometers. The problem comes in cases of timely measurement where the sections are very small. Thus, for example, for a section of 5 = 6 m, distance estimation errors should be less than 9cm in order to obtain a relative speed error of less than 3%. This fact calls into question the accuracy of speed measurement systems based on a single camera (for example, US 8184863, US 201010158321) since the measurement range that a single camera can cover is limited. The most extreme case is that of systems based on instantaneous speed calculation, that is, in the calculation of speed using consecutive images. In these cases, the distance of the section on which it is being measured, which will depend on the
    vehicle speed and the number of images per second that is being taken by the camera, will be a very small distance that in some cases can be of the same order of magnitude as the relative distance estimation errors themselves. These methodologies are, therefore, poorly suited to be used as precision cinemometers.
    In order to define a minimum error geometry, we need to obtain the optimum distance of the section S that provides the minimum error in the estimation of the velocity. For him / her, it is necessary to study the errors of distance estimation. For the specific case of the cameras, it is known that the error grows squared with distance. In this case we start from the calculation of the relative distance using for the / or known dimensions of the license plate, for example, its width. For a coordinate axis with a Z axis parallel to the direction of travel of the vehicle, let X 2 be the X coordinate of, by
    For example, the upper right corner of the license plate, and X the X coordinate of the upper left corner. The width of the license plate in 3D coordinates will be given by M = X2-X¡. After intrinsic camera calibration, the focal length in pixels is known fx = f / dx (where f is the focal length in millimeters, and dx the pixel size of the camera's CCD or CMOS sensor). After applying a registration location method (for example, [H. Bai et al. "A fast license plate extraction method on complex background", Intel IEEE /. Trans. Sys. 2003]), the pixel width of the license plate Llu = u2-u¡. Considering the simple case in which the normal license plate vector moves parallel to the camera's Z axis, the relative distance of the license plate and hence the vehicle, to the camera will be
    Z = fx (Ll X / Ll ul. In the event that the location error of the license plate limits in the image is zero, the image discretization process itself implies a location error that is given by the size of each pixel at the specific distance, that is, Dx = (Z / f) dx = Z / fx. If we take n as the pixel error of locating a point in the image (where n = 1 for the case of zero error), the estimate of the relative distance furthest from the real value will be given by the expression Z '= fx (M + nDx) / Llu. Thus, the relative error in the distance estimate will be:
    If we take into account that the pixel errors in the location of the points in the images are neither Y n2 for the distances ZI and Z2 respectively, the error in the velocity estimation will be given by the following expression:
    This equation can be adapted for the case in which the distances ZI and Z2 are each estimated with a different camera, with focal distances f XI and fX2 'that is:
    If we look at the representation of the error in the velocity estimation using the previous equation, for a hypothetical case in which the distance of the first point is fixed (for example, Z¡ = 3m), for various real velocity values, as a function of the distance of the second point Z2 'the curves that represent the minimum error geometry are obtained. What happens in this geometry for the calculation of the speed and the analysis of its error is that, on the one hand, the speed error decreases linearly with the length of the section (Z2-Z¡). But on the other hand, the error grows squared with the distance of each of the points of the section. For example, in the case of Z¡ = 3m for values of 3m <Z2 <6m the error in speed decreases squared to a minimum from which the error grows almost linearly with the distance Z2 'These curves would describe, by therefore, the geometry of minimum speed error for different speed values. In fact, although the speed error grows linearly with the same speed, the minimum error value (for ZI = 3m, this value is around Z2 = 8m) is independent of it. The optimal distance value can be obtained by deriving the error expression in speed
    with respect to the value of Z2, matching said expression to zero and solving. For that matter
    from a single camera you have that the optimal distance would be:
    s
    In the case of two cameras with different focal lengths, the optimal distance would be:
    10 Since the error in the estimation of distance from cameras decreases
    linearly with the focal length in pixels, it is important that the focal length of
    The cameras for these systems are as high as possible. However, the greater
    focal length, smaller opening angle. This implies that, in practice, it is almost
    impossible to design a high speed motor vehicle measurement system
    lS precision from a single camera pointing to a region, and complying with the
    requirement of optimal distance between measurements for minimization of calculation error
    of speed That is, the maximum region encompassed by a single large chamber
    focal length would not be large enough, and if more focal lengths are used
    low, the error would be increased linearly. The solution raised in this
    twenty invention patent, is based on the use of at least two cameras pointing at two
    different regions of the road with the greatest number of optimal distance pairs
    possible between both regions, that is, distances between both regions that are
    optimal for minimizing the error in speed estimation. Thus,
    each camera can have a focal length as high as possible,
    2S minimizing the error in the calculation of distance and speed.
    The geometry raised between the cameras and the road plan, and used to obtain
    The above equations can be generalized for the case in which the cameras
    are at a higher height and with a specific elevation angle, entering the
    30 rotation and translation matrices that relate the origin of coordinates of the
    cameras with respect to the main road plan. While the expressions are more
    complex, the shape of the curves that define the error in speed are equivalent, so that it is also possible to calculate the optimum point 22 and to be able to configure a geometry of minimum error for the calculation of the speed.
    5 In order to carry out high-precision measurements it is necessary that the cameras be high resolution cameras, and that the optics be as long as possible focal length so that the projection of the vehicle in the image plane is as large as possible, as long as the vehicle is captured and fully visible. The procedure is equally valid for measuring vehicle speeds
    10 from rear elements, (vehicles that move away), which in the case of vehicle speeds from elements obtained from its front (approaching vehicles), because there are structural elements, such as license plates, that appear both in the front and in the rear of the vehicle. The minimum speed geometry is valid regardless of
    15 sense of the progress of vehicles with respect to cameras.
    In low lighting conditions, one or more artificial lighting systems will be activated which can be of visible and / or infrared spectrum to improve the contrast of the images, especially to improve the contrast of the license plates, whose surface 20 is highly reflective. Both cameras and artificial lighting are synchronized thanks to an electronic synchronization card. Both the capture and the processing of the images and their storage and the results obtained are tasks carried out in the processor. The processor adapts to provide the capture order to the synchronism hardware, to capture images
    25 of the cameras, to process them and to store images and results. Each image will have a high precision timestamp thanks to the use of GPS systems with GPSD process.
    In order to solve the problem of distance estimation relative to cameras at
    30 Starting from known structural elements of the vehicle, in this case from the dimensions of the license plate, and taking into account that the cameras have different focal lengths and must be located at different heights and orientations, it is necessary to apply a process previous calibration. This process can be done largely in a laboratory environment to obtain the intrinsic parameters of the
    cameras, as well as their orientation with respect to a horizontal main plane and the extrinsic relationship between them (rotation and translation). For this, the cameras must maintain the same structure in the laboratory and at the actual operating point. In the installation of the system it will only be necessary to provide the final height data at which the cameras are installed and automatically calculate the steering angle from the calculation by means of an optical flow or similar of the main direction of advance of the vehicles. If it is possible to stop the traffic momentarily, it is also possible to use specific calibration patterns such as chessboard, or any other type, to obtain the extrinsic parameters (rotation and translation) of the cameras with respect to the main plane of the track, and always with the steering angle defined by the direction of advance of the vehicles.
    The error in the location of the structure elements of the vehicle in the image, such as in the location of the license plate, leads to errors in the estimation of the speed. The lower the resolution of the image, the greater the mistake is made for each pixel error in the location. That is, the cameras must be high resolution to minimize errors. Also, the procedures for locating structural elements, in this case the registration must be exact and precise. Once the position of the points belonging to the limits of the registration and even the points that define the limits of the alphanumeric characters in the registration is detected, a procedure is applied to estimate the relative distance of the registration to each of the cameras. To do this, the problem can be formulated based on the resolution of a system of equations in which the dimensions in the registration plate of the points belonging to it are known, or by estimating the homography formed between the points of the registration relative to an origin of coordinates located in the registration itself with a Z axis normal to the registration, and said points projected on the image plane. From the homography the rotation and the translation of the origin of coordinates located in the license plate with respect to the cameras can be obtained (see for example, [R. Hartley & A. Zisserman, "Multiple View Geometry in Computer Vis ion", Cambridge University Press 2004]). These values can be used to calculate the relative distance of some point of the license plate with respect to the cameras.
    For each of the images captured by the cameras, the relative distance of the license plate with respect to the position of the cameras is estimated, and a high precision time stamp is associated. For this, the time stamp generated by the processor itself is used, being synchronized with a GPSD (Global Positioning System Daemon) time service from an NTP (Network Time Protocol) server that receives an accurate signal of 1 PPS (1 Press Per Second) of a GPS connected to the processor. All relative distances and their corresponding timestamp are stored to be filtered in a subsequent process. The filtering process makes use of the minimum error geometry. All the relative distance measurements of both cameras that comply with the optimal distance previously calculated are combined to obtain a minimum error in the estimation of the speed of the motor vehicle. In this case, complying with the optimum distance implies that the distance of the motor vehicle or its license plate from the camera that points to the region furthest from the measurement point minus the distance of the motor vehicle or its license plate from the camera at the nearest region of the measurement point, equal to the optimum distance of minimum error plus-minus a percentage that should not be greater than 20%. All measurements that meet this criterion will be used to, in a last phase, estimate the vehicle speed by calculating for this the average of all the speeds calculated from the distance measurements that have passed the minimum error filtering process and their corresponding timestamps. BRIEF DESCRIPTION OF THE DRAWINGS
    Figure 1 shows a side view of the general operating scheme. In this figure the two vehicles (3) correspond to the same vehicle in two different time moments.
    Figure 2 shows a side view of a simplified geometry in which only a camera located at the same height as the vehicle registration itself is used. Similarly, the two vehicles (3) correspond to the same vehicle in two different time periods. This figure is used to illustrate the explanation of the minimum error geometry for speed calculation.
    Figure 3 shows a top view or bird's eye view (in this case, of the XZ plane) of a camera and the license plate of a vehicle. In this case, the detail and nomenclature of the projection are shown in the image of the points that make up the upper left and right corners of the license plate. This figure is used to illustrate the explanation of the minimum error geometry for speed calculation.
    5 Figure 4 shows the representation of the error curves in the speed estimation as a function of the distance at which the second measurement Z2 'is obtained considering that the first measurement Z1 has been taken at 3m with respect to the cameras, for different speeds.
    10 Figure 5 shows the fundamental elements necessary to estimate the speed of motor vehicles.
    Figure 6 shows the general procedure for estimating the speed of motor vehicles.
    Figure 7 illustrates a generic scheme for calculating the relative distance of the license plate to each of the cameras for two different time moments. This figure is used to expose the nomenclature and the elements necessary to expose a specific mode of realization for the calculation of the relative distance of the
    20 vehicle to the cameras, as well as for speed calculation.
    Figure 8 shows, by way of example, the various points of a license plate, which can be used to calculate the relative distance of the license plate from the cameras. PREFERRED EMBODIMENT OF THE INVENTION
    The device necessary to implement the procedure for timely measurement
    (1) speed (2) of motor vehicles (3) in short section (12) with geometry of
    The minimum error consists of at least two digital cameras (4), (5) integrated on the same mechanized structure that keeps the relative position of the cameras fixed, and installed in gantry, post or wall. The cameras can be in color or monochrome, with a minimum resolution of 1280x960, the recommended resolution being at least 1900x1200, high speed, with at least 60 capture capacity
    images per second, with a speed of between 120-160 images per second being recommended. The cameras have built-in optics with a focal length that must be configured so that the relative distance ranges between the regions of the cameras where the motor vehicles 5 captured by them will appear, are at an optimal distance that minimizes the error in Speed calculation. Taking into account that the cameras are located at the same point, either parallel to each other at the same height, or one on top of the other, but oriented differently, and that in the geometry of minimum error there will be a camera (5) pointing at a region further away than the other camera (4), the focal distance 10 of the camera (5) pointing to the furthest region will be greater than that of the camera (4) pointing to the nearest region. A possible installation that meets the minimum error geometry criteria has an optics of 75mm for the camera (5) and another of 50mm for the camera (4), at a height of 4m from the ground and with different orientation angles . The cameras can contain solid state sensors both CCD and CMOS. For low lighting conditions the device incorporates one or several infrared or visible illuminators (13) either in a ring-like structure around the optics of the cameras, or as an independent illuminator of the Raymax or similar type. The illuminators are arranged with a photosensitive cell for automatic ignition. Both infrared or visible illuminators 20 (13) and cameras (4), (5) are synchronized by means of an electronic synchronization card (14). The capture and shooting mode of the cameras is carried out via external shooting by means of a hardware signal, including the exposure time. This signal is provided by the synchronism hardware card (14) and this signal will also be responsible for activating the infrared or visible lighting (13). The
    25 cameras (4), (5) connect to the processor (15) via USB 3.0 type interface,
    o Firewire, or GigE Visiono The processor (15) is also connected to the synchronization card (14) in order to configure both the exposure time of the cameras and the capture speed of the cameras. The connection will be made every second synchronously with a 1PPS signal from a GPS (16)
    30 connected to the processor (15). The GPS (16) supports the NMEA-0183 protocol and through a serial connection with the processor (15) provides a high precision signal of one pulse per second that allows to configure an NTP (Network Time Protocol) server in the processor (15) by correction provided by a GPSD service (GPS daemon). In this way, the timestamp of the processor (15) will be of a level 1 precision (stratum 1). All the images coming from the cameras (4), (5) and captured by the processor (15) will have the precise time stamp in which these images were taken. These timestamps allow you to measure the time between images with a large
    5 precision.
    The installation process involves knowing the minimum error geometry in the speed calculation. For this it is necessary to formulate the error in speed based on the focal distances and the relative distances of the position of the vehicle with respect to
    10 both cameras. For the simplified case in which the optical axis of the cameras is parallel to the road plane and parallel to the direction of travel of the vehicle (see Figure 2) and the distance from the vehicle to the cameras is calculated from the detection of the license plate and its width measurement in pixels (Figure 3) said error would be as follows:
    For the generic case in which the cameras are not oriented in a way
    parallel to the plane of the track (Figure 1), the previous expression would incorporate the
    twenty elements of the rotation and translation matrices of the cameras relative to the plane of
    route. These matrices must be calculated using some calibration process (17),
    which can be carried out in the same application scenario by using
    chessboard type calibration patterns or through a mixed process in which
    the angles of elevation (pitch) and warping or roll (roll) are calculated in the laboratory and
    2S measuring the translation (height) in the application scenario, and obtaining an angle of
    adaptive direction (yaw) according to the passage of each vehicle through flow techniques
    Optical or similar. Once the expression of the velocity error has been defined, the
    first zonemeasurementclosest, camera (4), what definesarank of
    positions (6) for the variable Z l 'Deriving the general expression oferrorin
    30 speed with respect to the position taken Z2 by the second chamber (5) and equalizing
    zero, for the previous values of the variable Z 1the range of positions is obtained (7)
    for the variable Z2 that optimize theerror inSpeed calculation. So by
    For example, for the simplified case (Figure 2) with focal lengths of 50mm for the camera (4) and 75mm for the camera (5), for a range of positions (6) relative to the camera (4) between 2- 4m, the range of optimal positions (7) relative to the camera (5) to obtain measurements with minimum speed error should be between approximately 5-10m (depending on the pixel size of the sensor), following the following expression:
    In order to calculate the distance (20) of the vehicle to the reference point where the cameras are located, several procedures can be chosen. A possible solution is to use structural elements of the vehicle of which variables such as size, relative positions, etc. are previously known. The most standard element is the license plate itself. Both its width (Figure 3) and the relative position of the characters (Figure 8) can be known in advance. In order to obtain these positions, it is necessary to implement a precise registration location mechanism (19), also including the location of the characters. For this there are multiple technologies already available for the location and recognition of license plate characters called LPR (License Plate Recognition Systems) systems. Consider that a registration location mechanism has been implemented (see for example, [H. Bai et al. "A fast license plate extraction method on complex background", IEEE Intell. Trans. Sys. 2003)), and that of, for example, as shown in Figure 7, of the location in the images of the left and right corners of the license plate (either upper or lower) for the two cameras. After the calibration process, rotation matrices and translation vectors for the two chambers with respect to two are known
    systems located in the plane of the track [Rp Ti 1 being the index "i" the indicative of the first chamber or the second chamber ("B" and "A" following the nomenclature of Figure 7). The calibration process is carried out making sure that the rotation between both reference systems on the track is the identity matrix. Following the scheme in Figure 7, we consider that the lower corners of the
    registration of the same vehicle in the image at points p ~ = (u ~, v ~, 1) left and
    p ~ = (u ~, v ~, 1) right, in homogeneous coordinates. We consider that the height of
    these points with respect to the road plan is the same And w¡, and that is also known
    the width of the license plate relative to the coordinate axis placed in the plane of the
    via, thisisL1X = X ~ ¡-X ~ ¡. On the other hand, after the calibration process (17)be
    S know the intrinsic parameters of the cameras including the focal length in the
    two axes, and the optical center, and therefore the intrinsic matrix K can be formed!
    of the camera. Using homogeneous coordinates, the projection of each 3D point
    Pwf relative to the origin of coordinates locatedinthe road plan, inthe plane
    Image is given by the following expression:
    10
    X W¡
    You
    And w¡
    = M¡P w¡
    l w¡
    From this expression the following system of equations can be generated:
    lS sp¡1 = M¡P1 w¡ sp¡2 = M¡P2 ", ¡
    which can be ordered in a linear form of type Ax = b with A being a 5x5 matrix,
    1 2 1 2). TO
    20 b a 5x1 vector, and x = (X wi 'X wi' and wi 'l Wi' lwi. Since the matrix is square
    and has full range, the system has a single solution given by x = A-1 b. To use a single point, the midpoint located in the center of the registration is calculated, that is, the endpoint will be:
    2S
    Finally, the coordinates of the points captured by the two cameras are transferred to the coordinate axis of the cameras, the fixed measurement point, but
    maintaining the relative orientation given by the coordinate axes located in the
    road plan, this is:
    The final distance D between the points corresponds to the final distance traveled by the vehicle (12) and will be given by the expression:
    DJ
    This value is calculated and stored (21) for all possible combinations between images from the first camera (4) and the second camera (5) and will be
    j
    Associates the corresponding timestamp t (22). Subsequently, in a stage of filtering (23) they are calculated for all the relative distances obtained in the first
    chamber (4) P '~ B the relative optimal distances that would be required in the second
    chamber (5) P'cA, that is, D j = 11 p '~ -P' ~ BI I according to the expression of the geometry of minimum speed error. In this filtering stage (23) only those are accepted
    DJ
    distances that are equal to the corresponding optimal distance í) and plus-minus 20 a percentage that should not be greater than 20%. If we consider j = 1. .. N the number of distance measurements that have passed the filtering stage (23), in the final stage (24) the average vehicle speed is estimated as the average of all the speeds that have passed by the filtering stage and that are, therefore, estimates of minimum error rate, that is:
    DJ
    1 N
    V = -L-j -
    N j = 1 tA -tj o
    To check the accuracy of the system, a vehicle is equipped with a high-end differential GPS (high precision and 20Hz frequency), which provides 30 position measurements with errors less than 2.5cm (for a 6m stretch it implies relative speed errors less than 0.83%). Differential GPS data is synchronized from
    the same way from a GPSD service with an NTP client running on a processor shipped in the vehicle. In this way, the absolute position measurements given by the differential GPS can be associated with the measurements obtained from the cameras. An experienced driver is asked to pass four times through the
    5 measuring point, at speeds of {10, 20, 30, 40, 50, 60, 70, 80} kilometers per hour approximately. Speed errors are calculated after applying the procedure described above, and comparing them with the speeds calculated using the high precision differential GPS. The results for all speeds are shown in Table 1.
    TABLE I
    Approximate speed (km / h) Average absolute error (km / h)Standard Deviation (km / h)Maximum error (km / h)
    10 0.900.311.24
    twenty 1.280.271.51
    30 1.760.161.92
    40 1.630.272.01
    fifty 1.660.502.17
    60 1.700.812.62
    70 0.450.170.63
    80 2.130.412.52
    As you can see the maximum errors are always below 3 km / h, an essential requirement to obtain the model exam certificate by the 15 Spanish Center of Metrology. So also the average absolute error is 1.44km / h.
    Industrial application
    The patent object of this invention has its field of application in the industry of
    20 intelligent transport systems, and more specifically, in the companies in charge of the commercialization and maintenance of precision cinemometers for the detection of speed of motor vehicles in different types of roads. The fundamental advantage of the proposed technology is given by the associated cost that is much lower than the relative cost of point kineometers based on
    25 radar or laser.
    权利要求:
    Claims (6)
    [1]
    1. Procedure for the timely detection of motor vehicle speed characterized by:
    5 a. Understand at least two cameras located at the same measurement point, with different focal length, and different orientation with respect to the road plane, to capture images of motor vehicles in at least two different regions of the same lane separated by an optimal distance for Provide minimum error speed measurements.
    10 b. Calculate the optimal distance between these regions by setting the distances of the region closest to the measurement point and deriving the equation of the speed error with respect to the farthest distances, equalizing it to zero and solving for those distances.
    C. Calculate the intrinsic parameters of the cameras, as well as the rotation and
    15 translation of these with respect to the road plan through a calibration process.
    d. Calculate for each image taken by each camera the relative distance of the vehicle to the measurement point from the precise location in the license plate images, prior knowledge of its dimensions, according to
    20 the projection equations of the cameras.
    and. Store for the same vehicle all the relative distances calculated by the two cameras, and associate them with a precise time stamp.
    F. Associate all possible combinations of relative distances between the two
    cameras from all available images and calculate for each combination the distance traveled by the vehicle.
    g. Eliminate those combinations in which the distance traveled by the vehicle moves more than 20% away from the optimum distance required in each case to obtain a minimum error in the calculation of the speed.
    h. Calculate speed for all distance combinations between
    30 images of both cameras that are at the optimum distance plus less than 20%, such as the division between the distance and the time elapsed between each pair of specific images.
    i. Calculate the final speed of the motor vehicle as the average of all previously calculated speeds, that is, as the average of all speed measurements obtained from distances close to the optimum point
    of the curve that relates the error in speed to the distance of the
    section in which it is being measured.
    [2]
    2. Procedure for the timely detection of motor vehicle speed according to claim [1) characterized in that the relative distance of the vehicles to the cameras is calculated from the homography defined by the license plate plane and its characteristic points, that is, the corners that define the limits of the license plate as well as the corners of the regions that frame the alphanumeric characters of the license plate, with respect to the upper left corner of the license plate.
    [3]
    3. Method for the timely detection of speed of motor vehicles according to claims [1) and [2) characterized in that it comprises an artificial lighting system in the infrared or visible spectrum for nighttime visibility conditions, automatically activated by photosensitive cell, and pressed by means of a synchronism hardware at the same speed as the cameras and with a power-up time equal to the exposure time of the cameras.
    [4]
    Four. Method for the timely detection of motor vehicle speed according to claims [1) - [3) characterized in that the cameras can be configured to capture images of more than one lane.
    [5]
    5. Procedure for the timely detection of motor vehicle speed according to claims [1) - [4) characterized in that the cameras can be configured to measure the speed of vehicles in both directions, that is, vehicles that approach the measurement point or are get away from it.
    .....
    ~. -. r
    :
    ·
    ·
    ··
    ·
    EC · ·•.
    ··
    ·
    ·
    . : N
    .. 1i:
     "lJi ·:.
    ·: ·
    ·
    ·
    ·
    ID
    ··
    ·
    ·
    ·:.
    ___! ___ ..L
    .
    . :
    .
    ·: O:.
    ...
    ·: ::
    OR.
    ·
     Figure 1
    ,
    ,
    ,
    ,
    ,
    >
    ,
    ,
    ,
    .-4
    N
    ,
    .....-4 ,
    Figure 2
     (X1, Y, Z) (X2'Y'Z)
    ~~ _____ MUñ ~ _I_a ____ ~~,!
    ,
    · ,,,
    ·· '. ,
    ..
    ,
    ..
    ,
    ,
    ,
    ·
    · • ,,
    ,
    , ,,,,,
    ,
    ,
    ··,. ,,
    ·,. ,
    . ,
    '. · ·,.
    .
    ·,,
    ·· .. ..
    (u, v) ... ,! (u, v)1 ~ ~ 2
    .1 Image Plane
    , 'i
    ~~ c
    Figure 3
    '"
    '"
    '"
    •
    lfI
    ... E •
    ...
    ...
    ...
    .... ... ... or o
    ci Ó d d ci ci
    ('IIWlI) JCWjl
     Figure 4
    16 15
    Processor Hardware O ~ 13
    '. .... 5
    GPS, 1 PPS t ----- t ~
    Synchronism 1 ---'---. 14
    Figure 5
    ~ 5
    I
    19
    o ~ Location
    Location
    '-
    .
    Calibration
    enrollment
    enrollment
    ·
    • L ...-- "'T" "" "---' • L ..-_ ~ _-- '
    · T ......... ...
    .........._
    20:. - _.... L.-_--,
    J-
    Estimate
    .! ...
    distance
    21: r + ____
    ~ "-..- '
    Brand of
    '"I ---: --_ ~: ....... Distances.-¡
    p •
    Time and Times
    ~
    Estimate
     ...
    distance
    twenty-one,-.....
    ~ ...... -
    , .... Distances and Times
    '-...
    '-
    L - + Filtering 14 - Minimum Error
     /L---...,.---..J 
    23 l- / Speed Estimation
    Figure 6
    "T1
    you'
    N
    C
    ...
    ~
    To the
    [-]
    - .J
    Figure 8
    类似技术:
    公开号 | 公开日 | 专利标题
    US9066085B2|2015-06-23|Stereoscopic camera object detection system and method of aligning the same
    US10565867B2|2020-02-18|Detection and documentation of tailgating and speeding violations
    JP5642341B2|2014-12-17|Range measuring device
    EP2467827B1|2013-10-16|A method and system for measuring the speed of a vehicle
    US20160253575A1|2016-09-01|Object Detection Device and Vehicle Using Same
    CN106681353A|2017-05-17|Unmanned aerial vehicle | obstacle avoidance method and system based on binocular vision and optical flow fusion
    EP2588882B1|2018-01-24|Method for producing a digital photo wherein at least some of the pixels comprise position information, and such a digital photo
    CN109359409A|2019-02-19|A kind of vehicle passability detection system of view-based access control model and laser radar sensor
    CN103955920A|2014-07-30|Binocular vision obstacle detection method based on three-dimensional point cloud segmentation
    KR20180123558A|2018-11-16|Method and system for determining a global location of a first landmark
    JP5088401B2|2012-12-05|Road structure measuring method and road surface measuring device
    WO2021004548A1|2021-01-14|Vehicle speed intelligent measurement method based on binocular stereo vision system
    CN106197382B|2018-06-05|A kind of vehicle-mounted single camera target dynamic distance measuring method
    WO2012097077A1|2012-07-19|Mobile mapping system for road inventory
    CN108444449A|2018-08-24|It is a kind of to the object space attitude measurement method with parallel lines feature
    CN202130447U|2012-02-01|Novel lane line deviation detection device
    JP2020085886A|2020-06-04|Vehicle, vehicle positioning system, and method for positioning vehicle
    JP4760274B2|2011-08-31|Map update device
    ES2665939A1|2018-04-30|Procedure for the punctual measurement of speed of motor vehicles in short section with geometry of minimum error |
    JP4595773B2|2010-12-08|Vehicle control device
    CN105424059A|2016-03-23|Wide baseline infrared camera pose estimation method
    CN102981010A|2013-03-20|Method for verifying speed of appropriate vehicle by using camera
    US11210940B2|2021-12-28|Detection and documentation of speeding violations
    Belaroussi et al.2011|Vehicle attitude estimation in adverse weather conditions using a camera, a GPS and a 3D road map
    KR101320337B1|2013-10-29|Estimation system of the position and pose
    同族专利:
    公开号 | 公开日
    ES2665939B2|2018-08-20|
    WO2018078204A1|2018-05-03|
    引用文献:
    公开号 | 申请日 | 公开日 | 申请人 | 专利标题
    ES2137287T3|1993-05-28|1999-12-16|Cit Alcatel|PROCEDURE AND ARRANGEMENT FOR THE UNIVOQUE ASSIGNMENT OF MEASUREMENT RESULTS IN ROLLED TRAFFIC.|
    EP2597632A2|2010-07-20|2013-05-29|Obschestvo S Ogranichennoi Otvetstvennostyu "Tekhnologii Raspoznavaniya"|Method for determining the speed of a vehicle|
    GB2488890A|2011-03-09|2012-09-12|Xerox Corp|Speed enforcement system which triggers higher-accuracy active sensor when lower-accuracy passive sensor detects a speeding vehicle|
    US20140267733A1|2013-03-12|2014-09-18|Xerox Corporation|Single camera video-based speed enforcement system with a secondary auxiliary rgb traffic camera|
    法律状态:
    2018-08-20| FG2A| Definitive protection|Ref document number: 2665939 Country of ref document: ES Kind code of ref document: B2 Effective date: 20180820 |
    优先权:
    申请号 | 申请日 | 专利标题
    ES201600904A|ES2665939B2|2016-10-28|2016-10-28|Procedure for the punctual measurement of speed of motor vehicles in short section with minimum error geometry|ES201600904A| ES2665939B2|2016-10-28|2016-10-28|Procedure for the punctual measurement of speed of motor vehicles in short section with minimum error geometry|
    PCT/ES2017/070711| WO2018078204A1|2016-10-28|2017-10-24|Process for measuring instantaneous speed in motor vehicles|
    [返回顶部]