A vehicle chassis comprehensive detection system and method based on a vision detection method

Abstract

The present invention discloses a vehicle chassis comprehensive detection system and method based on a vision detection method. The system includes a Bluetooth sensor and a signal receiver to achieve wheel positioning, an inclination sensor and a weighing instrument to compensate for the angle inclination and center of gravity offset caused by ground inclination, and a vision to detect the parameters of the front wheel alignment angle, camber angle, kingpin inclination angle, and kingpin caster angle of the wheels in a stationary state of the vehicle, and accurate wheel positioning parameters are obtained after correction and compensation. The system of the present invention does not require frame shooting, the wheel positioning measurement can be completed independently, the system structure is simple, the equipment cost is significantly reduced, it can effectively cope with complex environmental factors such as uneven ground and vehicle dynamic inclination, provide a stable and accurate positioning solution, and is widely applicable to most wheel positionings.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle maintenance and detection, and particularly to a vehicle chassis comprehensive detection system and method based on a vision detection method. Background Art

[0002] Wheel alignment is the precise adjustment of a vehicle's wheels so that their angles conform to the specifications of the vehicle manufacturer. This adjustment mainly involves the contact angle of the tires with the ground and the relationship between the individual tires, with the aim of ensuring even tire wear when the vehicle is in motion, guaranteeing handling performance, driving stability, comfort, and at the same time extending the service life of the tires.

[0003] Currently, for wheel alignment methods, there are traditional instruments that use mechanical relationships for judgment and positioning, and there are also methods that combine traditional instruments with electronic devices. For the combination of electronic devices and traditional instruments, a relatively typical one is the three-dimensional detection method. Its structure consists of a crane, multiple front and rear rods, and multiple rods combined with industrial cameras. The cameras are used to take pictures of the tires, and image processing is performed on them. Combining with the weighing instrument on the flipping plate, the angles and the center of gravity position are calculated to determine whether the wheels are abnormal. This method has factors such as a large footprint and complex operation, and a more concise and equally accurate wheel alignment system is needed. Summary of the Invention

[0004] Object of the Invention: Aiming at the above problems, the present invention proposes a vehicle chassis comprehensive detection system and method based on a vision detection method.

[0005] Technical Solution:

[0006] The present invention provides a vehicle chassis comprehensive detection system based on a vision detection method. The system includes: a Bluetooth sensor module, where 4 Bluetooth sensors supporting AoA broadcasting are fixed at the wheel axle center positions, and 2 receivers supporting AoA are deployed at fixed positions around the vehicle for obtaining the coordinates of the wheels;

[0007] An inclination sensor detects the ground tilt angle. The inclination sensor measures the depression angle and the roll angle of each wheel in the vehicle coordinate system and records ; where the depression angle is the front - rear tilt rotating around the Y - axis and the roll angle is the left - right tilt rotating around the X - axis ; correction is performed using a rotation matrix, and then through inverse transformation, it is compensated into the wheel alignment parameters;

[0008] A weighing instrument detects the center - of - gravity offset. There are four such weighing instruments in each flipping plate, evenly distributed at the four corners of the rectangular flipping plate. Compensation is made for the wheel load according to the ground inclination angle to obtain the true pressure , the center of gravity coordinates are calculated based on moment balance; the wheel alignment parameters are compensated according to the offset of the center of gravity coordinates;

[0009] For visual inspection, industrial cameras are used for camera calibration and image preprocessing is performed on the PC side. The wheel pose parameters are obtained through visual inspection;

[0010] For the correction and compensation of wheel alignment parameters, angle correction and compensation are performed on the wheel pose parameters obtained by visual inspection according to the ground tilt angle and the center of gravity offset, and accurate wheel alignment parameters are obtained.

[0011] Preferably, the load on the wheel is compensated according to the ground inclination angle to obtain the true pressure It includes: correcting the data of each weighing instrument:

[0012] , i = 1, 2, 3, 4... 4n, where n is the number of wheels;

[0013] The compensation of the wheel alignment parameters according to the offset of the center of gravity coordinates includes: calculating the center of gravity correction and compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination angle according to the center of gravity coordinates.

[0014] Preferably, the angle correction and compensation of the wheel pose parameters obtained by visual inspection according to the ground tilt angle and the center of gravity offset includes: detecting the ground tilt angle by an inclination sensor; the weighing instrument cooperates with the inclination sensor to correct the center of gravity offset, and calculates the center of gravity correction and compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination angle; according to the wheel pose parameters of the camber angle, toe angle, caster angle, and kingpin inclination angle in the static state detected by visual inspection; the accurate wheel alignment parameters obtained by combining the above three.

[0015] Preferably, the ground tilt angle detected by the inclination sensor needs to be corrected using a rotation matrix, and the rotation matrix is:

[0016] , ;

[0017] where and are the depression angle and side inclination angle measured by the inclination sensor respectively; through Inverse transformation is compensated into the parameters of wheel alignment.

[0018] Preferably, the inclination sensor is installed at the center of the flip plate and is verified by comparison with a spirit level.

[0019] Preferably, the Bluetooth sensor module adopts the Bluetooth 5.1 communication protocol, supports the SPP protocol, and has a baud rate of 0.92 M bps; it adopts a one-master-multi-slave Bluetooth network configuration, where the master devices are the PC and the dual-mode Bluetooth module, which are used to control the slave devices, receive AoA data, execute the MUSIC angle estimation algorithm, and calculate coordinates; the slave devices are the wheel sensors, which are used to broadcast AoA signals and respond to SPP instructions.

[0020] The present invention also provides a comprehensive vehicle chassis detection method based on a vision detection method, including the following steps:

[0021] S1: Install Bluetooth sensors that support AoA broadcasting at the centers of each wheel, and deploy 2 AoA-supported receivers at fixed positions around the vehicle; the PC makes a Bluetooth connection; the PC sends a wheel position positioning operation instruction to perform wheel position positioning.

[0022] S2: The PC sends an instruction to start the inclination sensor and the level gauge, measure the pitch angle and the roll angle, and record ; where the pitch angle is the front-back tilt that rotates around the Y-axis and the roll angle is the left-right tilt that rotates around the X-axis ; use the rotation matrix for correction, and then through inverse transformation to compensate into the wheel positioning parameters;

[0023] S3: The PC sends a weighing instrument start instruction to measure the load on each wheel, compensate the wheel load according to the ground inclination angle to obtain the true pressure and calculate the center of gravity coordinates based on the moment balance; compensate the wheel positioning parameters according to the offset of the center of gravity coordinates;

[0024] S4: The PC starts the industrial camera detection to visually detect the camber angle, toe angle, caster angle, and kingpin inclination angle of the wheels;

[0025] S5: Perform angle correction and compensation on the camber angle, toe angle, caster angle, and kingpin inclination angle detected visually according to the ground tilt angle and the center of gravity offset to obtain accurate wheel positioning parameters.

[0026] Preferably, the specific method of step S1 is as follows:

[0027] S1.1 Align the vehicle body, and then install a Bluetooth sensor with AoA broadcasting support on each wheel;

[0028] S1.2 Adopt a "one-master-multi-slave" Bluetooth network architecture: the PC is used as the master device, and multiple wheel Bluetooth sensors are used as slave devices; the Bluetooth module supports the SPP protocol and sets the baud rate to 0.92 M bps to match the data transmission requirements;

[0029] S1.3 Construct the vehicle body coordinate system, with the track width W and wheelbase L. Set a subordinate range for each wheel. The PC sends signals through the communication channel. After receiving the instructions, the sensor determines the position of the tire center, stores its coordinate values, and then returns them to the main device, the PC, for range matching to determine the position of the tire.

[0030] Preferably, the specific method of step S3 includes:

[0031] S3.1 The PC sends an instruction to start the weighing instrument. Each tipping plate has four weighing instruments, which are distributed at the four corner points of the tipping plate. The weighing instruments measure the weight of each of the four corners of each tire to obtain , where i = 1, 2, 3... 16;

[0032] S3.2 Perform inclination compensation on each to obtain ;

[0033] S3.3 Establish a center of gravity solution equation based on moment balance to calculate the center of gravity coordinates; compensate the wheel alignment parameters according to the offset of the center of gravity coordinates, and calculate the center of gravity correction compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination angle.

[0034] Preferably, the specific method of step S5 is: Detect the ground inclination angle according to the inclination sensor; the weighing instrument cooperates with the inclination sensor to correct the center of gravity offset, and calculate the center of gravity correction compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination angle; according to the wheel alignment parameters of the camber angle, toe angle, caster angle, and kingpin inclination angle detected by visual inspection in the static state; combine the three to obtain accurate wheel alignment parameters.

[0035] The present invention has the following beneficial effects compared with the prior art: The present invention combines an inclination sensor and a level to obtain the ground inclination angle; for the center of gravity offset compensation, since the ground inclination degree is different, each tipping plate corresponds to a tire, and there are four weighing instruments in one tipping plate, which are distributed at the four corners of the tipping plate. Multiple-point compensation for the wheel load is used to obtain a more accurate load-bearing capacity; and then the compensation of the wheel alignment parameters is obtained; Bluetooth positioning is used to lock the wheels, and the visual inspection method is combined with the inclination sensor and the weighing instrument to correct the wheel alignment parameters, making the positioning more accurate; through multi-sensor fusion and image processing technology, efficient and accurate wheel positioning is realized. Compared with the traditional detection equipment and the industrial rack-type measurement method, the present invention simplifies the equipment device, reduces the cost, concentrates the measurement part on the tipping plate to complete independently, and reduces the measurement floor area while ensuring the accuracy. Improvements and optimizations have been made in both R & D costs and usage scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1is a flow chart of the vehicle chassis comprehensive detection method of the present invention;

[0037] Figure 2 It is a schematic diagram of the vehicle wheelbase and track;

[0038] Figure 3 This is a schematic diagram of a Bluetooth positioning wheel;

[0039] Figure 4 It is a top-down view and angle decomposition diagram;

[0040] Figure 5 It is a schematic diagram of the roll angle and angle decomposition;

[0041] Figure 6 This is a diagram illustrating the toe angle;

[0042] Figure 7 It is a diagram illustrating the camber angle;

[0043] Figure 8 It is a diagram illustrating the kingpin inclination angle;

[0044] Figure 9 This is a diagram illustrating the caster angle;

[0045] Figure 10 It is a spatial diagram of toe angle and camber angle;

[0046] Figure 11 It is a spatial diagram of the kingpin inclination angle and the kingpin caster angle;

[0047] Figure 12 It is the camera calibration flow chart;

[0048] Figure 13 It is a physical simulation diagram of the visual inspection part;

[0049] Figure 14 It is the flow chart of pose parameter calculation. DETAILED DESCRIPTION

[0050] Many modifications and changes made by those skilled in the art based on the principles of the present invention belong to the protection scope of the present invention.

[0051] Those skilled in the art can understand that, unless specifically stated otherwise, the singular forms "a", "an", "the", and "said" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the specification of the present invention means the presence of the described features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or their groups. It should be understood that when an element or component is "connected" to another element or component, it can be directly connected to other elements or components, or there may also be intermediate elements or components. The phrase "and / or" used herein includes any and all combinations of one or more of the associated listed items.

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of the present invention.

[0053] As Figure 1 shown is a flowchart of the vehicle chassis comprehensive detection method of the present invention. The embodiments of the present invention provide a vehicle chassis comprehensive detection method based on a vision detection method, including the following steps:

[0054] S1: Install Bluetooth sensors supporting AoA broadcast at the centers of each wheel, and deploy 2 AoA-supporting receivers at fixed positions around the vehicle; the PC end makes a Bluetooth connection; the PC end sends a wheel position positioning operation instruction to perform wheel position positioning.

[0055] Park the vehicle on a horizontal ground, with the vehicle head facing the reference direction (such as a wall marking line), calibrate the vehicle body axis through a laser level to ensure no deflection, and establish a vehicle body coordinate system. Input parameters: such as Figure 2 shown wheelbase W and wheel track L.

[0056] As Figure 3 Use non-metallic fixtures (such as nylon wheel clamps) to fix 4 Bluetooth 5.1+ sensors Nordic nRF5340 supporting AoA broadcast at the wheel axle center positions, and deploy 2 AoA-supporting receivers (such as Nordic nRF52833 + antenna array) at fixed positions around the vehicle, which are 1 meter directly in front of the vehicle head, 0.5 meters in height and 1 meter directly behind the vehicle tail, 0.5 meters in height respectively. Configure Bluetooth, and the PC end makes a Bluetooth connection.

[0057] Adopt a master-slave Bluetooth network configuration. The master device is the PC and the dual-mode Bluetooth module, which is used to control the slave devices, receive AoA data, execute the MUSIC angle estimation algorithm, and calculate the coordinates. The slave devices are wheel Bluetooth sensors, which are used to broadcast AoA signals and respond to SPP commands. The Bluetooth module supports the SPP protocol and sets the baud rate to 0.92 M bps to match the data transmission requirements.

[0058] The following details the AoA-based positioning method:

[0059] Assume that the antenna array spacing is d, the signal wavelength is λ, and the relationship between the phase difference Δϕ of the signals received by adjacent antennas and the angle of arrival θ is:

[0060] ;

[0061] Use the MUSIC algorithm to estimate the angle of arrival. First, construct the covariance matrix R of the received signals:

[0062] ; Perform eigenvalue decomposition on R to separate the signal subspace and the noise subspace; N is the number of signal acquisitions.

[0063] Search for the angle that maximizes the following formula :

[0064] ;

[0065] Coordinate calculation: If the coordinates of receiver 1 are (x 1 , y 1 ), and the measured angle of arrival is ; The coordinates of receiver 2 are (x 2 , y 2 ), and the measured angle of arrival is , then the target coordinates (x, y) satisfy:

[0066] ;

[0067] Solve to get: ;

[0068] After calculating the coordinate values of the four tires, perform the set interval matching on the coordinate values to determine the front and rear wheels and the left and right wheels, and the data displayed on the PC will also be output by tire position.

[0069] S2: The PC sends a command to start the tilt sensor and the level, and measures the pitch angle and the roll angle;

[0070] After the PC sends a detection command, the tilt sensor measures the pitch angle such as Figure 4 and the roll angle such as Figure 5The top angle refers to the front and back tilt when rotating around the Y axis, and the roll angle refers to the left and right tilt when rotating around the X axis.

[0071] The specific measurement process is as follows: the vehicle is parked on the flip board and the body is straightened. The tilt sensor measures the top view and roll angle of each wheel in the vehicle coordinate system and records the , and verify that the level is consistent.

[0072] Since the angle is tilted due to the uneven ground, a rotation matrix is ​​required Make corrections, including ; then through The inverse transformation is compensated into the parameters of the wheel alignment.

[0073] S3: The PC sends a command to start the load cell to measure the load of each wheel, and compensates the wheel load according to the ground inclination to obtain the actual pressure. , calculate the center of gravity coordinates according to the moment balance; compensate the wheel alignment parameters according to the offset of the center of gravity coordinates;

[0074] Each flip board corresponds to a tire. There are four weighing instruments on a flip board, distributed at the four corners of the flip board. Due to the different inclinations of the ground, the following are used: Figure 4 and Figure 5 (Incline force decomposition diagram) Multi-point compensation of wheel load to obtain more accurate load-bearing capacity. The correction of single load cell data is as follows:

[0075] , i=(1, 2, 3, 4…4n, n is the number of wheels); use the force and moment balance equation to solve the center of gravity coordinates .

[0076] (m is the mass of the car, g is the acceleration due to gravity);

[0077] Construct the moment balance equation in the vehicle body coordinate system: Assume that the coordinates of the contact point of each k-th wheel in the global coordinate system are , the total vertical force is: ,but:

[0078] ;

[0079] After obtaining the center of gravity coordinates, the compensation angles of camber angle, toe angle, caster angle, and kingpin inclination angle can be calculated.

[0080] Camber center of gravity correction: ;in It is the difference in suspension compression caused by the center of gravity height. . is the initial camber angle, The camber angle is increased by using a jack, which will be used in step S4.

[0081] Toe-in foot center of gravity correction: ; where is the difference in suspension compression caused by the center of gravity height, . is the initial camber angle, The camber angle is increased by using a jack, which will be used in step S4.

[0082] Caster angle center of gravity correction: ; where is the caster angle change rate, , is the difference in the caster angle change.

[0083] Kingpin inclination angle center of gravity correction: ; where is the kingpin inclination angle sensitivity coefficient, . It can be obtained by querying the manual.

[0084] The above formulas mention which are the pitch stiffness and roll stiffness;

[0085] , where is the spring stiffness, r is the leverage ratio, W is the track width, and L is the wheelbase.

[0086] S4: The industrial camera is started on the PC side for detection, and the camber angle, toe-in angle, caster angle, and kingpin inclination angle of the wheel are visually detected;

[0087] The present invention makes a legend illustration for the toe-in angle such as Figure 6 , the camber angle such as Figure 7 , the kingpin inclination angle such as Figure 8 , and the caster angle such as Figure 9 .

[0088] As Figure 10 shown, assume that the planes xoz, xoy, and yoz are the vehicle body horizontal plane, the vehicle body lateral plane, and the vehicle body longitudinal plane respectively, the plane ABCD is the wheel plane, and OG is the tire rotation axis vector perpendicular to ABCD. The wheel plane ABCD intersects the planes xoz and yoz at the straight lines EF and MN respectively. Let the direction vector of OG be N=(n 1 , n 2 , n 3 ).

[0089] According to the concept of toe angle, the angle between OF and the Z-axis is the toe angle. . Let the unit normal vector of the xoz plane be (0, 1, 0), then we have:

[0090] ;

[0091] Then the direction vectors of OF and the z-axis are respectively (n 3 , 0, n 1 ) and (0, 0, 1). According to the formula for the angle between two vectors in space, we have:

[0092] ;

[0093] That is, the toe angle is:

[0094] ;

[0095] The angle between the plane ABCD and the xoz is the camber angle , Let the normal vectors of the plane ABCD and the xoz be N = (n 1 , n 2 , n 3 ) and (0, 1, 0) respectively, then we can get:

[0096] ;

[0097] That is, the camber angle is:

[0098] ;

[0099] As Figure 11 shown, assume that OG' is the kingpin of the vehicle, and its direction cosine is set as E = (e 1 , e 2 , e 3 ). OE and OF are the projections of OG' on the vehicle body's transverse plane xoy and longitudinal plane yoz respectively. Thus, it can be known that YOE and YOF are the kingpin inclination and the kingpin caster respectively. Let the length of the kingpin axis be a, and the angles between OG' and the x, y, z axes be . From this, we can get:

[0100] ;

[0101] Thus, we can get:

[0102] ;

[0103] After obtaining the operation relationships of these angles in the visual space coordinates, next, we perform the visual inspection of the wheel parameters part. Figure 12It is the camera calibration flow chart for camera calibration. Figure 13 It is the physical simulation diagram of the vision detection part. The present invention adopts a planar calibration method, calibrates the reflection sheet on the wheel, and then performs image acquisition. Calibration board images are collected from multiple angles to reduce the error caused by randomness. First, four images are collected from directly in front of the calibration board, and then the calibration board is deflected by a certain angle from above, below, left, and right respectively, and five pictures are taken from each angle, for a total of 24 calibration images collected. Image acquisition for the toe angle and camber angle experiments: Utilize the self-locking effect of the steering wheel locking rod of the vehicle to lock the steering wheel, rotate the wheel, simulate the vehicle moving forward and backward, and then collect the images of the target board when the wheel rotates through a camera. Image acquisition for the kingpin inclination and caster angle: Fix the vehicle brake pedal to prevent the wheel from rolling when turning, and then turn the vehicle to the left and right by a certain angle respectively through the steering wheel, and during this process, collect images through the camera.

[0104] Make the 24 collected calibration pictures into an image list file lst.yaml in yaml format to facilitate the batch reading operation of images by OpenCV, where the yaml format is the standard format for OpenCV file reading.

[0105] Define relevant variables, and initialize variables such as the number of images, the size of the calibration board (the number of feature points in each image), the internal parameter matrix of the camera, and store the image feature points, and allocate storage space.

[0106] Load the step lst file, read each picture and perform binary processing. After the image is binary processed, detect and extract the feature points in the image, compare the number of detected feature points with the number of feature points in each image during initialization. If they are equal, save the extracted feature points. If the number is smaller than the preset value, do not save, and then extract the next image, and repeat this operation until all image feature points are extracted. After the feature points are successfully extracted, sort the feature points and draw the feature points in order. In this experiment, the feature points are sorted in the order from left to right and from top to bottom.

[0107] After obtaining the world coordinates and image coordinates of the calibration board center, according to the camera calibration parameter solution formula, calculate the internal parameters and distortion parameters, and calculate the calibration reprojection error: After calling the function calibrateCamera to solve the internal parameters and distortion parameters, calculate the projection coordinates of the world coordinates of the center on the image, and then compare the obtained coordinates with the coordinates extracted by the function findCirclesGrid to obtain the mean square error value.

[0108] After completing the camera calibration work, perform image preprocessing on the original image. The specific operations are as follows:

[0109] First, perform a binarization operation on the image. By setting an appropriate threshold and then comparing the grayscale value of each pixel with the set threshold, the part we need can be highlighted from the image. The binarization formula is:

[0110] ;

[0111] where g(x, y) represents the grayscale value of each pixel after the thresholding operation. Pixels with a grayscale value less than the threshold T are regarded as the background, and vice versa as the foreground.

[0112] Then, perform Gaussian filtering. The images collected in this invention are less affected by salt-and-pepper noise. Therefore, Gaussian filtering is used to smooth the images. For a Gaussian filter, the weight corresponding to a pixel is proportional to the distance between the pixel and the central pixel. The formula for a one-dimensional Gaussian function is:

[0113] ;

[0114] The normalization coefficient A is to ensure that the sum of all weights is equal to 1. The symbol value determines the width of the Gaussian function curve. The larger this value, the flatter the function curve. The Gaussian function is a symmetric bell-shaped curve and is very suitable for filtering. The weights corresponding to points farther from the center point are smaller, so the transition between pixels is more gentle. Since the Gaussian filter can be split into two one-dimensional filters, resulting in fewer multiplication operations after splitting, this provides convenience for matrix operations on images. Therefore, when applying a two-dimensional Gaussian filter to an image, first apply a one-dimensional Gaussian filter to the horizontal lines to filter out the high-frequency components in the horizontal direction, and then apply another one-dimensional Gaussian filter to the vertical lines to filter out the high-frequency components in the vertical direction.

[0115] Next, perform image segmentation processing. This invention uses the GrabCut algorithm for static image segmentation. The method steps are as follows:

[0116] The GrabCut algorithm is an algorithm based on a mixture of Gaussian models. The mixture of Gaussian models is:

[0117] ;

[0118] where, refers to the weight of the i-th Gaussian distribution in the mixture model, The sum of is 1, that is , and k is the number of Gaussian distributions in the mixture of Gaussian models. refers to the mean of the i-th Gaussian distribution, representing the central position of this Gaussian distribution. It refers to the covariance matrix of the i-th Gaussian distribution. In the image, a Gaussian mixture model with k Gaussian components is used to model the background and foreground respectively. Thus, each pixel in the image corresponds to a Gaussian component. An energy model for the image is established as E = U + V, where U is the regional term, representing the negative logarithm of the probability that a pixel belongs to the background or foreground, and V is the boundary term, representing the weight of a pixel belonging to the boundary.

[0119] ;

[0120] where represents the estimated class label of the n-th pixel in the image. Here, there are two classes, foreground and background. represents the index of the n-th pixel, indicating its position in the image. represents the true label of the n-th pixel, indicating its actual class.

[0121] In the Gaussian mixture model, , and are respectively used to describe the weight size, mean component, and covariance matrix in each Gaussian component. These three variables together constitute the parameters of the Gaussian function . As long as the values of these 3 variables are obtained, then the corresponding regional term U is obtained. At this time, the probability that each pixel belongs to the foreground or background can be calculated. Then it is to determine the boundary weight V in the energy model.

[0122] ;

[0123] Generally, the Euclidean distance is used to judge the similarity between pixels. The parameter b in the formula is the scaling factor. A lower image contrast will make the Euclidean distance between pixels m and n belonging to different target regions very small, misjudging m and n as belonging to the same target region, while a higher image contrast will make the Euclidean distance between pixels m and n with high similarity relatively large, misjudging m and n as belonging to different target regions. The role of the parameter b is to adjust this misjudgment operation, that is, to increase the difference between pixels when the image has low contrast through a larger b value, and to reduce the difference between pixels when the image has high contrast through a smaller b value. In this way, whether in high-contrast regions or low-contrast regions, the V term can work properly. The constant term g takes an empirical value of 50, and finally the boundary weight V is determined. The specific steps are as follows:

[0124] First, mark the area to be processed on the image. The pixels outside the marked area are background pixels, that is, the pixel label , and all the pixels within the marked area are initially set as pixels that may be foreground, with the pixel label The marked pixels are divided into k classes by the k-means algorithm for initializing k Gaussian components. For each pixel assigned to a Gaussian component, there is a corresponding Gaussian component corresponding to it, so there is:

[0125] ;

[0126] For the given image data, the GMM parameters are obtained. Its parameter means and covariance can be estimated from the RGB values of the pixel samples obtained by marking. Then, divide the number of pixels corresponding to this Gaussian component by the total number of pixels, and use the obtained value to calculate the weight corresponding to each Gaussian component.

[0127] ;

[0128] According to the energy distribution formula E and the method described above, the corresponding regional weights and boundary weights are obtained, and the image is segmented and estimated; the above operations are repeated to learn and optimize the GMM model parameters to obtain the final segmentation result, and the segmented image is smoothed.

[0129] Finally, the features of the image are extracted. In the present invention, the Hough transform is used for feature extraction. The purpose of the Hough transform is to seek a mapping relationship from the image region to the parametric equation so that the specific curves in the image region can be described by the parametric equation. For the pixel region, we establish a parametric equation with x, y, r (x and y are the center coordinates, and r is the radius) as parameters:

[0130] ;

[0131] We can regard the target image region as a set composed of multiple circular regions. Then, for each circle boundary, there is a corresponding equation with a specific radius in the parameter space. The Hough transform is to make each pixel in the target region correspond one by one in the parameter space. The boundary points of the circular region are composed of all pixel points with non-zero intensity values in the figure. If a pixel point (x, y) moves around the circular boundary determined by the characteristic parameters, then there will be multiple rings determined by different parameters intersecting at a point (a, b), and (a, b) is the center coordinate of a specific circular region. At this time, the calculation of the center becomes a problem of seeking the intersection of circles with different parameters. It can be seen that its basic strategy is: find the trajectory of specific points in the parameter space through the pixel points that may be boundary points, and count the obtained specific points through a counter. If the value is larger than the pre-set value, then the circle obtained at this time is the circle that meets the requirements. The steps for extracting the center coordinates are as follows:

[0132] Determine the set of analytical expressions of the curves to be recognized and establish a parametric equation;

[0133] Construct a counter N[a][b] for each parameter in the analytical expression set and initialize the value of the counter;

[0134] Traverse the valid pixels of the image and substitute the pixel coordinates in the image into the parametric equation. If it satisfies:

[0135] ; then the corresponding counter value is increased by 1; in the counter, find the parameter (a, b) corresponding to the maximum value, (a, b) is the coordinate of the center of the circle we want to obtain.

[0136] After the image is preprocessed, the image is combined with the obtained coordinates to solve the pose parameters. Figure 14 As shown, the wheel posture parameters are finally output.

[0137] Finally, the wheel alignment parameters are calculated, and the corresponding wheel alignment parameters are obtained according to the wheel alignment parameter calculation formula described earlier in this article.

[0138] Pseudo code for obtaining pose parameters:

[0139] Define variables to store 2D image coordinates A1, feature point camera coordinates A2, and world coordinates A3;

[0140] A1 and A2 are initialized to 0, and A3 is initialized according to the camera calibration method;

[0141] Extract 2D image coordinates;

[0142] Substitute the camera internal parameters to obtain the camera coordinates A2 of the two-dimensional image coordinates;

[0143] Using pose parameters as variables , combine A2 and A3 to establish the objective function:

[0144] ;

[0145] Before obtaining the extreme value of the objective equation, perform the following operations:

[0146] Least squares expansion of objective function;

[0147] Use SVD to calculate the corresponding location parameters ;

[0148] Pose parameters Substitute into the objective function.

[0149] There may be certain errors in the visual detection algorithm. The present invention adopts the weighting method to correct the angle and calculates the weight using the root mean square error. The specific process is as follows:

[0150] For each frame of image, we set m feature points, and the reprojection error of each feature point can be written as:

[0151] ;

[0152] where is the actual observed position of the feature point in the image coordinates, is the theoretical position projected after pose calculation, and i represents the i-th wheel.

[0153] The calculation formula of the mean square error is as follows:

[0154] ;

[0155] Regarding RMS as the standard deviation of this frame , the corresponding variance estimate is: ; m represents the number of feature points.

[0156] Thus, the weight of the j-th frame of each wheel is: ;

[0157] Finally, normalize the weights, and then perform a weighted average on the detected angles to obtain the final detected angle value: ;

[0158] The weighted average method, using the reciprocal of the mean square error as the weight, can solve problems such as the image shooting angle, pixel blurring of a certain frame, and light. It assigns better weights to images with good quality and relatively lower weights to images with relatively weaker effects. After processing by combining multiple frames of images, the errors existing in visual detection can be eliminated.

[0159] S5: According to the ground tilt angle and the center of gravity offset, perform angle correction and compensation on the camber angle, toe angle, caster angle, and kingpin inclination detected by visual inspection to obtain accurate wheel alignment parameters.

[0160] Detect the ground tilt angle according to the tilt sensor; the weighing instrument cooperates with the tilt sensor to perform center of gravity offset correction, and deduce the center of gravity correction and compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination; according to the wheel pose parameters of the camber angle, toe angle, caster angle, and kingpin inclination in the static state detected by visual inspection; the accurate wheel alignment parameters obtained by combining the three.

[0161] Considering the inclination of the ground unevenness and the change of the inclination angle caused by the center of gravity offset, for the camber angle detected by visual inspection 、toe angle 、caster angle 、kingpin inclination Perform ground inclination compensation and center-of-gravity offset inclination correction. The ground inclination compensation and center-of-gravity offset inclination correction angles have been described above. The angle after camber angle compensation correction is , the angle after toe angle compensation correction is , the angle after kingpin inclination angle compensation correction is , the angle after caster angle compensation correction is . Now, the formulas for each angle are listed comprehensively:

[0162] Calculation formula after camber angle compensation correction:

[0163] ;

[0164] ;

[0165] Calculation formula after toe angle compensation correction:

[0166] ;

[0167] Calculation formula after kingpin inclination angle compensation correction:

[0168] ;

[0169] Calculation formula after caster angle compensation correction:

[0170] .

[0171] Another embodiment of the present invention provides a vehicle chassis comprehensive detection system based on a vision detection method. The system includes: a Bluetooth sensor module, where 4 Bluetooth sensors supporting AoA broadcast are fixed at the wheel axle center positions, and 2 receivers supporting AoA are deployed at fixed positions around the vehicle for obtaining the coordinates of the wheels;

[0172] An inclination sensor detects the ground inclination angle. The inclination sensor measures the depression angle and roll angle of each wheel in the vehicle coordinate system and records ; where the depression angle is the front-back inclination rotating around the Y axis , and the roll angle is the left-right inclination rotating around the X axis ; use a rotation matrix for correction, , and then compensate it into the wheel alignment parameters through inverse transformation;

[0173] A weighing instrument detects the center-of-gravity offset. There are four such weighing instruments in each tipping plate, evenly distributed at the four corners of the rectangular tipping plate. Compensate the wheel load according to the ground inclination to obtain the true pressure , and calculate the center-of-gravity coordinates based on the moment balance; compensate the wheel alignment parameters according to the offset of the center-of-gravity coordinates;

[0174] Visual inspection uses an industrial camera for camera calibration and the PC side for image preprocessing, and obtains the wheel pose parameters through visual inspection;

[0175] Wheel alignment parameter correction and compensation correct and compensate the wheel pose parameters obtained by visual inspection according to the ground tilt angle and the center of gravity offset to obtain accurate wheel alignment parameters.

[0176] Furthermore, compensating the wheel load according to the ground inclination angle to obtain the true pressure includes: correcting the data of each weighing instrument:

[0177] , i = 1, 2, 3, 4…4n, n is the number of wheels;

[0178] The compensation of the wheel alignment parameters according to the offset of the center of gravity coordinates includes: calculating the center of gravity correction and compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination angle according to the center of gravity coordinates.

[0179] Furthermore, the angle correction and compensation of the wheel pose parameters obtained by visual inspection according to the ground tilt angle and the center of gravity offset includes: detecting the ground tilt angle by an inclination sensor; the weighing instrument cooperates with the inclination sensor to correct the center of gravity offset, and calculates the center of gravity correction and compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination angle; according to the wheel pose parameters of the camber angle, toe angle, caster angle, and kingpin inclination angle in the static state detected by visual inspection; the combination of the three obtains accurate wheel alignment parameters.

[0180] Furthermore, the ground tilt angle detected by the inclination sensor needs to be corrected using a rotation matrix, and the rotation matrix is:

[0181] ;

[0182] where are respectively the depression angle and the side inclination angle measured by the inclination sensor; through Inverse transformation is compensated into the parameters of wheel alignment.

[0183] Furthermore, the inclination sensor is installed at the center of the flip plate and verified by comparison with a spirit level.

[0184] Furthermore, the Bluetooth sensor module uses the Bluetooth 5.1 communication protocol, supports the SPP protocol, and the baud rate is 0.92 M bps; adopts a one-master multi-slave Bluetooth network configuration, the master device is the PC side, which realizes controlling the slave device, receiving AoA data, executing the MUSIC angle estimation algorithm, and calculating coordinates; the slave device is the wheel Bluetooth sensor, which realizes broadcasting the AoA signal and responding to the SPP instruction.

[0185] The applicant of the present invention has made a detailed description and illustration of the embodiments of the present invention in conjunction with the accompanying drawings. However, those skilled in the art should understand that the above embodiments are only the preferred implementation schemes of the present invention, and the detailed description is only to help readers better understand the spirit of the present invention, rather than a limitation on the protection scope of the present invention. On the contrary, any improvement or modification made based on the spirit of the present invention should fall within the protection scope of the present invention.

Claims

1. A vehicle chassis comprehensive detection system based on visual detection method, characterized in that: The system includes: a Bluetooth sensor module, wherein four Bluetooth sensors supporting AoA broadcasting are fixed at the axle center of the wheel, and two receivers supporting AoA are deployed at fixed positions around the vehicle to obtain the coordinates of the wheel; The tilt sensor detects the ground tilt angle. The tilt sensor measures the top view and roll angle of each wheel in the vehicle coordinate system and records ; The weighing instrument detects the center of gravity deviation. There are four weighing instruments in each flip board, evenly distributed in the four corners of the rectangular flip board. The wheel load is compensated according to the ground inclination to obtain the real pressure. , calculate the center of gravity coordinates according to the moment balance; compensate the wheel alignment parameters according to the offset of the center of gravity coordinates; Visual inspection uses industrial cameras for camera calibration and PC for image preprocessing to obtain wheel posture parameters through visual inspection. Wheel alignment parameter correction and compensation: angle correction and compensation are performed on the wheel posture parameters obtained by visual inspection according to the ground inclination angle and center of gravity offset to obtain accurate wheel alignment parameters.

2. A vehicle chassis comprehensive detection system based on a visual detection method according to claim 1, characterized in that: The wheel load is compensated according to the ground inclination to obtain the real pressure Including: correcting the data of each weighing instrument, , i=1, 2, 3, 4…4n, n is the number of wheels; The compensation of wheel alignment parameters according to the offset of the center of gravity coordinates includes: calculating the center of gravity correction compensation angles of the camber angle, toe angle, caster angle and kingpin inclination angle according to the center of gravity coordinates.

3. A vehicle chassis comprehensive detection system based on a visual detection method according to claim 2, characterized in that: The method of performing angle correction and compensation on wheel posture parameters obtained by visual detection according to the ground inclination angle and the center of gravity offset comprises: detecting the ground inclination angle according to the inclination sensor; performing center of gravity offset correction on the load cell in cooperation with the inclination sensor to deduce the center of gravity correction compensation angle of the camber angle, the toe angle, the kingpin castor angle and the kingpin inclination angle; performing angle correction and compensation on the wheel posture parameters in a static state according to visual detection of the camber angle, the toe angle, the kingpin castor angle and the kingpin inclination angle; and performing angle correction and compensation on the wheel posture parameters in a static state according to the ground inclination angle and the center of gravity correction compensation angle to obtain accurate wheel alignment parameters.

4. A vehicle chassis comprehensive detection system based on a visual detection method according to claim 3, characterized in that: The ground inclination angle detected by the inclination sensor needs to be corrected using a rotation matrix, which is: ， ； in and are respectively the top angle and roll angle measured by the tilt sensor; The inverse transformation is compensated into the parameters of the wheel alignment.

5. The vehicle chassis comprehensive detection system based on the visual detection method according to claim 4 is characterized in that: The inclination sensor is installed at the center of the flip plate and is compared and verified using a level meter.

6. A vehicle chassis comprehensive detection system based on a visual detection method according to claim 5, characterized in that: The Bluetooth sensor module adopts Bluetooth 5.1 communication protocol, supports SPP protocol, and has a baud rate of 0.92 M bps. It adopts a one-master-multiple-slave Bluetooth network configuration, with the master device being the PC end, which controls the slave device, receives AoA data, executes the MUSIC angle estimation algorithm, and calculates coordinates. The slave device is the wheel Bluetooth sensor, which broadcasts AoA signals and responds to SPP instructions.

7. A vehicle chassis comprehensive detection method based on a visual detection method, using a vehicle chassis comprehensive detection system based on a visual detection method according to any one of claims 1 to 6, characterized in that: The steps include: S1: A Bluetooth sensor supporting AoA broadcast is installed at the center of each wheel, and two AoA-supported receivers are deployed at fixed positions around the vehicle; the PC is connected via Bluetooth; the PC sends a wheel position positioning operation instruction to perform wheel position positioning; S2: The PC sends a command to start the inclination sensor and level to measure the pitch angle and roll angle. The pitch angle is the forward and backward tilt around the Y axis. The roll angle is the left-right tilt around the X axis. ; S3: The PC sends a command to start the load cell to measure the load of each wheel, and compensates the wheel load according to the ground inclination to obtain the actual pressure. , calculate the center of gravity coordinates according to the moment balance; compensate the wheel alignment parameters according to the offset of the center of gravity coordinates; S4: PC starts industrial camera detection to visually detect the camber angle, toe angle, caster angle, and kingpin inclination angle of the wheel; S5: According to the ground inclination angle and the center of gravity offset, the visually detected camber angle, toe angle, caster angle and kingpin inclination angle are corrected and compensated to obtain accurate wheel alignment parameters.

8. The vehicle chassis comprehensive detection method based on visual detection method according to claim 7 is characterized in that: The specific method of step S1 is as follows: S1.1 puts the car body in a straight position and then installs a Bluetooth sensor with AoA broadcast support on each wheel; S1.2 adopts the "one master and multiple slaves" Bluetooth network architecture: the PC is the master device and multiple Bluetooth sensors are slave devices; the Bluetooth module supports the SPP protocol and sets the baud rate to 0.92 Mbps to match the data transmission requirements; S1.3 builds the vehicle body coordinate system and sets a wheel slave range. The PC sends a signal through the communication channel. After receiving the command, the Bluetooth sensor determines the center position of the tire, stores and accesses its coordinate value, and returns it to the main device PC for range matching to determine the position of the tire.

9. The vehicle chassis comprehensive detection method based on visual detection method according to claim 8, characterized in that: The specific method of step S3 includes: S3.1 The PC sends a start command to the weighing instrument. Each flip board has four weighing instruments, which are distributed at the four corners of the flip board. The weighing instruments measure the weight of each tire at the four corners. , where i=1, 2, 3…16; S3.2 For each Perform tilt compensation and obtain ; S3.3 The center of gravity solution equation is established based on the moment balance to deduce the center of gravity coordinates; the wheel alignment parameters are compensated based on the offset of the center of gravity coordinates to deduce the center of gravity correction compensation angles of the camber angle, toe angle, caster angle, and kingpin inclination angle.

10. A vehicle chassis comprehensive detection method based on a visual detection method according to claim 9, characterized in that: The specific method of step S5 is: detecting the ground inclination angle according to the inclination sensor; performing gravity center offset correction with the load cell in cooperation with the inclination sensor, and calculating the gravity center correction compensation angle of the camber angle, toe angle, kingpin castor angle, and kingpin inclination angle; detecting the wheel posture parameters of the camber angle, toe angle, kingpin castor angle, and kingpin inclination angle in a static state according to visual detection; performing angle correction compensation on the wheel posture parameters in a static state according to the ground inclination angle and the gravity center correction compensation angle, and obtaining accurate wheel alignment parameters.

Images