A vehicle-mounted method for measuring wheel steering angle based on image processing
By using an image processing-based method and an onboard camera to identify graphic patterns on the steering mechanism, combined with Gaussian filtering and the RANSAC algorithm, the problems of increased sensor error and installation difficulties are solved, enabling fast and low-cost wheel steering angle measurement, applicable to various vehicle types.
Patent Information
- Application Number
- CN202310585552.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-23
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-05-23
AI Technical Summary
Existing methods for measuring vehicle steering angle rely on sensors inside the steering column assembly. These methods suffer from mechanical wear leading to increased errors, difficulty in installation and replacement, and high costs, making them unsuitable for sensorless vehicles and impacting driving safety.
An image processing-based method is employed, which uses an onboard camera to identify simple graphic patterns fixed to the steering mechanism. By combining Gaussian filtering, gradient calculation, Hough transform, and RANSAC algorithm, the wheel steering angle is calculated, enabling fast and low-cost measurement.
It enables fast and simple wheel steering angle measurement with high accuracy, does not change the key structure of the vehicle, is applicable to different vehicle types, and the detection markings are easy to relocate, reducing installation and maintenance costs.
Smart Images

Figure CN116793262B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to an image processing-based vehicle-mounted method for measuring wheel steering angle, which belongs to the field of intelligent driving for electric vehicles. Background Technology
[0002] The steering wheel angle is an essential key data point for establishing a vehicle dynamics model and plays a vital role in the safety decision-making and execution of intelligent driving systems. Intelligent driving systems make full use of real-time vehicle steering angle data to achieve closed-loop feedback control and ensure driving safety.
[0003] Currently, vehicle-mounted steering angle measurement mainly relies on incremental photoelectric encoders or angle sensors installed inside the steering column assembly. Although the measurement accuracy is high, the sensor error increases due to mechanical wear and tear during vehicle use. There is a large nonlinear error between the steering column-measured steering angle and the actual wheel steering angle. Furthermore, installation and replacement are difficult and costly. Some steering columns do not have the above-mentioned sensors and cannot measure the steering angle. If retrofitting is required, it may change the critical structure of the vehicle and affect driving safety. Therefore, this method has limitations.
[0004] Therefore, it is indeed necessary to improve existing technologies to address their shortcomings. Summary of the Invention
[0005] This invention provides an image processing-based vehicle-mounted method for measuring wheel steering angle to address the problems existing in the prior art. This method is quick and easy to install and replace, has low cost, and does not alter the key structure of the vehicle.
[0006] The technical solution adopted in this invention includes: an image processing-based vehicle-mounted measurement method for wheel steering angle, wherein a detection marker, i.e., a specified graphic pattern, is fixed to the steering mechanical structure, and an on-board camera is used to acquire images, ensuring that each frame of the image contains at least two complete graphics, with the graphic patterns being simple circles, rectangles, and triangles; the specific steps are as follows:
[0007] Step 1: The vehicle-mounted camera acquires images at different steering angles, records the corresponding actual wheel turning angles, and establishes dataset D;
[0008] Step 2: Image preprocessing, using Gaussian filtering for noise reduction, the mathematical expression is as follows: ,in σ represents the pixel coordinates, σ represents the variance of the Gaussian function, k represents the dimension of the kernel matrix, and a 3×3 Gaussian convolution kernel is used.
[0009] Step 3: Calculate the gradient magnitude and direction using the first-order partial derivative finite difference to extract image edges. The mathematical expression is: , ,in It is the gradient value in the horizontal direction. It is the gradient value in the vertical direction, which is used to determine the image contour edges;
[0010] Step 4: Calculate the area and perimeter of the contour edges from Step 3, filter edges that meet the threshold conditions, and remove noise interference;
[0011] Step 5: Use Hough transform to identify the image category;
[0012] Step 6: Calculate the minimum bounding rectangle: Obtain the maximum and minimum values of the horizontal and vertical coordinates of the edge. , , , area of the circumscribed rectangle Rotate the profile and solve for the new one. , , , Then calculate the area of the circumscribed rectangle. After rotating 360°, take Record the minimum bounding rectangle at this time. , , , ;
[0013] Step 7: If the identification result in Step 5 is a rectangle, use the aspect ratio of the smallest bounding rectangle in Step 6 to further distinguish between rectangles and squares;
[0014] Step 8: Fit the parametric equations using the RANSAC sampling algorithm;
[0015] Step 9: Calculate the coordinates of the center of the smallest bounding rectangle using the results from Step 6. , Calculate the wheel steering angle based on the parametric equations corresponding to the graphic category;
[0016] The specific process for step eight is as follows:
[0017] Step 8.1: Determine the form of the parametric equation as follows: Where deg is the wheel rotation angle. , These are the coordinates of the center of the smallest bounding rectangle. , , , , , The target parameters are to be determined.
[0018] Step 8.2: Randomly select n data points from dataset D, n≥6, to form a minimum subset, and use the least squares method to fit and estimate the target parameters;
[0019] Step 8.3: Using the parameter model obtained in the previous step, traverse each sample in dataset D and record the angle prediction error. The samples constitute a consensus set. , where t is the threshold;
[0020] Step 8.4: Calculation , where N represents the cardinality of the set;
[0021] Step 8.5: If r < T, where T is the threshold, then proceed to step 8.2;
[0022] Step 8.6: Record Then proceed to step 8.2 and repeat the experiment until the specified number of experiments C is reached;
[0023] Step 8.7: Take Time corresponding ,use All elements are re-estimated for model parameters.
[0024] Further, in step 8.5, the threshold is: the minimum number of data points applicable to the model.
[0025] Furthermore, in step 8.6, the number of experiments C is: the number of algorithm iterations.
[0026] The present invention has the following advantages: The vehicle-mounted wheel steering angle measurement method based on image processing is quick and easy to install and replace, has low cost, does not change the key structure of the vehicle, and the detection mark is easy to transfer for different types of vehicles and different measurement needs, and has high measurement accuracy. Attached Figure Description
[0027] Figure 1 A schematic diagram showing the inspection mark fixed to the steering mechanism structure.
[0028] Figure 2 This is a schematic diagram of the vehicle-mounted measurement method for wheel steering angle in image processing according to the present invention. Detailed Implementation
[0029] The invention will now be further described with reference to the accompanying drawings.
[0030] The present invention provides a vehicle-mounted wheel steering angle measurement method based on image processing, which mainly obtains specific graphic poses through image pattern recognition and calculates the wheel steering angle.
[0031] like Figure 1As shown, the detection mark (i.e., the specified graphic style) is fixed to the steering mechanical structure. The image is acquired using an on-board camera. By designing the appropriate size of the detection mark and the camera object distance, it can be ensured that each frame of the image contains at least two complete graphics. Through image recognition, graphic pose calculation and calibration of the mathematical transformation relationship between graphic pose and wheel angle, the actual wheel angle value can be obtained.
[0032] This invention presents an image processing-based vehicle-mounted wheel steering angle measurement method, divided into two stages: a calibration stage and an application stage. The calibration stage obtains the parametric equations of the aforementioned mathematical transformation relationship. The application stage involves actual use, calculating the actual wheel steering angle value based on these parametric equations. To meet real-time requirements and reduce computational load, the graphics are designed as simple circles, rectangles, and triangles. Hough transform is used to identify the graphic category in real time, and the graphic pose is expressed using the minimum bounding rectangle. The pose transformation relationship is determined using multi-sample calibration. The principle is as follows: Figure 2 As shown.
[0033] The present invention relates to an image processing-based vehicle-mounted method for measuring wheel steering angle, the specific steps of which are as follows:
[0034] Step 1: The vehicle-mounted camera acquires images at different steering angles, records the corresponding actual wheel turning angles, and establishes a dataset. ;
[0035] Step 2: Image preprocessing, using Gaussian filtering for noise reduction, the mathematical expression is as follows: ,in These are the pixel coordinates. Let V be the variance of the Gaussian function. The dimension of the kernel matrix is used. Gaussian convolution kernel;
[0036] Step 3: Calculate the gradient magnitude and direction using the first-order partial derivative finite difference to extract image edges. The mathematical expression is: , ,in It is the gradient value in the horizontal direction. It is the gradient value in the vertical direction, which is used to determine the image contour edges;
[0037] Step 4: Calculate the area and perimeter of the contour edges from Step 3, filter edges that meet the threshold conditions, and remove noise interference;
[0038] Step 5: Use Hough transform to identify the image category;
[0039] Step 6: Calculate the minimum bounding rectangle: Obtain the maximum and minimum values of the horizontal and vertical coordinates of the edge. , , , area of the circumscribed rectangle Rotate the profile and solve for the new one. , , , Then calculate the area of the circumscribed rectangle. Rotation After that, take Record the minimum bounding rectangle at this time. , , , ;
[0040] Step 7: If the identification result in Step 5 is a rectangle, use the aspect ratio of the smallest bounding rectangle in Step 6 to further distinguish between rectangles and squares;
[0041] Step 8: Fit the parametric equations using the RANSAC sampling algorithm. The main process is as follows:
[0042] Step 8.1: Determine the form of the parametric equation as follows: ,in For the turning angle of the wheel, , These are the coordinates of the center of the smallest bounding rectangle. , , , , , Here, k represents the target parameter to be determined, and k is the index of the graphic category.
[0043] Step 8.2: From the dataset Random selection The data are used to form the smallest subset, and the target parameters are estimated by fitting the least squares method.
[0044] Step 8.3: Using the parameter model obtained in the previous step, traverse the dataset. For each sample, record the angle prediction error. , forming a consistent set , where t is the threshold;
[0045] Step 8.4: Calculation , where N represents the cardinality of the set;
[0046] Step 8.5: If r < T, where T is the threshold, then proceed to step 8.2;
[0047] Step 8.6: Record Then proceed to step 8.2 and repeat the experiment until the specified number of experiments C is reached;
[0048] Step 8.7: Take Time corresponding ,use All elements are re-estimated for model parameters.
[0049] In step 8.5, the threshold is the minimum number of data points applicable to the model.
[0050] In step 8.6, the number of experiments C is the number of algorithm iterations.
[0051] Step 9: In practical applications, use the results from Step 6 to calculate the coordinates of the center of the smallest bounding rectangle. , The wheel steering angle is calculated based on the parametric equations corresponding to the graphic category.
[0052] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several improvements without departing from the principle of the present invention, and these improvements should also be considered within the scope of protection of the present invention.
Claims
1. A vehicle-mounted method for measuring wheel steering angle based on image processing, characterized in that: The detection markers, which are designated graphic patterns, are fixed to the steering mechanism. Images are acquired using an onboard camera, ensuring that each frame contains at least two complete graphics. The graphic patterns are simple circles, rectangles, and triangles. The specific steps are as follows: Step 1: The vehicle-mounted camera acquires images at different steering angles, records the corresponding actual wheel turning angles, and establishes dataset D; Step 2: Image preprocessing, using Gaussian filtering for noise reduction, the mathematical expression is as follows: ,in σ represents the pixel coordinates, σ represents the variance of the Gaussian function, k represents the dimension of the kernel matrix, and a 3×3 Gaussian convolution kernel is used. Step 3: Calculate the gradient magnitude and direction using the first-order partial derivative finite difference to extract image edges. The mathematical expression is: , ,in It is the gradient value in the horizontal direction. It is the gradient value in the vertical direction, which is used to determine the image contour edges; Step 4: Calculate the area and perimeter of the contour edges from Step 3, filter edges that meet the threshold conditions, and remove noise interference; Step 5: Use Hough transform to identify the image category; Step 6: Calculate the minimum bounding rectangle: Obtain the maximum and minimum values of the horizontal and vertical coordinates of the edge. , , , area of the circumscribed rectangle Rotate the profile and solve for the new one. , , , Then calculate the area of the circumscribed rectangle. After rotating 360°, take Record the minimum bounding rectangle at this time. , , , ; Step 7: If the identification result in Step 5 is a rectangle, use the aspect ratio of the smallest bounding rectangle in Step 6 to further distinguish between rectangles and squares; Step 8: Fit the parametric equations using the RANSAC sampling algorithm; Step 9: Calculate the coordinates of the center of the smallest bounding rectangle using the results from Step 6. , Calculate the wheel steering angle based on the parametric equations corresponding to the graphic category; The specific process for step eight is as follows: Step 8.1: Determine the form of the parametric equation as follows: Where deg is the wheel rotation angle. , These are the coordinates of the center of the smallest bounding rectangle. , , , , , The target parameters are to be determined. Step 8.2: Randomly select n data points from dataset D, n≥6, to form a minimum subset, and use the least squares method to fit and estimate the target parameters; Step 8.3: Using the parameter model obtained in the previous step, traverse each sample in dataset D and record the angle prediction error. The samples constitute a consensus set. , where t is the threshold; Step 8.4: Calculation , where N represents the cardinality of the set; Step 8.5: If r < T, where T is the threshold, then proceed to step 8.2; Step 8.6: Record Then proceed to step 8.2 and repeat the experiment until the specified number of experiments C is reached; Step 8.7: Take Time corresponding ,use All elements are re-estimated for model parameters.
2. The vehicle-mounted wheel steering angle measurement method based on image processing as described in claim 1, characterized in that: In step 8.5, the threshold is the minimum number of data points applicable to the model.
3. The vehicle-mounted wheel steering angle measurement method based on image processing as described in claim 2, characterized in that: In step 8.6, the number of experiments C is the number of algorithm iterations.
Citation Information
Patent Citations
Wheel steering angle acquisition method and device, equipment and storage medium
CN115092254A
Vehicle wheel center position measuring device by image processing
JP2008102063A
