Tire pressure calibration method based on image analysis

Abstract

The invention relates to the technical field of tire monitoring, and discloses a tire pressure calibration method based on image analysis, which comprises the following steps: step 1, acquiring a vehicle tire image; 2, preprocessing the vehicle tire image; step 3, carrying out contour identification and positioning on the vehicle tire image; 4, detecting and extracting tire contour data; 5, building a finite element model of the vehicle tire, and calibrating the tire pressure of the tire; according to the method, on the basis of not adding tire end electronic products, the tire image data of different shooting angles and different tire pressures are obtained, the outer contour of the tire is identified by using an image analysis method, and then the tire pressure is calibrated and calculated based on the contour data, so that tire pressure calibration based on image processing can be realized; powerful guarantee is provided for safe driving of the vehicle; the problem that in the prior art, a digital image processing technology cannot be effectively integrated into vehicle tire pressure calibration, and the tire pressure of a special vehicle cannot be calibrated and monitored is solved.

Description

Technical Field

[0001] The invention belongs to the technical field of tire monitoring, and in particular relates to a tire pressure calibration method based on image analysis. Background Art

[0002] With the leapfrog development of my country's national defense strength, various types of special vehicles have become an indispensable part of the equipment of various military services. These special vehicles undertake important tasks in complex battlefield environments, so their safety issues are particularly important. Among the various safety hazards of special vehicles, tire safety issues account for a relatively large proportion. Among them, tire blowout is a major cause of tire-induced accidents, which may cause the vehicle to lose control and even cause serious accidents.

[0003] In addition to the external factor of poor road conditions, abnormal tire pressure of special vehicle tires is the main cause of accidents such as tire blowouts. Too high or too low tire pressure may have an adverse effect on the performance and service life of the tire, thereby threatening the safe driving of special vehicles. Therefore, real-time monitoring and calibration of tire pressure of special vehicles is a key link to ensure the safe driving of vehicles. Since tire blowout accidents are accidental and unpredictable, once a tire blowout occurs, the driver often panics due to lack of experience and makes operating errors, making the consequences of the accident disastrous. Relevant experts remind that when the car is traveling at a speed of more than 100km / h, the mortality rate of a tire blowout accident is close to 100%.

[0004] Digital image processing is a technology that uses computer programs to process images to meet human visual senses or application requirements. The earliest purpose of digital image processing technology was to improve the image information used in human analysis. With the development of computer technology and artificial intelligence technology, its main purpose has evolved into processing image data obtained by automatic devices, namely computer vision and pattern recognition. Therefore, how to effectively integrate digital image processing into the tire pressure calibration of special vehicles and effectively calibrate and monitor the tire pressure of special vehicles has become an urgent problem to be solved in this field. Summary of the invention

[0005] 1. Technical issues to be solved

[0006] In view of the shortcomings of the prior art, the present invention provides a tire pressure calibration method based on image analysis. This method, without adding electronic products on the tire side, obtains tire image data of different shooting angles and different tire pressures, uses image analysis methods to identify the outer contour of the tire, and then calibrates and calculates the tire pressure based on the contour data. This can realize tire pressure calibration based on image processing, and provide strong guarantee for the safe driving of special vehicles. It solves the problem that the prior art cannot effectively integrate digital image processing technology into vehicle tire pressure calibration, and effectively calibrate and monitor the vehicle tire pressure.

[0007] (II) Technical solution

[0008] In order to achieve the above-mentioned purpose of calibrating and monitoring the tire pressure of a vehicle by using digital image processing technology, the present invention provides the following technical solutions:

[0009] A tire pressure calibration method based on image analysis, comprising:

[0010] Step 1: Acquire vehicle tire image;

[0011] Step 2: preprocessing the vehicle tire image obtained in step 1 based on an image recognition algorithm;

[0012] Step 3: Using a tire contour extraction algorithm based on morphological image processing to identify and locate the tire contour of the vehicle tire image preprocessed in step 2;

[0013] Step 4: Perform ground line detection on the tire profile extracted in step 3 and extract tire profile data;

[0014] Step 5: Use the tire profile data extracted in step 4 to build a finite element model of the vehicle tire and calibrate the tire pressure.

[0015] According to a specific implementation of the embodiment of the present disclosure, the process of preprocessing the vehicle tire image in step 2 includes:

[0016] Step 2.1: Perform geometric transformation processing on the special vehicle tire image obtained in step 1;

[0017] Step 2.2: grayscale the image processed by the geometric transformation in step 2.1;

[0018] Step 2.3: Use Gaussian filtering to perform denoising filtering on the image after grayscale processing in step 2.2;

[0019] Step 2.4: Use the OTSU algorithm to binarize the image after denoising and filtering in step 2.3;

[0020] Step 2.5: Perform edge detection and analysis on the image binarized in step 2.4.

[0021] According to a specific implementation of the embodiment of the present disclosure, the process of performing geometric transformation processing on the vehicle tire image in step 2.1 includes:

[0022] Step 2.1.1: Taking the origin of the vehicle tire image obtained in step 1 as the center, the image is rotated. The process includes:

[0023] Assume |OP0|=|OP1|=L, then

[0024] cosα=x0 / L,sinα=y0 / L

[0025] sin(α+θ)=y1 / L=cosθsinα+cosαsinθ

[0026] cos(α+θ)=x1 / L=cosθcosα-sinαsinθ

[0027] have

[0028] x1=cosθx0-sinθy0

[0029] y1=cosθy0-sinθx0

[0030] The rotation transformation matrix is

[0031]

[0032] The corresponding inverse operation is

[0033]

[0034] Step 2.1.2: Correct the rotated image using perspective transformation

[0035] The perspective transformation matrix transformation formula is:

[0036]

[0037] According to a specific implementation of the embodiment of the present disclosure, the process of graying the image in step 2.2 includes:

[0038] Step 2.21: Use the weighted average method to grayscale the image;

[0039] Step 2.22: Perform grayscale conversion on the grayscaled image, including

[0040] The gray value s(x,y) of the original image is mapped to a new gray value d(x,y) through the transformation function T, d(x,y) = T[s(x,y)], and the gray value d(x,y) is a piecewise linear transformation function;

[0041]

[0042] Step 2.23: Enhance the tire image after the grayscale transformation in step 2.22.

[0043] According to a specific implementation of the embodiment of the present disclosure, the process of performing edge detection and analysis on the image in step 2.5 includes:

[0044] Step 2.5.1: Filter and denoise the image and calculate the amplitude and direction by difference. The gradient amplitude is:

[0045] The gradient direction is:

[0046] Among them, G x is the horizontal gradient operator, G y is the vertical gradient operator;

[0047] Step 2.5.2: Perform non-maximum suppression on the image and locate the edge information of the image by combining the gradient direction of the detection point;

[0048] Step 2.5.3: Perform hysteresis threshold processing on the image to obtain a closed curve of the edge of the entire image.

[0049] According to a specific implementation of the embodiment of the present disclosure, the process of step 3 of identifying and locating the tire contour of the image includes:

[0050] Step 3.1: Use the tire contour extraction algorithm of morphological image processing to perform morphological closing operation on the tire edge image to connect the discontinuous edges, including

[0051] Step 3.1.1: Use the structure element SE to perform a dilation operation on the tire edge image;

[0052] Step 3.1.2: Use the structure element SE to perform corrosion operation on the expansion result of the tire image;

[0053] Step 3.2: Use an iterative dilation algorithm to fill holes in the edge image.

[0054] According to a specific implementation of the embodiment of the present disclosure, the process of performing ground line detection on the tire contour and extracting tire contour data in step 4 includes:

[0055] Step 4.1: Measuring the tire contact line and tire radius of the tire profile based on the detection algorithm of Hough transform, including a straight line detection process and a curve detection process;

[0056] Step 4.2: Extract, detect and analyze the features of the vehicle tire profile to detect the tire contact line and tire radius of the tire profile.

[0057] According to a specific implementation of the embodiment of the present disclosure, step 5 uses ABAQUS software to construct a finite element model of a vehicle tire, including:

[0058] Step 5.1: Perform load characteristic analysis of a heavy-duty tire fine parameterized finite element model;

[0059] Step 5.2: Establish a tire pressure calibration model.

[0060] According to a specific implementation of the embodiment of the present disclosure, in the tire pressure calibration model, the direct method and the sine value method are used to calculate the vehicle tire overload condition.

[0061] (III) Beneficial effects

[0062] Compared with the prior art, the present invention provides a tire pressure calibration method based on image analysis, which has the following beneficial effects:

[0063] 1. This tire pressure calibration method can complete the tire pressure calibration of the vehicle without adding electronic products on the tire side. It can provide strong guarantee for the safe driving of the vehicle by obtaining tire image data of different shooting angles and different tire pressures, preprocessing the tire image, identifying and locating the tire contour, detecting the tire grounding line, and building a finite element model of the vehicle tire to calibrate the tire pressure.

[0064] 2. This method uses a preprocessing algorithm for special vehicle tire images based on image recognition to perform geometric transformation, grayscale, denoising filtering, binarization, edge detection and analysis on the vehicle tire image, eliminating background noise and other irrelevant information in the image, ensuring that only the tire part is retained as the object of further processing, improving the accuracy and reliability of the measurement results, and providing a strong guarantee for the safe operation of the vehicle;

[0065] 3. This method uses a tire contour extraction algorithm based on morphological image processing to identify and locate the tire contour of the preprocessed image, shields other parts of the image except the required target area, accurately identifies the outer contour of the tire, and obtains a circular or quasi-circular binary image, which provides a basic object for the next step of extracting the tire's feature information;

[0066] 4. This method uses Hough transform to detect the tire contour for grounding to identify the morphological changes of the tire, thereby determining its tire pressure and load status; at the same time, other parameters such as the tire radius are extracted to provide the necessary conditions for the subsequent tire pressure calibration algorithm, and improve the accuracy of tire pressure calibration, and improve the accuracy and reliability of the measurement results;

[0067] 5. This method constructs a finite element model of the vehicle tire, and uses the finite element model to perform fine parameterized finite element model load characteristic analysis of the heavy-load tire and establish a tire pressure calibration model. It uses the direct method and the sine value method to calculate and detect the heavy-load condition of the vehicle tire, and can accurately calibrate the tire pressure model; it achieves the purpose of quickly and accurately calibrating and monitoring the tire pressure of special vehicles using digital image processing technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 The algorithm flow chart of the special vehicle tire for image recognition of the present invention;

[0069] Figure 2 This is a complete front view of the tire of the present invention;

[0070] Figure 3 The rotation transformation coordinate diagram of the present invention;

[0071] Figure 4 This is a comparison diagram of the rotation effect of the present invention;

[0072] Figure 5 A comparison diagram of the tire image after perspective transformation of the present invention;

[0073] Figure 6 is the weighted average image histogram of the tire of the present invention;

[0074] Figure 7 It is a schematic diagram of the piecewise linear transformation of the present invention;

[0075] Figure 8 The original image and the image after grayscale conversion of the present invention;

[0076] Fig. 9 The original image of the present invention and the histogram after grayscale transformation;

[0077] Fig.10 The original image and the image after image enhancement of the present invention;

[0078] Fig.11 This is the tire Gaussian filter diagram of the present invention;

[0079] Fig.12 The original image and the binarized image of the present invention;

[0080] Fig.13 It is the canny algorithm flow chart of the present invention;

[0081] Fig.14 It is the Canny operator graph of the present invention;

[0082] Fig.15 This is a diagram of a tire after expansion according to the present invention;

[0083] Fig.16 The corrosion rear tire diagram of the present invention;

[0084] Fig.17 Fill in the rear tire map for the present invention;

[0085] Fig.18 It is the principle diagram of Hough transformation in rectangular coordinates of the present invention;

[0086] Fig.19 This is a straight line detection diagram under heavy load state of the present invention;

[0087] Fig. 20 This is the ideal circular detection diagram under no load of the present invention;

[0088] Fig.21 It is a schematic diagram of tire detection variables of the present invention.

[0089] Among them, Figure 4 In the figure, Figure (a) is a photo of the tire before rotation, and Figure (b) is a photo of the tire after rotation;

[0090] exist Figure 5 In the figure, Figure (a) is a photo of the tire before perspective transformation, and Figure (b) is a photo of the tire after perspective transformation;

[0091] exist Figure 6 In the figure, (a) is the weighted average tire photo, and (b) is the weighted average grayscale histogram;

[0092] exist Figure 8 In the figure, (a) is the original photo of the tire, and (b) is the photo of the tire after grayscale transformation;

[0093] exist Fig. 9 In the figure, (a) is the histogram of the original tire image, and (b) is the histogram of the tire after grayscale transformation;

[0094] exist Fig.10 In the figure, (a) is the original photo of the tire, and (b) is the photo of the tire after image enhancement;

[0095] exist Fig.12 In the figure, Figure (a) is the original photo of the tire, and Figure (b) is the photo of the tire after binarization;

[0096] exist Fig.17 In the figure, Figure (a) is the image of the tire after filling, and Figure (b) is the image of the tire in the under-inflated state; DETAILED DESCRIPTION

[0097] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0098] Example 1: Please refer to Figure 1-Figure 21 , the present invention provides a technical solution:

[0099] A tire pressure calibration method based on image analysis, comprising:

[0100] Step 1: Obtain the tire image of the special vehicle, and take photos to obtain tire images at different shooting angles and different tire pressures;

[0101] Step 2: Preprocess the special vehicle tire image obtained in step 1 based on the image recognition algorithm

[0102] Step 2.1: Perform geometric transformation on the special vehicle tire image obtained in step 1

[0103] Step 2.1.1: The special vehicle tire image obtained in step 1 ( Figure 2 ) as the center, and rotate the image, such as Figure 3 As shown, the basic principles are as follows:

[0104] Assume |OP0|=|OP1|=L, then

[0105] cosα=x0 / L,sinα=y0 / L

[0106] sin(α+θ)=y1 / L=cosθsinα+cosαsinθ

[0107] cos(α+θ)=x1 / L=cosθcosα-sinαsinθ

[0108] Then there is

[0109] x1=cosθx0-sinθy0

[0110] y1=cosθy0-sinθx0

[0111] The rotation transformation matrix is

[0112]

[0113] The corresponding inverse operation is

[0114]

[0115] The rotation of the tire image is realized by Matlab. Figure 4 As shown;

[0116] Step 2.1.2: Correct the rotated image using perspective transformation

[0117] Perspective transformation is also called projection transformation. It can maintain "linearity" in image processing, that is, the straight lines in the original image remain straight lines after perspective transformation. The perspective transformation matrix transformation formula is:

[0118]

[0119] The contrast of the tire image after perspective transformation is obtained. Figure 5 As shown;

[0120] Step 2.2: Grayscale the image processed by the geometric transformation in step 2.1

[0121] Normally collected tire images are basically color images, which contain a large amount of information, which is not conducive to clearly identifying the contour of special vehicle tires. In addition, since tire images may be affected by various factors such as lighting and camera resolution during the collection process, the color information in the image may generate a lot of noise, which will affect the accuracy of tire contour recognition and tire pressure calibration. By graying the tire image, these color noises can be eliminated, making the image information clearer and more accurate.

[0122] Secondly, grayscale images can simplify the image processing process; grayscale images only contain brightness information and the amount of data is relatively small, which enables the processor to complete tasks faster when calculating and processing images, reducing the amount of calculation and storage space required; at the same time, simplifying the image processing process also helps to improve the real-time performance of tire contour recognition and tire pressure calibration, enabling it to respond to and process tire status changes more quickly;

[0123] Finally, grayscale images can better highlight the geometric and texture features of tires. In the process of tire contour recognition and tire pressure calibration, the main focus is on the geometric shape and texture changes of the tire, which can be better reflected in grayscale images. Through grayscale processing, the contour, edge and surface texture information of the tire can be observed more clearly, providing more accurate data support for subsequent recognition and calibration work.

[0124] Step 2.21: Grayscale the image using the weighted average method

[0125] This method assigns different weights to the red, green and blue channels according to the sensitivity of the human eye to different colors or other indicators, and calculates the weighted average of the values ​​of the three channels to obtain the grayscale value; specifically, it is to set R=G=B=WR+VG+UB, where W, V, and U represent the weights of the red, green, and blue channels respectively; because the human eye is most sensitive to green, followed by red, and least sensitive to blue, W>V>U should be set when setting the weights; experiments and theory have proved that when W=0.299, V=0.587, U=0.114, Gray=0.299*R+0.587*G+0.114*B, the most reasonable grayscale image can be obtained; the weighted average method can perform targeted processing according to the importance of different colors during the processing process, so the output result is more in line with the visual characteristics of the human eye, and the tire weighted average image histogram is obtained as shown below Figure 6 As shown;

[0126] Step 2.22: Grayscale conversion of the grayscaled image

[0127] Grayscale processing of tire images is a necessary preprocessing step. Grayscale processing simplifies the image processing process by removing the color information in the image and converting the image into a grayscale image. However, grayscale processing cannot completely solve the image quality problems caused by uneven brightness and shadow covering. Therefore, after grayscale processing, the image needs to be further grayscale transformed.

[0128] Grayscale transformation is the simplest and most effective method of contrast enhancement based on image grayscale value adjustment. Its basic principle is to map the grayscale value s(x,y) of the original image to a new grayscale value d(x,y) through a transformation function T, that is, d(x,y)=T[s(x,y)]. This transformation can be flexibly adjusted according to the characteristics of the image and processing requirements to achieve contrast enhancement or reduction. This method actually enhances the contrast of each part of the original image, that is, enhances the grayscale area we need in the input tire image and relatively suppresses those unnecessary grayscale areas. In tire image processing, piecewise linear transformation functions are widely used in the grayscale transformation process.

[0129] The piecewise linear transformation function divides the grayscale range of the image into multiple intervals and applies different linear transformation functions to each interval. By reasonably setting the transformation parameters of each interval, the area of ​​interest can be enhanced and the area of ​​no interest can be suppressed. This method can effectively improve the contrast of the tire image, making the target area such as the tire contour more prominent.

[0130] From a theoretical point of view, grayscale transformation expands the grayscale dynamic range of an image by adjusting the grayscale distribution of the image; this adjustment can expand the contrast of the image, making the image details clearer and the features more obvious; for tire images, grayscale transformation can significantly improve the contrast difference between the tire contour and the background, providing a more reliable data basis for subsequent contour extraction. The schematic diagram of piecewise linear transformation is shown in Figure 7 As shown; the functional form of the piecewise linear transformation is as follows:

[0131]

[0132] Step 2.23: Get the original image and the grayscale transformed image as follows Figure 8 As shown, the histogram of the original image and the grayscale transformed image is as follows Fig. 9 As shown in the figure, due to reasons such as light, the collected tire image may be dark or bright or have low or high contrast as a whole, or the shadow at the bottom of the tire may be heavy, causing the target area (tire) to not stand out, making it difficult to extract the effective tire area image from it, which brings difficulties to the detection. Therefore, after the tire image is grayed, it is also necessary to perform image enhancement processing on it; as shown in the figure, Fig.10 As shown;

[0133] Step 2.3: Perform denoising filtering on the image after grayscale processing in step 2.2

[0134] The collected tire image is affected by the material of the shooting device itself, the internal electronic components or circuit structure, the shooting environment, etc., and various noises will be generated, such as thermal noise, light response non-uniformity noise, photon noise, dark current noise. The image may be affected by electromagnetic interference, signal attenuation or errors in the digitization process during transmission, resulting in transmission noise; these noises may appear as random changes in pixel values ​​or stripes, spots, etc. in the image, which will interfere with the image and make it unclear, which has a great impact on the extraction of later features. Therefore, the tire needs to be denoised and filtered before image recognition; in this embodiment, the image after grayscale processing in step 2.2 is denoised and filtered by using the Gaussian filtering method;

[0135] Gaussian filtering is a linear filter that can effectively suppress Gaussian noise. The name of Gaussian filtering comes from the two-dimensional Gaussian function, that is, the two-dimensional normal distribution density function, with a mean of 0 and a variance of σ. 2 The two-dimensional Gaussian function is:

[0136]

[0137] Average smoothing treats all pixels in the neighborhood equally. In order to reduce blur points in the smoothing process and obtain a more natural smoothing effect, the weight of the center point of the template is appropriately increased. As the template moves away from the center point, its weight decreases rapidly, thereby ensuring that the center looks closer to the points closer to it. Based on this consideration, the template obtained is a Gaussian template. The commonly used 3x3 Gaussian template is shown in the following formula:

[0138]

[0139] The Gaussian template is a discretized representation of the continuous two-dimensional Gaussian function. Therefore, a Gaussian template of any size can be obtained by establishing a (2k+1)×(2k+1) matrix M, and the element value at the (i, j) position can be determined by the following formula:

[0140]

[0141] When the standard deviation σ takes different values, the shape of the two-dimensional Gaussian function will change greatly. Therefore, it is very important to choose a suitable standard deviation σ value in practical applications: if σ is too small, the weights of all pixels off the center will be very small, which is equivalent to the weighted sum response basically not considering the role of neighboring pixels. In this way, the filtering operation degenerates into a point operation of the image and cannot smooth the noise; on the contrary, if σ is too large and the neighborhood is relatively small, the Gaussian template in the neighborhood will degenerate into an average template; only when σ takes a suitable value can a better estimate of the pixel value be obtained; the default value of σ in MATLAB is 0.5. In practical applications, σ is usually taken as about 0.8 for a 3x3 template. For larger templates, the value of σ can be appropriately increased; in this embodiment, a 5x5 template is used in MATLAB, and the standard deviation σ=1 is taken to perform Gaussian filtering on the tire image, such as Fig.11 As shown;

[0142] Step 2.4: Binarize the image after denoising and filtering in step 2.3

[0143] After filtering the image, in order to extract the outer contour of the tire, we need to binarize the tire image. Thresholding is a basic and important means of image segmentation technology. Its core idea is to convert a grayscale image into a binary image with only black and white colors. By setting a grayscale threshold, the pixels in the image are divided into two categories: one category of pixels has a grayscale value higher than the threshold, and the other category has a grayscale value lower than the threshold. The former is usually assigned a maximum grayscale value, i.e. 255, which is white; the latter is assigned a minimum grayscale value, i.e. 0, which is black. In this way, the details and features in the original image are highlighted, making the image present a sharp black and white contrast effect, that is, the entire tire image presents a clear visual effect of only black and white. Through the image binarization method, we can simplify a complex grayscale image into a binary image with only black and white colors, so that it is easier to extract useful information from the image, which provides convenience for subsequent image analysis and processing. Since the binary image data is simple enough, many visual algorithms rely on binary images. Through binary images, the shape and contour of objects can be better analyzed.

[0144] This embodiment uses the OTSU algorithm to perform binarization processing on the tire grayscale image, and the system automatically calculates a threshold value suitable for the tire image to perform binarization processing on the image; Fig.12 As shown, it is a comparison between the original image and the binary image processed by the Otsu algorithm;

[0145] Step 2.5: Perform edge detection and analysis on the image binarized in step 2.4

[0146] The core basis of machine visual perception lies in the feature information of the edge of the image; as the part with the most significant characteristic changes in the image, it not only carries key information such as the shape and structure of the object; for the image of special vehicle tires, the edge feature corresponds to the tire contour line, and also provides us with a basis for determining the position of the tire; edge detection technology is a basic method in image processing and computer vision. By identifying the pixel points with step changes in grayscale in the digital image, the edge contour of the tire can be accurately identified, and the position of the tire in the image can be effectively located; for the processing of vehicle tire images, by accurately detecting the edge contour of the tire, we can better understand and analyze the tire image; there are two types of algorithms for edge detection, namely, search-based algorithms and zero-crossing-based algorithms. This embodiment uses a zero-crossing-based algorithm to perform edge detection and analysis on the image after binarization processing in step 2.4, specifically, by finding the zero crossing of the second-order derivative of the image to find the boundary information. This embodiment uses the Canny operator zero-crossing point or the zero-crossing point represented by the nonlinear difference. The goal of the Canny algorithm is to find an optimal edge detection solution or to find the position where the grayscale intensity changes the most in the image;

[0147] like Fig.13 As shown, this is the Canny algorithm process:

[0148] Step 2.5.1: First, filter and remove noise and calculate the amplitude and direction by difference. The gradient amplitude is:

[0149] The gradient direction is:

[0150] Among them, G x is the horizontal gradient operator, G y is the vertical gradient operator;

[0151] Step 2.5.2: Then perform non-maximum suppression, which is to suppress the non-maximum data in the image, which can also be understood as excluding the possibility that the non-maximum data is an edge; the larger the element value in the image gradient amplitude matrix in the neighborhood, the larger the gradient value of the point in the image. Combined with the gradient direction of the detection point, the approximate edge information can be located;

[0152] Step 2.5.3: Perform the hysteresis threshold again. In simple terms, it is to use double thresholds to screen the binary image. By selecting appropriate large and small thresholds, an edge image that is closest to the real edge of the image can be obtained. The specific implementation method is: an edge image is obtained according to the high threshold. Such an image contains very few false edges, but due to the high threshold, the edges of the generated image may not be closed. To solve this problem, another low threshold is used; in the high threshold image, the edges are linked into contours. When the endpoint of the contour is reached, the algorithm will find a point that meets the low threshold in the neighborhood of the breakpoint, and then collect new edges based on this point until the edge of the entire image is closed; such as Fig.14 That is the tire image processed by the canny algorithm;

[0153] Step 3: The tire contour extraction algorithm based on morphological image processing is used to identify and locate the tire contour of the image preprocessed in step 2.

[0154] In order to extract the characteristic information of the tire, we need to shield other parts in the image after complex edge detection, retain the target area we need (i.e., the outer contour of the tire), accurately identify the outer contour of the tire, and obtain a circular or quasi-circular binary image, which provides a basic object for the next step of extracting the characteristic information of the tire; This embodiment uses a tire contour extraction algorithm based on morphological image processing to extract and analyze the tire contour

[0155] Step 3.1: Observe the tire edge detection image and find that due to reasons such as light and photo quality, the detected tire edge is not a completely connected continuous edge, which brings inconvenience to the subsequent tire extraction; the first step of the morphological tire contour extraction in this embodiment is to perform a morphological closing operation on the tire edge image, which can not only connect the discontinuous edges, but also make the tire edge smoother.

[0156] Step 3.1.1: First, use the structural element SE to Fig.14 Dilation operation is performed on the tire edge image

[0157] The structural element SE is a set, which is the basis of mathematical morphology operations. Each structural element has an origin. The structural element is located at point P, which means that the origin of this structural element is P.

[0158] The dilation operation is to expand the outline of the image. 2 The set A and S of elements in , use S to expand A, denoted as It is expressed as:

[0159]

[0160] The specific operation method is to use a 3×3 rectangle as a template, and perform the following processing on each pixel x in the image: pixel x is placed at the center of the template, and according to the size of the template, all other pixels covered by the template are traversed, and the value of pixel x is modified to the maximum value of all pixels. The result of this operation will connect the prominent points on the periphery of the image and extend them outward; inflating the tire can effectively connect the discontinuous edges of the tire, and at the same time will not connect the edges of other objects near the tire to it and cause interference. Fig.14 The result of expansion is Fig.15 As shown;

[0161] Step 3.1.2: Then the same applies to the structural element SE pair Fig.15 The expansion result of the tire image is corroded, and the effect is as follows Fig.16 As shown; the top-down operation is opposite to the expansion operation, specifically:

[0162] Z 2 The set A and S of elements in , use S to corrode A, denoted as It is expressed as:

[0163]

[0164] Let the structural element S originally located at the origin of the image be 2 Move on the plane. If S can be completely contained in A when the origin of S is translated to point z, then the set of all such points z is the eroded image of S on A. The specific operation method is to use a 3×3 rectangle as a template and perform the following processing on each pixel x in the image: pixel x is at the center of the template, and according to the size of the template, all other pixels covered by the template are traversed, and the value of pixel x is modified to the minimum value of all pixels. The result of this operation is that the protruding points on the periphery of the image will be eroded.

[0165] After the morphological closing operation, the edge of the tire in the image is clearer, smoother and interconnected, and the edge features of the tire are more completely displayed, laying the foundation for more accurate and convenient extraction of its geometric features; it can be easily found from the tire image that the tire area occupies most of the image area and is the object with the largest area among all objects in the image, that is, the object containing the most pixels. Therefore, this embodiment uses the maximum connected domain method in mathematical morphological image segmentation to extract the tire; however, the edge detection image only contains the edge of the object, that is, the edge of the candidate connected area, so the edge image needs to be hole-filled next;

[0166] Step 3.2: Fill holes in edge images

[0167] For binary images, a hole is a closed image background area surrounded by the edge of the image. Hole filling (region filling) is to fill each hole area with 1 that is not connected to the boundary. Its purpose is to fill the entire object area with 1 starting from a point on the object boundary, and finally divide the entire image into several connected areas. There are many hole filling algorithms, and the simplest filling method in mathematical morphology is to use an SE to perform an expansion operation starting from a point in a hole, and fill the entire area in the region of interest, that is, the object, through an iterative expansion algorithm; get the hole filling algorithm. Fig.16 The result of hole filling in tire image is as follows Fig.17 As shown; So far, this embodiment has successfully detected the tire contour using the mathematical morphology image processing method;

[0168] Step 4: Perform ground wire detection on the tire profile extracted in step 3 and extract tire profile data

[0169] When a tire faces overload or abnormal tire pressure, its shape will change significantly. This change is mainly reflected in the height change of the tire and the size of the contact area between the tire and the ground. Specifically, when the tire pressure is too low or the tire is subjected to too much load, the overall height of the tire will drop significantly, resulting in a significant increase in its contact area with the ground compared to the normal state. This is because insufficient tire pressure or overload weakens the internal support of the tire, and the tire body is flattened under the action of gravity, thereby increasing the contact range with the ground. On the contrary, if the tire pressure is too high, the elasticity of the tire body will be significantly reduced, and the rigidity will increase accordingly. This will cause the tire to reduce the contact area with the ground during driving, because the excessively high air pressure makes the tire body harder and less likely to deform, thereby reducing the actual contact area with the ground. Based on such physical characteristics, the tire pressure and overload detection algorithm based on image processing technology proposed in this embodiment is constructed. The core idea of ​​the algorithm is to identify the shape change of the tire by analyzing the image information of the tire, so as to judge its tire pressure and load state. In practical applications, in order to quantify the contact area between the tire and the ground, we can achieve this by measuring the length of the contact line between the side of the tire and the ground. The length of this contact line can actually be used as an effective indicator to reflect the size of the contact area between the tire and the ground; when the tire pressure or load state of the tire changes, the length of the contact line will also change accordingly; therefore, by accurately identifying and measuring the length of this contact line through image processing technology, we can indirectly infer the tire pressure and load state of the tire, providing an important basis for subsequent tire pressure calibration and overload detection; in order to explore the calibration method of tire pressure of special vehicles, this embodiment detects the contact length between the tire and the ground (tire ground contact line) through image recognition, and measures the length of the straight line in the image through a straight line detection algorithm based on Hough transform; at the same time, it is also necessary to extract other parameters such as the radius of the tire through feature extraction to provide necessary conditions for the subsequent tire pressure calibration algorithm;

[0170] Step 4.1: Measure the tire contact line and tire radius of the tire profile based on the Hough transform detection algorithm

[0171] Step 4.1.1: Line Detection

[0172] (1) In a rectangular coordinate parameter space, such as the space of image xy coordinates, the expression for a straight line passing through a point (x, y) is:

[0173] y i =ax i +b

[0174] In the above formula, a is the slope; b is the intercept;

[0175] There are countless straight lines passing through the point (x, y), corresponding to different values, and they all satisfy the above formula;

[0176] (2) If x and y are considered constants and the original parameters a and b are considered variables, the above formula can be expressed as

[0177] b=-x i a+y i

[0178] is transformed to the parameter plane ab; this transformation is the rectangular coordinate for the point (x i ,y i ) is the Hough transform of the point (x i ,y i ) is the only equation in parameter space; consider another point (x j ,y j ), which also has a corresponding straight line in parameter space:

[0179] b=-x j a+y j

[0180] This line and the point (x i ,y i ) in the parameter space intersects at a point (a0, b0), such as Fig.18 As shown;

[0181] (3) In the image coordinate space, the point (x i ,y i ) and (x j ,y j ) corresponds to a straight line in the parameter space ab. These straight lines intersect at point (a0, b0), and a0, b0 is the midpoint (x i ,y i ) and (x j ,y j ) is the parameter of the line determined by the Hough transform. Conversely, all lines that intersect at the same point in the parameter space have collinear points corresponding to them in the image coordinate space; based on this property, given some edge points in the image coordinate space, the equation of the line connecting these points can be determined by Hough transform;

[0182] In specific calculations, the parameter space can be regarded as discrete; a two-dimensional accumulation array A(a,b) is established, the range of the first dimension is the possible range of the slope of the line in the image coordinate space, and the range of the second dimension is the possible range of the intercept of the line in the image coordinate space; at the beginning, A(a,b) is initialized to 0, and then for each foreground point (x i ,y i ), substitute the discrete value of each a in the parameter space into the formula b = -xj a+y j , and thus calculate the corresponding b value; every time a pair (a, b) is calculated, the corresponding array element A(a, b) is increased by 1, that is, A(a, b) = A(a, b) + 1; after all calculations are completed, the maximum peak value of A(a, b) is found in the parameter space voting result, and the corresponding a0, b0 are the parameters of the straight line equation with the largest number of collinear points in the original image;

[0183] Step 4.1.2: Curve detection

[0184] (1) Hough transform is also applicable to curve detection with known equations. A known curve equation in the image coordinate space can also be mapped to its corresponding parameter space. Therefore, a point in the image coordinate space can be mapped to a corresponding trajectory curve or surface in the parameter space. If the curves or surfaces corresponding to each discontinuity point in the parameter space can intersect, the maximum of the parameter space and the corresponding parameters can be found. If the curves or surfaces corresponding to each discontinuity point in the parameter space cannot intersect, it means that the discontinuity point does not conform to a known curve.

[0185] When using Hough transform for curve detection, the most important thing is to write the transformation formula from image coordinate space to parameter space; for example, for a known circle equation, the general equation of its rectangular coordinates is

[0186] (x) 2 +(yb) 2 =r 2

[0187] Where (a, b) is the coordinate of the center of the circle, r is the radius of the circle, and they are the parameters of the image; then, the parameter space can be expressed as (a, b, r), and a circle in the image coordinate space corresponds to a point in the parameter space;

[0188] The specific calculation is the same as the method discussed above, except that the array accumulator is three-dimensional (a, b, r); the calculation process is to increase ab within the value range and solve the r value that satisfies the above formula. Every time a (a, b, r) value is calculated, 1 is added to the array element A (a, b, r); after the calculation is completed, the a, b, r corresponding to the largest (a, b, r) found is the parameters of the circle;

[0189] Step 4.2: Extract, detect and analyze the features of special vehicle tire profiles

[0190] For the above Fig.17 The image of the tire after filling (the right image of the tire under ideal pressure) is tested by Hough transform of straight lines and curves, as shown below Fig.19 As shown Fig. 20 As shown;

[0191] Fig.19 For the state of the vehicle tire under load under simple background and ideal conditions, the longest straight line in the image is detected by Hough transform straight line, which is the tire ground contact line, and the detected length is 227.00 pixels; by merging, discarding and other screening, adjusting the range of the straight line of the algorithm to detect the image, the length of the straight line can be accurately detected, and the longest straight line is output, which is what we want;

[0192] Fig. 20 The vehicle tire is in a simple background, ideally without load. Through Hough transform curve detection, a red circle that best fits the tire image is detected, which is the most suitable tire outer contour model simulated by the MATLAB algorithm. The radius length of the circle is detected to be 241.27 pixels. By adjusting the accuracy of the algorithm detection and the approximate range of the radius, the fitting circle can be accurately detected and the radius of the fitting circle can be output, which is what we want.

[0193] So far, this embodiment has successfully used the Hough transform detection algorithm to detect data such as the vehicle tire profile radius and the grounding wire length;

[0194] Step 5: Use ABAQUS software to build a finite element model of the vehicle tire and calibrate the tire pressure

[0195] Through finite element model simulation, we can simulate the deformation and stress distribution of tires under different tire pressure and load conditions, so as to more accurately understand the performance of tires in actual use; this simulation method not only overcomes the limitations of physical experimental environment, but also can significantly reduce costs and shorten the R&D cycle;

[0196] Step 5.1: Perform load characteristic analysis on a fine parameterized finite element model of a heavy-duty tire

[0197] The focus of this step is to explore the changing law of tire pressure of special vehicles under fixed loads; through simulation experiments, we can obtain a large amount of data on the relationship between tire pressure and tire performance, and then provide strong support for the optimization of tire pressure calibration algorithm; simulate special vehicle tires through the vehicle tire finite element model established by ABAQUS software; first use AutoCAD to complete the drawing of the heavy-duty tire two-dimensional model, import the two-dimensional model into HyperMesh software to mesh it as needed, obtain a two-dimensional structural model, and then complete the remaining pre-processing modeling steps to obtain a two-dimensional finite element model of the heavy-duty tire;

[0198] When constructing a three-dimensional model of a heavy-duty tire, the center is used as a reference point. The two-dimensional finite element model of the tire is rotated by editing the inp file command stream in the finite element simulation software Abaqus to obtain a three-dimensional structural model of the tire. During the rotation process, a single rotation step is used to obtain node information at a constant sampling frequency while maintaining a certain accuracy. However, considering that the stress distribution and strain degree of different parts of the tire during operation have large gradients, the tire analysis results will be quite different. If the step size is small, on the one hand, the calculation accuracy will be improved, but at the same time, the amount of calculation during simulation will be increased; the non-contact area does not require high model accuracy, so if this part of the model is calculated with high accuracy, it will cause a waste of computing resources and may even lead to non-convergence during the calculation process; on the contrary, if a large rotation step size is set, the model calculated by simulation cannot represent the actual behavior and stress of the tire. Therefore, when constructing the three-dimensional model, the variable step rotation method is selected to simplify the non-contact area model while improving the model accuracy before and after the tire-road contact area, reducing the calculation cost while ensuring the research needs, and more realistically simulating the actual behavior of the tire;

[0199] The research focus of this step is to explore the changing law of tire pressure of special vehicles under fixed load. Through simulation experiments, we can obtain a large amount of data on the relationship between tire pressure and tire performance, which can provide strong support for the optimization of tire pressure calibration algorithm. The special vehicle tire is simulated by the vehicle tire finite element model established by ABAQUS software. The experiment mainly collects data under two different load conditions, and the different states of the tire under 2 tons and 3 tons. The collected data are shown in Table 1:

[0200] Table 1: Tire pressure and ground wire data under different working conditions

[0201]

[0202] The difference between the grounding wire length data collected by the finite element model and the grounding wire length measurement value detected by the image recognition processing algorithm is because the unit of the tire grounding wire length processed by image recognition is pixel point, while the unit of the grounding wire length of the finite element tire model is millimeter. The tire outer contour radius measurement value processed by image recognition is R1=335.52 pixels, while the original actual value radius of the finite element tire model is R0=660mm. In order to facilitate the subsequent exploration of the tire pressure calibration model and establish the corresponding relationship between the actual value and the measured value, this step sets a relationship coefficient K. The basic data of the tire outer contour radius measurement value processed by image recognition and the original actual value radius of the finite element tire model are the ratio, then K=R1 / R0=0.5083;

[0203] There is an error between the ground wire length data collected by the finite element model and the ground wire length measurement value detected by the image recognition processing algorithm, which is a common problem in engineering practice; in order to evaluate this method more accurately, we need to calculate its relative error and absolute error; absolute error refers to the absolute value of the difference between the measured value and the actual value, which reflects the degree of deviation of the measurement result; and relative error is the ratio of absolute error to actual value, usually expressed in percentage form, which reflects the relative degree of deviation of the measurement result; in order to determine whether the ground wire length data obtained by the tire finite element model processed by image recognition in this article is accurate, we now judge whether the image preprocessing detection data meets the requirements through the relative error and absolute error between the actual value and the measured value; the absolute error can be obtained by a simple subtraction operation, that is, the measured value minus the actual value (taking the absolute value); the relative error requires the absolute error to be divided by the actual value and multiplied by 100% to obtain the percentage form;

[0204] Next, we calculate these two errors based on the given data; assuming that the grounding wire length data collected by the finite element model is the actual value, and the grounding wire length measurement value detected by the image recognition processing algorithm is the measured value; we can calculate the absolute error and relative error of each data point one by one, and organize the results into a table; then the actual value and measured value of the corresponding pixel value calculated by the relationship coefficient of the grounding wire in Table 1 above, as well as the relative error and absolute error are shown in Tables 2 and 3:

[0205] Table 2: Actual and measured values ​​of tire contact line and their errors under 2t condition

[0206]

[0207] Table 3: Actual and measured values ​​of tire contact line and their errors under 3t condition

[0208]

[0209] In the above tables 2 and 3, we can see the actual value, corresponding value, measured value, absolute error and relative error of each data point; by calculating and analyzing these error data, we can draw the following conclusions: compared with the measured value, the relative error of the ground contact line data of the finite element model tire is less than 5%, indicating that the accuracy of the data of the ground contact length detected in the image recognition processing algorithm of this embodiment is high, and the subsequent exploration of the tire pressure calibration model can be carried out;

[0210] Step 5.2: Build a tire pressure calibration model

[0211] Establish a tire detection variable diagram as follows Fig.21As shown, in the special tire pressure detection algorithm discussed in this step, a crucial concept is the ground secant (GS for short); the ground secant actually describes the boundary line of the tire side that contacts the ground; in order to understand this concept more intuitively, we can imagine the ground as an infinitely extended straight line, and the side of the tire is approximately regarded as an arc or a part of a circle; when the tire contacts the ground, these two "lines" and "arcs" will intersect at a certain point or in a certain area, and the contact line formed is the ground secant. The length of the ground contact line is called the ground contact line length, which is represented by D. This length not only reflects the size of the contact area between the tire and the ground, but is also closely related to factors such as tire pressure, load and driving status. Therefore, by accurately measuring and analyzing the ground contact line length, we can indirectly infer the current state of the tire and provide strong data support for tire pressure detection. It is worth noting that the radius R of the tire when it is suspended is also an important parameter. When the tire is not subject to ground pressure, its shape remains a perfect circle or an approximate circle, and the radius at this time is R. The relationship between the suspended radius R and the ground contact line length D can provide us with information on the deformation of the tire when it is under pressure, which further helps us understand the working state and performance of the tire. Through the precise measurement and analysis of parameters such as the ground contact line length D and the radius R of the tire when it is suspended, we can have a deeper understanding of the working state of the tire and explore the tire pressure calibration model.

[0212] For this step, the straight line segments used in the special vehicle tire overload condition detection algorithm are all horizontal or vertical in the image, and the simplest representation method is to calculate the number of pixels between the two pixel points at their endpoints. Therefore, this step proposes the direct method and the sine value method to realize the special vehicle tire overload condition detection.

[0213] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0214] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A tire pressure calibration method based on image analysis, characterized in that: include: Step 1: Acquire vehicle tire image; Step 2: preprocessing the vehicle tire image obtained in step 1 based on an image recognition algorithm; Step 3: Using a tire contour extraction algorithm based on morphological image processing to identify and locate the tire contour of the vehicle tire image preprocessed in step 2; Step 4: Perform ground line detection on the tire profile extracted in step 3 and extract tire profile data; Step 5: Use the tire profile data extracted in step 4 to build a finite element model of the vehicle tire and calibrate the tire pressure.

2. The tire pressure calibration method based on image analysis according to claim 1, characterized in that: Step 2: The process of preprocessing the vehicle tire image includes: Step 2.1: Perform geometric transformation processing on the special vehicle tire image obtained in step 1; Step 2.2: grayscale the image processed by the geometric transformation in step 2.1; Step 2.3: Use Gaussian filtering to perform denoising filtering on the image after grayscale processing in step 2.2; Step 2.4: Use the OTSU algorithm to binarize the image after denoising and filtering in step 2.3; Step 2.5: Perform edge detection and analysis on the image binarized in step 2.

4.

3. The tire pressure calibration method based on image analysis according to claim 2, characterized in that: Step 2.1 The process of geometric transformation of the vehicle tire image includes: Step 2.1.1: Taking the origin of the vehicle tire image obtained in step 1 as the center, the image is rotated. The process includes: Assume |OP0|=|OP1|=L, then cosα=x0 / L,sinα=y0 / L sin(α+θ)=y1 / L=cosθsinα+cosαsinθ cos(α+θ)=x1 / L=cosθcosα-sinαsinθ have x1=cosθx0-sinθy0 y1=cosθy0-sinθx0 The rotation transformation matrix is The corresponding inverse operation is Step 2.1.2: Correct the rotated image using perspective transformation The perspective transformation matrix transformation formula is:

4. The tire pressure calibration method based on image analysis according to claim 2, characterized in that: Step 2.2 The process of graying the image includes: Step 2.21: Use the weighted average method to grayscale the image; Step 2.22: Perform grayscale conversion on the grayscaled image, including: The gray value s(x,y) of the original image is mapped to a new gray value d(x,y) through the transformation function T, d(x,y) = T[s(x,y)], and the gray value d(x,y) is a piecewise linear transformation function; Step 2.23: Enhance the tire image after the grayscale transformation in step 2.

22.

5. The tire pressure calibration method based on image analysis according to claim 2, characterized in that: Step 2.5 The process of edge detection and analysis of the image includes: Step 2.5.1: Filter and denoise the image and calculate the amplitude and direction by difference Then the gradient amplitude is: The gradient direction is: Among them, G x is the horizontal gradient operator, G y is the vertical gradient operator; Step 2.5.2: Perform non-maximum suppression on the image and locate the edge information of the image by combining the gradient direction of the detection point; Step 2.5.3: Perform hysteresis threshold processing on the image to obtain a closed curve of the edge of the entire image.

6. The tire pressure calibration method based on image analysis according to claim 1, characterized in that: Step 3: The process of identifying and locating the tire contour in the image includes: Step 3.1: Use the tire contour extraction algorithm of morphological image processing to perform morphological closing operation on the tire edge image to connect the discontinuous edges, including Step 3.1.1: Use the structure element SE to perform a dilation operation on the tire edge image; Step 3.1.2: Use the structure element SE to perform corrosion operation on the expansion result of the tire image; Step 3.2: Use an iterative dilation algorithm to fill holes in the edge image.

7. The tire pressure calibration method based on image analysis according to claim 1, characterized in that: Step 4: The process of detecting the grounding line of the tire contour and extracting the tire contour data includes: Step 4.1: Measuring the tire contact line and tire radius of the tire profile based on the detection algorithm of Hough transform, including a straight line detection process and a curve detection process; Step 4.2: Extract, detect and analyze the features of the vehicle tire profile to detect the tire contact line and tire radius of the tire profile.

8. The tire pressure calibration method based on image analysis according to claim 1, characterized in that: Step 5: Use ABAQUS software to build a finite element model of the vehicle tire, including: Step 5.1: Perform load characteristic analysis of a heavy-duty tire fine parameterized finite element model; Step 5.2: Establish a tire pressure calibration model.

9. The tire pressure calibration method based on image analysis according to claim 1, characterized in that: In the tire pressure calibration model, the direct method and the sine value method are used to calculate the vehicle tire overload condition.

Images