A method for travel path control based on monocular vision recognition and an unmanned laying vehicle for crack-resistant base fabric.

Abstract

This invention provides a method for controlling the travel path based on monocular vision recognition and an unmanned laying vehicle for crack-resistant base fabric, belonging to the technical field of machine vision recognition. The method includes using a monocular video acquisition device to collect video information of a target reference object on the crack-resistant base fabric laying equipment and converting it into image data to determine the initial center coordinate position of the target reference object in the image; adjusting the acquisition angle of the monocular image acquisition device on the target reference object according to the initial center coordinate position of the target reference object in the image, so that the distance between the initial center coordinate position of the target reference object in the image and a preset target center coordinate position in the image meets a preset initial distance; acquiring the travel center coordinate position of the target reference object in the image in real time and calculating the deviation between the travel center coordinate position and the target center coordinate position; and controlling the travel path based on the calculated deviation. This invention can significantly improve the flatness and efficiency of crack-resistant base fabric laying and effectively reduce labor costs.

Description

Technical Field

[0001] This invention relates to the field of machine vision recognition technology, and particularly to a method for controlling the travel path based on monocular vision recognition and an unmanned laying vehicle for anti-crack base fabric. Background Technology

[0002] During road construction or maintenance, a layer of crack-resistant base fabric is typically laid on the old road surface to prevent cracks from spreading and to extend the road's service life. Common crack-resistant base fabrics are made of polyester fiber or fiberglass mesh composites, possessing high tensile strength, wear and corrosion resistance, and excellent waterproof performance, and they bond well with materials such as concrete and asphalt.

[0003] Traditional crack-resistant base fabric laying relies primarily on manual operation, which is not only inefficient but also significantly affected by human factors in terms of laying smoothness. In recent years, with the development of automation technology, various semi-automatic crack-resistant base fabric laying devices have emerged to improve work efficiency and laying smoothness, but they still cannot completely eliminate manual control. For example, CN218621705U discloses a crack-resistant base fabric laying device that lays the base fabric by installing a placing boom on a loader. Although existing laying devices improve the laying quality of crack-resistant base fabric to some extent compared to manual laying, the movement path of the laying device is mainly controlled manually, and the movement path is prone to deviation, which greatly affects the laying quality of the crack-resistant base fabric. Summary of the Invention

[0004] To address the above problems, this invention provides a method for controlling the travel path based on monocular vision recognition and an unmanned laying vehicle for crack-resistant base fabric. It can automatically determine the travel path for laying crack-resistant base fabric based on machine vision recognition results, and adjust the laying direction in real time when the travel path of the laying equipment deviates, which can significantly improve the flatness and efficiency of laying crack-resistant base fabric.

[0005] This invention provides a method for controlling a walking path based on monocular vision recognition, comprising:

[0006] The video information of the target reference object on the anti-crack base cloth laying device is acquired by a monocular video acquisition device and converted into image data. The initial center coordinate position of the target reference object in the image is determined. The target reference object can move horizontally during the movement, which changes the movement center coordinate position of the target reference object in the image.

[0007] The monocular image acquisition device adjusts the video information acquisition angle of the target reference object based on the initial center coordinate position of the target reference object in the image, so that the distance between the initial center coordinate position of the target reference object in the image and the preset target center coordinate position in the image meets the preset initial distance.

[0008] The system acquires the coordinates of the travel center of the target reference object in the image in real time, and calculates the deviation between the coordinates of the travel center of the target reference object in the image and the coordinates of the target center.

[0009] The travel path of the anti-crack base fabric laying device is controlled based on the calculated deviation between the travel center coordinates of the target reference object in the image and the target center coordinates.

[0010] As a further improvement of the present invention, the step of using a monocular video acquisition device to acquire video information of the target reference object on the anti-crack base fabric laying device and converting it into image data, and determining the initial center coordinate position of the target reference object in the image, includes converting the RGB video information of the target reference object on the anti-crack base fabric laying device acquired by the monocular video acquisition device into grayscale image data, and preprocessing the grayscale image data to improve the quality of the grayscale image data.

[0011] As a further improvement of the present invention, the preprocessing of the grayscale image data includes using contrast-limited adaptive histogram equalization to preprocess the grayscale image to enhance the contrast of the grayscale image data.

[0012] As a further improvement of the present invention, after preprocessing the grayscale image using contrast-limited adaptive histogram equalization, a Region of Interest (ROI) containing the target reference is selected from the preprocessed grayscale image, and a gamma correction algorithm is used to adjust the overall brightness of the ROI image. The formula for adjusting the overall brightness of the ROI image using the gamma correction algorithm is as follows:

[0013]

[0014] In the formula, I(x,y) represents the pixel values ​​of the original image. γ represents the pixel value of the corrected image, and γ is the gamma correction coefficient.

[0015] As a further improvement of the present invention, determining the initial center coordinate position of the target reference object in the image includes:

[0016] In the first frame of the video information, a region of interest (ROI) containing the target reference is selected as a training sample for the filter tracker.

[0017] The training sample set X is obtained by amplification reconstruction-cyclic shift sampling method, and the ROI is multisampled using a cyclic matrix. The resulting sample set constitutes a typical cyclic matrix.

[0018] The Sobel operator is used to convolve with the training sample set X to obtain gradient information in the horizontal and vertical directions. Then, the feature information in each block is statistically analyzed according to the FHOG feature classification method, and finally combined to form the feature information of the training sample set X. The block is obtained by dividing the ROI region using the FHOG feature extraction method, and each block is composed of multiple adjacent cells.

[0019] Based on the training sample set X, through the training samples Establish Gaussian regression labels within the Fourier domain , along with It decreases as the number of cyclic shifts increases, where Filter tracker Training is performed using regularized least squares, so that the samples... and Gaussian regression labels The squared error is minimized:

[0020]

[0021] In the formula, It is a regularization parameter that controls overfitting and prevents... Overfitting during training;

[0022] The unique closed-form solution is:

[0023]

[0024] In the formula, It is a sample The sample matrix formed It is the conjugate transpose, that is , yes The complex conjugate matrix, It is the identity matrix. It is a Gaussian regression label. The vector formed;

[0025] Using mapping Sample Mapped to Hilbert space, filter tracker Represented as a sample Linear combination:

[0026]

[0027] In the formula, These are the combination coefficients;

[0028] Introduce a kernel function in the dual space:

[0029]

[0030] but The nuclear version The solution is:

[0031]

[0032] In the formula, This is the kernel autocorrelation matrix of the sample:

[0033]

[0034] Utilizing the property that the training sample set X has diagonalizable circumcyclic matrices, matrix operations are transformed into dot product operations in the Fourier domain, ultimately yielding the kernel correlation filter tracker. The closed-form solution is as follows:

[0035]

[0036] In the formula, the symbol This represents the discrete Fourier transform of the variable. It is a sample Kernel autocorrelation matrix The first line in;

[0037] Using the kernel correlation filter tracker The position of the target reference object in the image is detected and located.

[0038] The present invention provides an unmanned laying vehicle for crack-resistant base fabric, which uses the following travel path control method based on monocular vision recognition for travel path control:

[0039] The video information of the target reference object on the anti-crack base cloth laying device is acquired by a monocular video acquisition device and converted into image data. The initial center coordinate position of the target reference object in the image is determined. The target reference object can move horizontally during the movement, which changes the movement center coordinate position of the target reference object in the image.

[0040] The monocular image acquisition device adjusts the video information acquisition angle of the target reference object based on the initial center coordinate position of the target reference object in the image, so that the distance between the initial center coordinate position of the target reference object in the image and the preset target center coordinate position in the image meets the preset initial distance.

[0041] The system acquires the coordinates of the travel center of the target reference object in the image in real time, and calculates the deviation between the coordinates of the travel center of the target reference object in the image and the coordinates of the target center.

[0042] The travel path of the anti-crack base fabric laying device is controlled based on the calculated deviation between the travel center coordinates of the target reference object in the image and the target center coordinates.

[0043] As a further improvement of the present invention, it includes a laying mechanism for smoothly laying the crack-resistant base fabric onto the road surface, a trajectory recognition component for monitoring the travel path, and a travel control module for adjusting the path. The trajectory recognition component acquires the travel path deviation of the unmanned crack-resistant base fabric laying vehicle in real time and feeds back the deviation information to the travel control module to adjust the travel path of the unmanned crack-resistant base fabric laying vehicle, so that the unmanned crack-resistant base fabric laying vehicle lays the crack-resistant base fabric according to the expected travel path.

[0044] As a further improvement of the present invention, the trajectory recognition component includes a monocular video acquisition device, a gimbal, a trajectory reference device, a telescopic connector, and an image data processing unit. The trajectory reference device serves as a target reference object. The monocular video acquisition device is mounted on the laying mechanism via the gimbal, and the trajectory reference device is mounted on the laying mechanism via the telescopic connector. The monocular video acquisition device is used to acquire video information containing the trajectory reference device and transmit the acquired video information to the image data processing unit for processing. The deviation between the travel center coordinate position of the trajectory reference device in the image and the target center coordinate position is calculated, and the deviation information is fed back to the travel control module to adjust the travel path of the laying mechanism in real time.

[0045] As a further improvement of the present invention, the travel control module includes a steering controller, a travel controller, and a brake controller. The travel control module controls the steering controller to work based on the deviation information provided by the image data processing unit, and adjusts the travel direction of the laying mechanism. The travel control module controls the work of the travel controller and the brake controller to make the laying mechanism move or stop.

[0046] As a further improvement of the present invention, the travel control module also includes an early warning unit and multiple environmental monitors. The early warning unit is equipped with a pre-trained YOLO detection model for obstacles, and the environmental monitors are used to monitor the environment around the unmanned laying vehicle for crack-resistant base fabric in real time. After receiving the surrounding environment video collected by the environmental monitors, the early warning unit analyzes the video information using the YOLO detection model. When a safety hazard is detected in the surrounding area based on the video information from the environmental monitors, the travel control module will feed back to the brake controller to apply the brakes. If the surrounding environment is safe, the brakes will be released and the travel controller will be controlled to move the laying mechanism forward.

[0047] This invention provides a method for controlling the travel path based on monocular vision recognition and an unmanned vehicle for laying crack-resistant base fabric, which has at least the following beneficial effects:

[0048] 1. The travel center coordinates of the target reference object in the image are obtained in real time through monocular vision recognition and compared with the preset target center coordinates. The difference is calculated to obtain the deviation between the travel center coordinates of the target reference object in the image and the preset target center coordinates. The travel path of the anti-crack base fabric laying device is controlled according to the deviation, so as to accurately control the travel direction of the anti-crack base fabric laying device, ensuring that the anti-crack base fabric can be laid in the correct direction, which can significantly improve the laying accuracy.

[0049] 2. Automated laying replaces the traditional manual laying method, reducing the need for manual intervention and improving the efficiency of laying operations.

[0050] 3. It can monitor the environmental safety around the anti-crack base fabric laying equipment in real time. When potential risks are detected, it can take immediate braking measures to prevent the equipment from colliding or other accidents, thus improving the safety of the laying equipment operation. Attached Figure Description

[0051] Figure 1 This is a flowchart illustrating the movement path control method based on monocular vision recognition according to Embodiment 1 of the present invention.

[0052] Figure 2 This is a schematic diagram of the unmanned laying vehicle for anti-crack base fabric according to Embodiment 2 of the present invention.

[0053] Figure 3 This is a partial structural schematic diagram of the anti-crack base fabric laying mechanism of the unmanned laying vehicle for anti-crack base fabric according to Embodiment 2 of the present invention.

[0054] Figure 4 This is a side view of the unmanned laying vehicle for anti-crack base fabric according to Embodiment 2 of the present invention.

[0055] Figure 5 This is a schematic diagram of the trajectory reference device of the unmanned laying vehicle for anti-crack base fabric according to Embodiment 2 of the present invention.

[0056] Figure 6 This is a schematic diagram of the travel control module of the unmanned laying vehicle for anti-crack base fabric according to Embodiment 2 of the present invention.

[0057] Explanation of reference numerals in the attached drawings: 1. Laying mechanism; 11. Powered traveling machinery; 111. Pole sleeve; 12. Support frame; 121. Vertical frame; 122. Horizontal frame; 123. Connecting rod; 124. Damping telescopic sleeve; 13. Roll material roller; 14. Tensioning roller; 15. Paving rubber wheel; 2. Track recognition component; 21. Monocular video acquisition device; 22. Pan-tilt unit; 23. Track reference device; 24. Telescopic connector; 3. Travel control module; 31. Steering controller; 32. Travel controller; 33. Brake controller; 34. Early warning unit; 35. Environmental monitor. Detailed Implementation

[0058] The following describes specific embodiments and appendices. Figure 1-6 The invention is described in detail so that those skilled in the art can more fully understand its purpose, features and effects.

[0059] Unless otherwise specified, all technical and scientific terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. In the event of any discrepancy between the definitions of terms in this invention and their commonly understood meaning by one of ordinary skill in the art to which this invention pertains, the definitions provided in this invention shall prevail.

[0060] This invention provides a method for controlling the travel path based on monocular vision recognition and an unmanned laying vehicle for crack-resistant base fabric, which improves the existing crack-resistant base fabric laying method. It automatically adjusts the travel path of the crack-resistant base fabric laying, effectively solves the problems of manual laying of crack-resistant base fabric, and improves the flatness and efficiency of the laying of crack-resistant base fabric. Example 1

[0061] As a specific embodiment of the present invention, this embodiment provides a method for controlling a walking path based on monocular vision recognition, referring to... Figure 1 The specific steps are as follows:

[0062] S100. Use a monocular video acquisition device to collect video information of the target reference object on the anti-crack base cloth laying device and convert it into image data, and determine the initial center coordinate position of the target reference object in the image. The target reference object can move horizontally during the movement, so that the movement center coordinate position of the target reference object in the image changes.

[0063] Specifically, in S100, the monocular video acquisition device can be an existing device capable of acquiring video. Preferably, in this embodiment, the monocular video acquisition device is a monocular industrial camera, which acquires video information of the target reference object. The video information acquired by the monocular industrial camera is RGB video information. The video information includes the target reference object, the milled road surface, and the original road surface.

[0064] The RGB video information is then converted into image data. Preferably, in this embodiment, the RGB video information is converted into grayscale image data.

[0065] Furthermore, the grayscale image data is preprocessed to improve its quality. Preferably, contrast-limited adaptive histogram equalization (CLAHE) is used to preprocess the grayscale image to enhance its contrast.

[0066] Specifically, the converted grayscale image data is first neatly divided into multiple 8×8 pixel square regions, and the size of the regions can be adjusted according to the complexity of the road conditions.

[0067] Then, for each of the divided square regions, the grayscale histogram is calculated, and the number of pixels corresponding to each grayscale level is recorded.

[0068] Based on the road surface conditions, a suitable clipping threshold is predetermined. The histogram portion exceeding the clipping threshold is clipped, and the clipped pixels are evenly distributed to all gray levels to effectively limit local contrast.

[0069] After cropping and redistribution, the histogram is equalized. By calculating the cumulative distribution function, the original gray values ​​in the small square area are accurately mapped to new gray values.

[0070] Finally, based on the mapping function of the four adjacent square regions, the final gray value of each pixel in the image is calculated using a bilinear interpolation algorithm, thereby completing the contrast-limited adaptive histogram equalization processing of the entire image and significantly improving the contrast and visual quality of the image.

[0071] By employing contrast-limited adaptive histogram equalization, the detail representation of the target reference object in the image can be significantly enhanced, which is beneficial for subsequent extraction of the target reference object's features in the image.

[0072] Furthermore, to avoid over-enhancement or distortion in certain areas of the image due to contrast-limited adaptive histogram equalization, a Region of Interest (ROI) containing the target reference is selected from the preprocessed grayscale image data, and the image brightness is optimized. Preferably, a gamma correction algorithm is used to adjust the overall brightness of the ROI image. In addition, selecting the ROI region can also reduce the workload of subsequent data processing.

[0073] Specifically, the formula for adjusting the overall brightness of a ROI image using the gamma correction algorithm is as follows:

[0074]

[0075] In the formula, I(x,y) represents the pixel values ​​of the original image. γ represents the pixel value of the corrected image, and γ is the gamma correction coefficient, which is adjusted according to the actual situation to optimize image brightness.

[0076] Preferably, the first frame of the video information is used to determine the initial center coordinates of the target reference object in the image. Since the first frame of the video information typically has a relatively clear image of the target reference object, and under stable environmental conditions, it can ensure that the position and motion features of the target reference object are captured relatively accurately, the ROI region containing the target reference object is selected in the first frame image as the training selection sample for the filter tracker.

[0077] Determine the center coordinates of the ROI region containing the target reference object, and use the center coordinates of the ROI region as the initial center coordinates of the target reference object in the image. At this time, the target reference object is located at the center of the ROI region.

[0078] Determine the video center coordinates of the first frame image, and use the video center coordinates of the first frame image as the preset target center coordinate position. At this time, the distance between the center coordinates of the ROI region and the video center coordinates of the first frame image is the initial distance.

[0079] Furthermore, the training sample set X is obtained by amplification reconstruction-cyclic shift sampling method, and the ROI is multisampled using a cyclic matrix. The resulting sample set constitutes a typical cyclic matrix. This structure can be diagonalized in Fourier space by Discrete Fourier Transform (DFT), thereby transforming matrix operations into simple vector dot product operations, which significantly reduces the complexity of the training sample set in subsequent calculations.

[0080] Preferably, the ROI image is magnified to ensure that the target reference is not decomposed and reconstructed during cyclic shift sampling, thereby ensuring that subsequent feature extraction and target tracking are performed with stable image quality.

[0081] To address the issue that the increased number of pixels within a region of interest (ROI) image after magnification may slow down program execution, a linear interpolation sampling method is used to downsample the magnified ROI image to reduce redundant data. While reducing the amount of data, this method preserves as much of the target's key feature information as possible, including key visual features such as the target's shape, edges, and texture.

[0082] The ROI image is divided into several cells using the FHOG feature extraction method. Each cell consists of 4×4 pixels, and four adjacent cells form a block.

[0083] The Sobel operator is used to convolve with the training sample set X to obtain gradient information in the horizontal and vertical directions. Then, the feature information in each "block" is statistically analyzed according to the FHOG feature classification method, and finally combined to form the feature information of the training sample set X.

[0084] Furthermore, a kernel correlation filter tracker is used to locate the target reference object.

[0085] Specifically, based on the training sample set X, through the training samples Establish Gaussian regression labels within the Fourier domain For the target itself (i.e., the base samples in the training sample set) The value is 1. along with The value decreases with increasing cyclic shift count. (Filter tracker) It was trained using regularized least squares, making the samples... and Gaussian regression labels The squared error is minimized, as shown below:

[0086]

[0087] In the formula, It is a regularization parameter that controls overfitting and prevents... Overfitting during training.

[0088] The unique closed-form solution is:

[0089]

[0090] In the formula, It is a sample The sample matrix formed It is the conjugate transpose, that is , yes The complex conjugate matrix, It is the identity matrix. It is a Gaussian regression label. The vector formed.

[0091] However, typical target tracking tasks are performed in a nonlinear space, requiring samples from a low-dimensional space to be processed. Mapping to a higher-dimensional space. At this point, a mapping can be used. Sample Mapped to Hilbert space, according to the representation theorem, the filter tracker It can be represented as a sample Linear combination:

[0092]

[0093] In the formula, These are the combination coefficients.

[0094] Introduce a kernel function in the dual space:

[0095]

[0096] but The nuclear version The solution is:

[0097]

[0098] In the formula, It is the kernel autocorrelation matrix of the sample, that is:

[0099]

[0100] Utilizing the property that the training sample set X has diagonalizable circumcyclic matrices, matrix operations are transformed into dot product operations in the Fourier domain, ultimately yielding the kernel correlation filter tracker. The closed-form solution is as follows:

[0101]

[0102] In the formula, the symbol This represents the discrete Fourier transform of the variable. It is a sample Kernel autocorrelation matrix The first line in the text.

[0103] Using a kernel correlation filter tracer The position of the target reference object in the image is detected and located, and the initial center coordinate position is determined. The initial center coordinate position is represented by coordinate values.

[0104] S200. Adjust the video information acquisition angle of the monocular image acquisition device for the target reference object according to the initial center coordinate position of the target reference object in the image, so that the distance between the initial center coordinate position of the target reference object in the image and the preset target center coordinate position in the image meets the preset initial distance.

[0105] The initial distance can be adjusted by changing the angle at which the monocular image acquirer collects video information of the target reference object until the initial distance meets the preset initial distance, at which point no further adjustment is needed.

[0106] Preferably, in this embodiment, the center coordinates of the ROI region in the first frame image coincide with the video center coordinates of the first frame image. At this time, the preset target center coordinate position is the center of the image, and the initial distance is 0, thus facilitating the adjustment of the video information acquisition angle of the monocular image acquisition device for the target reference object. By adjusting the acquisition angle of the monocular industrial camera, the initial center coordinate position of the target reference object in the image coincides with the center of the image.

[0107] S300. In real time, acquire the travel center coordinate position of the target reference object in the image, and calculate the deviation between the travel center coordinate position of the target reference object in the image and the target center coordinate position.

[0108] During the laying of the anti-crack base fabric, the monocular industrial camera acquires video information of the target reference object in real time and converts it into image data. For each frame of the image, the filter tracker obtained in S100 acquires the travel center coordinates of the target reference object in the image. The travel center coordinates refer to the real-time center coordinates of the target reference object in the image during the laying of the anti-crack base fabric. Then, the travel center coordinates of the target reference object in the image are compared and the difference is calculated to obtain the deviation of the travel path of the anti-crack base fabric laying equipment.

[0109] For example, during the laying of anti-crack base fabric, for the second detection image, a filter tracker is used to detect the processed second frame image of the video. The tracker automatically generates a new ROI region for the detected trajectory. At this time, the center coordinate position of the ROI region is the travel center coordinate position of the target reference object. The distance between the travel center coordinate position and the preset target center coordinate position is calculated and compared with the preset initial distance to obtain the travel deviation.

[0110] S400. Based on the calculated deviation between the travel center coordinates of the target reference object in the image and the target center coordinates, the travel path of the anti-crack base fabric laying device is controlled.

[0111] Optionally, in this embodiment, the center coordinates of the ROI region in the first frame image are set to coincide with the video center coordinates of the first frame image, with an initial distance of 0. When the deviation between the traveling center coordinate position (the center coordinates of the ROI region in each frame image) and the target center coordinate position (the video center coordinates) is 0, that is, when the traveling center coordinate position coincides with the target center coordinate position, it indicates that the traveling path has not deviated, and the anti-crack base fabric laying device continues to travel along the existing path; when there is a deviation between the traveling center coordinate position and the target center coordinate position, it indicates that the traveling path has deviated, and the traveling path needs to be adjusted. The traveling path of the anti-crack base fabric laying device is adjusted according to the deviation value until the traveling center coordinate position of the target reference object in the image coincides with the target center coordinate position, and the traveling center coordinate position of the target reference object in the image no longer deviates from the image center. At this point, the adjustment stops, and the traveling center coordinate position of the target reference object in the image is monitored in real time. This process is repeated once after each frame is completed.

[0112] In other feasible embodiments, the initial distance may not be 0, that is, the center coordinates of the ROI region of the first frame image do not coincide with the video center coordinates of the first frame image.

[0113] The present invention provides a travel path control method based on monocular vision recognition. This method uses monocular vision recognition to obtain the travel center coordinates of the target reference object in the image in real time. These coordinates are then compared and the difference is calculated with a preset target center coordinate position to obtain the deviation between the travel center coordinates of the target reference object in the image and the preset target center coordinate position. Based on this deviation, the travel path of the anti-crack base fabric laying device is controlled, thereby precisely controlling the travel direction of the anti-crack base fabric laying device and ensuring that the anti-crack base fabric is laid in the correct direction, significantly improving laying accuracy.

[0114] Furthermore, while using a binocular camera mounted on the anti-crack base fabric laying equipment to calibrate the travel path during construction is feasible, the binocular camera system requires two cameras to work together, maintaining a fixed baseline distance and precise relative position calibration, which complicates installation and maintenance. Additionally, the binocular camera needs to process large amounts of data, performing complex video calibration and parallax calculations, increasing the difficulty of video processing and real-time applications, potentially leading to response delays and affecting laying efficiency. In environments with significant lighting variations, such as when the road surface is shaded or reflective, the binocular camera is prone to misjudgment. Moreover, the failure of any camera in the binocular camera system can cause the entire system to malfunction, affecting the smooth progress of the laying operation. Therefore, the application of binocular cameras in anti-crack base fabric laying has many limitations. This invention employs a travel path control method based on monocular vision recognition, which not only significantly improves laying accuracy but also avoids the complexity of binocular vision systems.

[0115] Example 2

[0116] As a specific embodiment of the present invention, this embodiment provides an unmanned laying vehicle for crack-resistant base fabric. Under the path control of the path control method based on monocular vision recognition provided in Embodiment 1, the crack-resistant base fabric is laid. (Refer to...) Figure 2 , Figure 3 It includes a laying mechanism 1 for smoothly laying the crack-resistant base fabric onto the road surface, a trajectory recognition component 2 for monitoring the travel path, and a travel control module 3 for adjusting the path.

[0117] In this invention, the anti-crack base fabric is laid by the laying mechanism 1. During the laying process, the trajectory recognition component 2 is used to obtain the deviation of the unmanned laying vehicle of the anti-crack base fabric in real time, and the deviation information is fed back to the travel control module 3 to adjust the travel path of the unmanned laying vehicle of the anti-crack base fabric so that the unmanned laying vehicle of the anti-crack base fabric lays the anti-crack base fabric according to the expected travel path, thus ensuring the quality of the anti-crack base fabric laying.

[0118] The laying mechanism 1 includes a powered traveling mechanism 11, a support 12, a roll material roller 13, a tensioning roller 14, and a paving wheel 15. The support 12 is mounted on the powered traveling mechanism 11, and the roll material roller 13, tensioning roller 14, and paving wheel 15 are mounted on the support 12. The roll material roller 13 is used to load the crack-resistant base fabric, which is wound around the roll material roller 13. The tensioning roller 14 is used to tighten the crack-resistant base fabric. When the powered traveling mechanism 11 moves, the crack-resistant base fabric is laid onto the road surface under the coordinated action of the roll material roller 13, tensioning roller 14, and paving wheel 15.

[0119] Specifically, the powered motor 11 refers to a self-moving device with the ability to travel, brake and steer, which can move automatically under the control of the travel control module 3. Existing equipment in the prior art can be used, and this embodiment does not make any improvements to the powered motor 11.

[0120] The support 12 is installed on the front side of the powered traveling machine 11 in the direction of travel. The support 12 includes two opposing vertical tubular supports 121. Each support 121 is fixedly connected to a horizontal tubular crossbeam 122 extending in the direction of travel. Optionally, the crossbeam 122 is welded to the support 121. In actual installation, it is not necessary to ensure that the support 121 is completely vertical and the crossbeam 122 is completely horizontal; a certain angle of inclination between the support 121 and the crossbeam 122 is allowed.

[0121] The upright frame 121 or the horizontal frame 122 is stably connected to the powered travel mechanism 11 via connecting rods 123, thereby enabling the bracket 12 to be mounted on the powered travel mechanism 11. Preferably, in this embodiment, the distance between the two upright frames 121 is greater than the width of the powered travel mechanism 11. The connecting rod 123 is fixed to the side of the upright frame 121 facing the powered travel mechanism 11, and the end of the connecting rod 123 away from the upright frame 121 is stably connected to the powered travel mechanism 11. There are two connecting rods 123. Optionally, the connecting rods 123 are welded to the upright frame 121.

[0122] Optionally, the connecting rod 123 is detachably and fixedly connected to the powered traveling mechanism 11. In this embodiment, a sleeve 111 that mates with the connecting rod 123 is fixed to the side of the powered traveling mechanism 11. The connecting rod 123 is inserted into the sleeve 111, and the connecting rod 123 and the sleeve 111 are detachably and fixedly connected by bolts.

[0123] The roll 13 is located between the upper ends of the opposing uprights 121, and the roll 13 is horizontally arranged. The roll 13 is rotatably connected to the uprights 121. The roll 13 enables the connection to the opposing uprights 121.

[0124] The tension roller 14 is located between the ends of the opposing crossbeams 122 away from the uprights 121, and the tension roller 14 is horizontally arranged. The tension roller 14 is rotatably connected to the crossbeams 122. The tension roller 14 enables the connection to the opposing crossbeams 122.

[0125] Optionally, both the roll material roller 13 and the tensioning roller 14 are steel rods. The rotatable connection structure between the roll material roller 13 and the upright frame 121, and the rotatable connection structure between the tensioning roller 14 and the cross frame 122, can adopt existing structures in the prior art, and will not be described in detail in this embodiment.

[0126] Combination Figure 4 A vertical damping telescopic sleeve 124 is fixed to the bottom end of the upright 121, and the damping telescopic sleeve 124 is provided at the bottom end of both uprights 121. Optionally, the damping telescopic sleeve 124 is welded to the bottom end of the upright 121. Optionally, the damping telescopic sleeve 124 is a hydraulic sleeve, which is a sleeve with a certain damping and can extend and retract.

[0127] The paving roller 15 is located between the lower ends of the opposing damping telescopic sleeve 124. The paving roller 15 is rotatably connected to the damping telescopic sleeve 124. Pressure is applied to the paving roller 15 by the damping telescopic sleeve 124, allowing the paving roller 15 to adhere tightly to the ground, thus ensuring that the anti-crack base fabric is laid flat and firmly during operation. The rotatable connection structure between the paving roller 15 and the damping telescopic sleeve 124 can adopt existing structures in the prior art, and will not be described in detail in this embodiment.

[0128] The anti-crack base fabric is installed on the roll roller 13. When laying the anti-crack base fabric, first use a jack to lift the lower end of the damping telescopic sleeve 124, causing the spreading rubber roller 15 to leave the ground. Then, pull the anti-crack base fabric, passing it around the tension roller 14 and through the bottom of the spreading rubber roller 15. Then release and remove the jack. Manually rotate the roll roller 13 in the opposite direction to make the anti-crack base fabric surface flat. Afterwards, start the power-driven traveling mechanism 11 to lay the anti-crack base fabric. Figure 3 .

[0129] The damping telescopic sleeve 124 applies a certain pressure to the paving roller 15 to ensure that the anti-crack base fabric remains flat and firm during the laying process, thus avoiding uneven laying caused by insufficient pressure.

[0130] The trajectory recognition component 2 includes a monocular video acquisition unit 21, a gimbal 22, a trajectory reference unit 23, a telescopic connector 24, and an image data processing unit.

[0131] Specifically, the monocular video acquisition device 21 is mounted on the upper end of the side of the support frame 121 of the bracket 12 opposite to the roll material 13 via a pan-tilt unit 22. The pan-tilt unit 22 is fixedly connected to the support frame 121, and the monocular video acquisition device 21 is rotatably mounted on the pan-tilt unit 22. Optionally, the monocular video acquisition device 21 is a monocular industrial camera.

[0132] The trajectory reference device 23 is located on the same side of the support frame 121 and below the monocular video acquisition device 21. The trajectory reference device 23 is fixed to the side of the damping telescopic sleeve 124 via the telescopic connector 24. Optionally, the telescopic connector 24 is a hydraulic rod with only horizontal freedom, perpendicular to the direction of travel. During the laying of the anti-crack base fabric, the side of the trajectory reference device 23 away from the telescopic connector 24 is pressed against the side of the junction between the milled road surface and the original road surface. The trajectory reference device 23 serves as a target reference for trajectory control.

[0133] In the initial starting position, the hydraulic rod is in a certain compressed state, ensuring that the trajectory reference device 23 can still rest against the side of the junction between the milled and original road surfaces when it is away from the intersection. That is, when the powered travel mechanism 11 approaches the side of the junction between the milled and original road surfaces, it compresses the hydraulic rod, shortening it, and the trajectory reference device 23 is tightly against the side of the junction. When the powered travel mechanism 11 moves away from the side of the junction, the hydraulic rod extends, ensuring that the trajectory reference device 23 remains tightly against the side of the junction. Optionally, the trajectory recognition component 2 can be installed on one or both sides as needed; typically, installation on only one side is sufficient to achieve the desired effect.

[0134] The monocular video acquisition device 21 is used to acquire video information containing the trajectory reference device 23. The working angle of the monocular video acquisition device 21 is adjusted by the gimbal 22 to ensure that the trajectory reference device 23 remains within the video range. In addition, the gimbal 22 has an image stabilization function to ensure that the monocular video acquisition device 21 can acquire video information smoothly.

[0135] The video information acquired by the monocular video acquisition device 21 is transmitted to the image data processing unit through a shielded video signal line. After receiving the video information containing the trajectory reference device 23, the image data processing unit processes the received video information using the steps described in Embodiment 1. It then calculates the deviation between the coordinates of the trajectory reference device 23's travel center and the coordinates of the target center in the image, obtaining the lateral deviation value of the traveling direction of the powered traveling machine 11 relative to the original path. This deviation information is then fed back to the travel control module 3, which adjusts the travel path of the powered traveling machine 11 in real time. Figure 5 Show.

[0136] By determining the position of the trajectory reference 23 in the image based on monocular vision recognition and calculating its deviation from the target center coordinate position, the travel direction of the powered traveling machine 11 can be precisely controlled, ensuring that the anti-crack base fabric can be laid in the correct direction, which can significantly improve the laying accuracy.

[0137] Combination Figure 1 The travel control module 3 is mounted on the powered travel mechanism 11. The travel control module 3 is connected to the image data processing unit via a shielded signal cable.

[0138] Specifically, refer to Figure 6 The travel control module 3 includes a steering controller 31, a travel controller 32, and a brake controller 33. The travel control module 3 controls the steering controller 31 based on the deviation information provided by the image data processing unit, adjusting the travel direction of the powered travel machinery 11 to ensure that the travel direction of the powered travel machinery 11 remains parallel to the road reference line (i.e., the side of the junction between the milled road surface and the original road surface). Figure 5 Show.

[0139] The travel control module 3 controls the travel controller 32 and the brake controller 33 to make the powered travel machine 11 move or stop.

[0140] Automated laying replaces the traditional manual laying method, reducing the need for human intervention and improving the efficiency of laying operations; at the same time, through real-time monitoring and adjustment based on monocular vision, it can quickly respond to and correct laying deviations, further shortening the laying time.

[0141] Furthermore, the travel control module 3 also includes an early warning unit 34 and an environmental monitor 35. Optionally, the environmental monitor 35 is a camera. The environmental monitor 35 is deployed around the unmanned anti-crack base fabric laying vehicle to monitor the environment around the vehicle in real time. Figure 2 As shown, in this embodiment, the environmental monitor 35 is installed on the powered motor 11 and the support 12. Specifically, the environmental monitor 35 is installed on the rear end of the powered motor 11 and on the crossbeam 122 of the support 12.

[0142] By adjusting the working angles of the multiple environmental monitors 35, it is ensured that the image within 30cm around the anti-crack base fabric laying mechanism 1 can be clearly displayed in the video data.

[0143] The early warning unit 34 is equipped with an environmental perception system based on the YOLO algorithm. For road construction scenarios, the YOLO algorithm is specifically trained using obstacles such as construction workers, vehicles, and safety cones as training samples. Data augmentation enhances the model's generalization ability and optimizes obstacle localization accuracy. A dynamic anchor-frame matching strategy further enhances the detection effect of small targets. Ultimately, a detection model that can automatically identify obstacles such as personnel, vehicles, and safety cones in the construction environment is constructed, outputting their location coordinates and category information in real time, providing reliable environmental perception support for the travel control module 3.

[0144] Based on the road safety information monitored by the environmental monitor 35, the travel control module 3 can further control the working status of the travel controller 32 and the brake controller 33 to ensure the adaptability of the paving equipment to the surrounding environment.

[0145] After receiving the surrounding environment video collected by the environmental monitor 35, the early warning unit 34 analyzes the video information using a trained YOLO detection model. When the video information from the environmental monitor 35 detects a potential safety hazard in the surrounding area, the travel control module 3 will feed back to the brake controller 33 to apply the brakes. If the surrounding environment is safe, the brakes will be released and the travel controller 32 will be controlled to move the powered travel machine 11 forward.

[0146] After the anti-crack base fabric laying work begins, multiple environmental monitors 35 use a dynamic detection program to monitor the situation within 30cm around the anti-crack base fabric laying mechanism 1 in real time. When a dangerous object or other safety hazard appears in the field of view of any of the multiple environmental monitors 35, the travel control module 3 will trigger the brake controller 33 to perform emergency braking to prevent the equipment from colliding or other safety accidents. Once the surrounding environment is confirmed to be safe, the brake is released, and the travel controller 32 is used to control the travel mechanism 11 to resume its movement.

[0147] The travel control module 3 is equipped with multiple environmental monitors 35, which can monitor the environmental safety around the laying mechanism 1 in real time. When a potential risk is detected, the travel control module 3 can immediately take braking measures to prevent the equipment from colliding or other safety accidents, thereby improving the safety of equipment operation.

[0148] The travel control module 3 automatically adjusts the travel direction of the powered walking machinery based on the deviation information provided by the image data processing unit, and adjusts the travel and braking of the powered walking machinery 11 based on the monitoring data of the environmental monitor 35, thereby achieving precise control of unmanned crack-resistant base fabric paving operation.

[0149] Furthermore, to make it easier for operators to adjust the working angles of the monocular video acquisition device 21 and the environmental monitor 35, the video information collected by the monocular video acquisition device 21 and the environmental monitor 35 is wirelessly transmitted to a mobile terminal, and the working angles of the monocular video acquisition device 21 and the environmental monitor 35 are adjusted through the mobile terminal.

[0150] The unmanned laying vehicle for crack-resistant base fabric of the present invention can adjust the laying direction according to real-time video information, automatically correct the travel path, and control the movement and braking according to the monitoring video information of multiple environmental monitors 35. This can significantly improve the flatness and efficiency of crack-resistant base fabric laying, enhance the safety of the laying process, effectively reduce labor costs, and further improve the construction quality and efficiency of road maintenance projects.

[0151] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any modifications or equivalent changes made based on the technical essence of the present invention shall still fall within the scope of protection claimed by the present invention.

Claims

1. A method for controlling a travel path based on monocular vision recognition, characterized in that, The method includes: The RGB video information of the target reference object on the anti-crack base cloth laying equipment is acquired using a monocular video acquisition device and converted into grayscale image data. Contrast-limited adaptive histogram equalization is used to preprocess the grayscale image data to improve its quality. The initial center coordinate position of the target reference object in the preprocessed grayscale image is determined. The video information contains information about the target reference object, the milled road surface, and the original road surface. The monocular video acquisition device is a monocular industrial camera. During the movement of the target reference object, it can move horizontally due to the influence of the side of the junction between the milled road surface and the original road surface, which causes the movement center coordinate position of the target reference object in the image to change dynamically. After preprocessing the grayscale image using contrast-limited adaptive histogram equalization, a Region of Interest (ROI) containing the target reference is selected from the preprocessed grayscale image, and a gamma correction algorithm is used to adjust the overall brightness of the ROI image. The formula for adjusting the overall brightness of the ROI image using the gamma correction algorithm is as follows: In the formula, I(x,y) represents the pixel values ​​of the original image. (x,y) represents the pixel value of the corrected image, and γ is the gamma correction coefficient; Determining the initial center coordinates of the target reference object in the preprocessed grayscale image includes: In the first frame of the video information, a region of interest (ROI) containing the target reference is selected as a training sample for the filter tracker. The training sample set X is obtained by amplification reconstruction-cyclic shift sampling method, and the ROI is multisampled using a cyclic matrix. The resulting sample set constitutes a typical cyclic matrix. The Sobel operator is used to convolve with the training sample set X to obtain gradient information in the horizontal and vertical directions. Then, the feature information in each block is statistically analyzed according to the FHOG feature classification method, and finally combined to form the feature information of the training sample set X. The block is obtained by dividing the ROI region using the FHOG feature extraction method, and each block is composed of multiple adjacent cells. Based on the training sample set X, through the training samples Establish Gaussian regression labels within the Fourier domain , along with It decreases as the number of cyclic shifts increases, where Filter tracker Training is performed using regularized least squares, so that the samples... and Gaussian regression labels The squared error is minimized: In the formula, It is a regularization parameter that controls overfitting and prevents... Overfitting during training; The unique closed-form solution is: In the formula, It is a sample The sample matrix formed It is the conjugate transpose, that is , yes The complex conjugate matrix, It is the identity matrix. It is a Gaussian regression label. The vector formed; Using mapping Sample Mapped to Hilbert space, filter tracker Represented as a sample Linear combination: In the formula, These are the combination coefficients; Introduce a kernel function in the dual space: but The nuclear version The solution is: In the formula, This is the kernel autocorrelation matrix of the sample: Utilizing the property that the training sample set X has diagonalizable circumcyclic matrices, matrix operations are transformed into dot product operations in the Fourier domain, ultimately yielding the kernel correlation filter tracker. The closed-form solution is as follows: In the formula, the symbol Represents the discrete Fourier transform of the variables. It is a sample Kernel autocorrelation matrix The first line in; Using the kernel correlation filter tracker The position of the target reference object in the image is detected and located; The monocular video acquisition device adjusts the video information acquisition angle of the target reference object based on the initial center coordinate position of the target reference object in the image, so that the initial center coordinate position of the target reference object in the image coincides with the preset target center coordinate position in the image. During the laying of the crack-resistant base fabric, the monocular video acquisition device acquires video information of the target reference object in real time and converts it into image data, which is then processed by the kernel correlation filter tracker. The travel center coordinates of the target reference object in the ROI region of the image are obtained in real time, and the deviation between the travel center coordinates of the target reference object in the image and the target center coordinates is calculated to obtain the lateral deviation value of the travel direction of the anti-crack base fabric laying equipment relative to the original path. The travel path of the anti-crack base fabric laying device is controlled based on the calculated deviation between the travel center coordinates of the target reference object in the image and the target center coordinates.

2. The method for controlling a travel path based on monocular vision recognition according to claim 1, characterized in that, The preprocessing of the grayscale image data using contrast-limited adaptive histogram equalization includes: The grayscale image data is divided into multiple 8×8 pixel square regions; For each of the square regions, its grayscale histogram is calculated, and the number of pixels corresponding to each grayscale level is recorded. According to the preset cropping threshold, the portion of the grayscale histogram that exceeds the cropping threshold is cropped, and the cropped pixels are evenly distributed to all grayscale levels; An equalization operation is performed on the grayscale histogram after cropping and redistribution. By calculating the cumulative distribution function, the original grayscale values ​​in the small square area are mapped to new grayscale values. Based on the grayscale value mapping function of the four adjacent square regions, the final grayscale value of each pixel in the grayscale image data is calculated using a bilinear interpolation algorithm, thus completing the contrast-limited adaptive histogram equalization processing.

3. The method for controlling a travel path based on monocular vision recognition according to claim 2, characterized in that, In the FHOG feature classification method, the ROI region is divided into 4×4 pixel cell units, and each block unit consists of 4 adjacent cell units.

4. A crack-resistant base fabric unmanned laying vehicle, characterized in that, The driving path control method based on monocular vision recognition as described in any one of claims 1-3 is used for driving path control.

5. The unmanned laying vehicle for anti-crack base fabric according to claim 4, characterized in that, The system includes a laying mechanism (1) for laying the crack-resistant base fabric flat on the road surface, a trajectory recognition component (2) for monitoring the travel path, and a travel control module (3) for adjusting the path. The trajectory recognition component (2) acquires the travel path deviation of the unmanned crack-resistant base fabric laying vehicle in real time and feeds back the deviation information to the travel control module (3) to adjust the travel path of the unmanned crack-resistant base fabric laying vehicle so that the unmanned crack-resistant base fabric laying vehicle lays the crack-resistant base fabric according to the expected travel path.

6. The unmanned laying vehicle for anti-crack base fabric according to claim 5, characterized in that, The laying mechanism (1) includes a power-driven machine (11), a support (12), a roll material roller (13), a tensioning roller (14), and a paving wheel (15). The roll material roller (13), the tensioning roller (14), and the paving wheel (15) are mounted on the support (12). The support (12) is fixedly connected to the sleeve (111) of the power-driven machine (11) through a connecting rod (123). The bottom of the support (12) is provided with a damping telescopic sleeve (124), and the paving wheel (15) is pressed tightly against the ground through the damping telescopic sleeve (124).

7. The unmanned laying vehicle for anti-crack base fabric according to claim 6, characterized in that, The trajectory recognition component (2) includes a monocular video acquisition device (21), a gimbal (22), a trajectory reference device (23), a telescopic connector (24), and an image data processing unit. The trajectory reference device (23) serves as a target reference object. The monocular video acquisition device (21) is mounted on the support (12) of the laying mechanism (1) via the gimbal (22). The trajectory reference device (23) is mounted on the support (12) of the laying mechanism (1) via the telescopic connector (24). The trajectory reference device (23) and the monocular video acquisition device (21) are located on the same side of the support (12), and the trajectory reference device (23) is located below the monocular video acquisition device (21). When laying the anti-crack base cloth, the side of the trajectory reference device (23) away from the telescopic connector (24) is close to the side of the junction between the milled road surface and the original road surface. The monocular video acquisition device (21) is used to acquire video information containing the trajectory reference device (23). The length of the telescopic connector (24) is affected by the side of the junction between the milled road surface and the original road surface. During the paving process, the dynamic change causes the position of the trajectory reference device (23) in the video to change. The acquired video information is transmitted to the image data processing unit for processing. The deviation between the coordinate position of the travel center of the trajectory reference device (23) in the image and the coordinate position of the target center is calculated. The lateral deviation value of the travel direction of the power travel machinery (11) relative to the original path is obtained. The deviation information is fed back to the travel control module (3) to adjust the travel path of the paving mechanism (1) in real time.

8. The unmanned laying vehicle for anti-crack base fabric according to claim 7, characterized in that, The travel control module (3) includes a steering controller (31), a travel controller (32), and a brake controller (33). The travel control module (3) controls the steering controller (31) to work based on the deviation information provided by the image data processing unit, and adjusts the travel direction of the laying mechanism (1). The travel control module (3) controls the travel controller (32) and the brake controller (33) to work, so that the laying mechanism (1) can travel or stop.

9. The unmanned laying vehicle for anti-crack base fabric according to claim 8, characterized in that, The travel control module (3) also includes an early warning unit (34) and multiple environmental monitors (35). The early warning unit (34) is equipped with a pre-trained YOLO detection model for obstacles. The environmental monitors (35) are used to monitor the environment around the unmanned laying vehicle for crack-resistant base fabric in real time. After receiving the surrounding environment video collected by the environmental monitors (35), the early warning unit (34) uses the YOLO detection model to analyze the video information. When the video information of the environmental monitors (35) detects a safety hazard in the surrounding area, the travel control module (3) will feed back to the brake controller (33) to brake. If the surrounding environment is safe, the brake will be released and the travel controller (32) will be controlled to make the laying mechanism (1) move forward.

Images