A road subgrade flatness detection method, system, device and program

Abstract

The present application belongs to the technical field of machine vision detection, and specifically discloses a road subgrade flatness detection method, system, device and program, which performs image cropping and conversion processing on the gap detection image between the road subgrade surface and the straight ruler, obtains a gap detection segmented binary image, then performs gap detection and anchor gap region feature extraction on the gap detection segmented binary image, determines the three-dimensional gap features of each gap detection segmented binary image, further assembles a three-dimensional gap feature set of the detection image, and finally uses the three-dimensional gap feature set for neural network flatness detection to determine the subgrade flatness of the corresponding straight ruler section. The present application can efficiently detect the gap distribution features between the road subgrade surface and the straight ruler through machine vision detection technology, and further accurately determine the flatness of the road subgrade by using the gap distribution features, thereby improving the efficiency and reliability of the road subgrade flatness detection.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of machine vision detection, and particularly relates to a road subgrade flatness detection method, system, device and program. BACKGROUND

[0002] The road surface flatness is one of the main technical indexes for evaluating the road surface quality, which is related to the safety and comfort of driving, the impact force on the road surface and the service life. The uneven road surface will increase the driving resistance and cause additional vibration of the vehicle, which will cause the driving to be bumpy, affect the driving speed and safety, and affect the driving stability and passenger comfort. At the same time, the vibration will also exert an impact force on the road surface, thereby accelerating the damage of the road surface and the vehicle parts and the wear of the tires, and increasing the energy consumption. In addition, the uneven road surface will also accumulate rainwater and accelerate the damage of the road surface.

[0003] The road surface flatness refers to the longitudinal concave-convex deviation of the road surface, and mainly reflects the flatness of the longitudinal section profile curve of the road surface. When the longitudinal section profile curve of the road surface is relatively smooth, it indicates that the road surface is relatively flat, or the flatness is relatively good, and vice versa. The straight ruler method is a common road surface flatness detection method at present, that is, a straight ruler with a specified length is placed on the road surface, and the gap between the straight ruler and the road surface is measured manually to determine the flatness of the road surface. This flatness detection method relies on manual measurement experience, and the efficiency and reliability of the detection need to be improved. SUMMARY

[0004] The present application provides a road subgrade flatness detection method, system, device and program to solve the above problems in the prior art.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical scheme:

[0006] In a first aspect, a road subgrade flatness detection method is provided, comprising:

[0007] An initial gap detection image between the road subgrade surface and the straight ruler collected by the image collector along the length direction of the straight ruler is acquired, and the initial gap detection image contains the road subgrade surface and the straight ruler;

[0008] The initial gap detection image is subjected to image rotation, scaling and cropping processing to obtain a gap detection image with a fixed direction and a fixed size, and the gap detection image is cut into a plurality of gap detection sub-images based on the scale marks of the straight ruler on the gap detection image;

[0009] The gap detection segmented image is subjected to gray scale processing to obtain a corresponding gap detection segmented smooth gray scale image, and the gap detection segmented smooth gray scale image is subjected to image binarization processing to obtain a corresponding gap detection segmented binary image;

[0010] The gap detection segmented binary image is subjected to edge detection to determine each edge-closed gap region in the gap detection segmented binary image, and the corresponding anchor gap region is extracted from each edge-closed gap region;

[0011] The direction perpendicular to the flat ruler in the gap detection segmented binary image is determined as the vertical direction, and the direction parallel to the flat ruler is determined as the horizontal direction, and the vertical gap parameter and the horizontal gap parameter of each anchor gap region in the gap detection segmented binary image are determined;

[0012] Based on the number of anchor gap regions and the vertical gap parameter and the horizontal gap parameter of each anchor gap region, a three-dimensional gap feature of the corresponding gap detection segmented binary image is constructed;

[0013] The three-dimensional gap features of the gap detection segmented binary images are used to form a three-dimensional gap feature set, and the three-dimensional gap feature set is input into a preset flatness detection model for flatness detection to obtain a flatness detection result of the corresponding road embankment section under the flat ruler, and the flatness detection result is output.

[0014] In one possible design, the gray scale processing of the gap detection segmented image to obtain the corresponding gap detection segmented smooth gray scale image, and the image binarization processing of the gap detection segmented smooth gray scale image to obtain the corresponding gap detection segmented binary image, includes:

[0015] The three-color channel component values R, G and B of each pixel point of the gap detection segmented image are substituted into a gray scale conversion formula H=0.39*R+0.50*G+0.11*B to obtain the gray scale value H of each pixel point, and the gap detection segmented gray scale image is constructed based on the gray scale value H of each pixel point;

[0016] The gap detection segmented gray scale image is subjected to median filtering processing to obtain the gap detection segmented smooth gray scale image;

[0017] The maximum inter-class variance method is used for the image binarization processing of the gap detection segmented smooth gray scale image to obtain the gap detection segmented binary image.

[0018] In one possible design, the edge detection of the gap detection segmented binary image to determine each edge-closed gap region in the gap detection segmented binary image includes:

[0019] The Canny edge detection operator is used to perform edge detection on the gap detection segmented binary image, and each edge-closed gap region in the gap detection segmented binary image is determined.

[0020] In a possible design, the extracting of the corresponding anchor gap region from each edge-closed gap region comprises:

[0021] The number of pixels contained in each pixel column in the edge-closed gap region is determined, and the number of pixels is taken as an anchor value of the corresponding pixel column.

[0022] A region composed of pixel columns in which a plurality of consecutive anchor values are greater than a set reference threshold in the edge-closed gap region is taken as an anchor gap region.

[0023] In a possible design, the determining of the vertical gap parameter and the horizontal gap parameter of each anchor gap region in the gap detection segmented binary image comprises:

[0024] The maximum image interval of the anchor gap region in the vertical direction is determined, and the maximum image interval of the anchor gap region in the vertical direction is taken as the vertical gap parameter of the anchor gap region.

[0025] The maximum image interval of the anchor gap region in the horizontal direction is determined, and the maximum image interval of the anchor gap region in the horizontal direction is taken as the horizontal gap parameter of the anchor gap region.

[0026] In a possible design, the constructing of the three-dimensional gap feature of the corresponding gap detection segmented binary image based on the number of anchor gap regions contained and the vertical gap parameter and the horizontal gap parameter of each anchor gap region comprises:

[0027] The number L of anchor gap regions contained in the gap detection segmented binary image is determined, the maximum value M of the vertical gap parameter of each anchor gap region in the gap detection segmented binary image is determined, and the maximum value N of the horizontal gap parameter of each anchor gap region in the gap detection segmented binary image is determined.

[0028] The three-dimensional gap feature (L, M, N) of the gap detection segmented binary image is constructed by using L, M, and N.

[0029] In a possible design, before the three-dimensional gap feature set is input into the preset flatness detection model for flatness detection, the method further comprises:

[0030] An initialized BP neural network model is constructed, and the initialized BP neural network model is trained and optimized by using a training set until a set training condition is met, to obtain a trained flatness detection model, wherein the training set contains a plurality of three-dimensional gap feature set samples, and each three-dimensional gap feature set sample is labeled with a corresponding flatness label.

[0031] In a second aspect, a road subgrade flatness detection system is provided, comprising an image acquisition unit, an image segmentation unit, an image conversion unit, a gap detection unit, a parameter determination unit, a feature construction unit, and a flatness detection unit, wherein:

[0032] The image acquisition unit is configured to acquire an initial gap detection image between a road subgrade surface and a straight ruler captured by an image acquisition device along a length direction of the straight ruler, and the initial gap detection image comprises the road subgrade surface and the straight ruler.

[0033] The image segmentation unit is configured to perform image rotation, scaling, and cropping on the initial gap detection image to obtain a gap detection image with a fixed direction and a fixed size, and cut the gap detection image into a plurality of gap detection sub-images based on scale marks of the straight ruler on the gap detection image.

[0034] The image conversion unit is configured to perform grayscale processing on each gap detection sub-image to obtain a corresponding gap detection sub-smooth grayscale image, and perform image binarization processing on each gap detection sub-smooth grayscale image to obtain a corresponding gap detection sub-binary image.

[0035] The gap detection unit is configured to perform edge detection on the gap detection sub-binary image, determine each edge-closed gap region in the gap detection sub-binary image, and extract a corresponding anchor gap region from each edge-closed gap region.

[0036] The parameter determination unit is configured to determine a vertical direction perpendicular to the straight ruler and a horizontal direction parallel to the straight ruler in the gap detection sub-binary image, and determine a vertical gap parameter and a horizontal gap parameter of each anchor gap region in the gap detection sub-binary image.

[0037] The feature construction unit is configured to construct a three-dimensional gap feature of the corresponding gap detection sub-binary image based on a number of anchor gap regions and the vertical gap parameter and the horizontal gap parameter of each anchor gap region.

[0038] The flatness detection unit is configured to use the three-dimensional gap features of each gap detection sub-binary image to form a three-dimensional gap feature set, input the three-dimensional gap feature set into a preset flatness detection model for flatness detection, obtain a flatness detection result of a corresponding road subgrade section under the straight ruler, and output the flatness detection result.

[0039] In a third aspect, a road subgrade flatness detection device is provided, comprising:

[0040] A memory configured to store instructions.

[0041] A processor is configured to read instructions stored in the memory and execute the method according to any one of the first aspect.

[0042] In a fourth aspect, a computer readable storage medium is provided, and the computer readable storage medium stores instructions, and when the instructions are executed on a computer, the computer is caused to execute the method according to any one of the first aspect. Meanwhile, a computer program is also provided, and when the computer program is executed on a computer, the method according to any one of the first aspect is executed.

[0043] Beneficial effects: the present application obtains the gap detection segmented binary image by collecting the gap detection image between the road subgrade surface and the flat ruler, and then performing image cropping and conversion processing, and then performs gap detection and anchor gap region feature extraction on the gap detection segmented binary image, determines the three-dimensional gap features of each gap detection segmented binary image, and then assembles the three-dimensional gap feature set of the detection image, and then uses the three-dimensional gap feature set to detect the flatness, and determines the subgrade flatness of the corresponding flat ruler section, so that the rapid and intelligent road subgrade flatness detection can be realized. The machine vision detection technology can efficiently detect the gap distribution features between the road subgrade surface and the flat ruler, and then accurately determine the flatness of the road subgrade by using the gap distribution features, so that the traditional manual fixed-length ruler detection method can be replaced, and the efficiency and reliability of the road subgrade flatness detection can be improved. BRIEF DESCRIPTION OF DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0045] Figure 1 The figure is a schematic diagram of the steps of the method in embodiment 1 of the present application.

[0046] Figure 2 The figure is a schematic diagram of the system in embodiment 2 of the present application.

[0047] Figure 3 The figure is a schematic diagram of the system in embodiment 2 of the present application. DETAILED DESCRIPTION

[0048] It is to be understood that the embodiments are merely illustrative of the principles of the present application. Certain structural and functional details are set forth in order to provide a thorough understanding of the example embodiments described herein. However, the present application can be embodied in many different forms and should not be construed as limited to the embodiments set forth herein.

[0049] It should be understood that, unless otherwise explicitly specified and limited, the corresponding terms should be interpreted broadly, for example, "connection" can be fixed connection, can also be detachable connection, or integrally connected; can be electrical connection, can be directly connected, can also be indirectly connected through intermediate medium, can be internal communication of two elements. For those skilled in the art, the specific meaning of the above terms in the embodiments can be understood according to the specific circumstances.

[0050] In the following description, specific details are provided to facilitate a thorough understanding of the example embodiments. However, those skilled in the art will understand that the example embodiments can be practiced without these specific details. For example, devices can be shown in block diagrams in order not to obscure the examples with unnecessary detail. In other embodiments, well-known processes, structures, and techniques can not be shown in detail in order not to obscure the embodiments.

[0051] Embodiment 1:

[0052] The embodiment provides a road subgrade flatness detection method, which can be applied to a corresponding image processing terminal, such as Figure 1 As shown in the figure, the method comprises the following steps:

[0053] S1. Obtain an initial gap detection image between the road subgrade surface and the straight ruler collected by the image collector along the length direction of the straight ruler, wherein the initial gap detection image contains the road subgrade surface and the straight ruler.

[0054] In specific implementation, a straight ruler with scale marks can be placed on the road subgrade in advance. If the road subgrade surface is uneven, a corresponding gap will be formed between the road subgrade surface and the straight ruler. At this time, the image collector can be used to collect the initial gap detection image between the road subgrade surface and the straight ruler by keeping a fixed distance between the straight ruler side and the straight ruler and sticking to the ground along the length direction of the straight ruler. The initial gap detection image contains the road subgrade surface and the straight ruler. The image collector uploads the collected initial gap detection image to the image processing terminal, and subsequent image analysis is performed by the image processing terminal.

[0055] S2. Perform image rotation, scaling and cropping on the initial gap detection image to obtain a gap detection image with fixed direction and fixed size, and cut the gap detection image into a plurality of gap detection sub-images based on the scale marks on the straight ruler in the gap detection image.

[0056] In specific implementation, after obtaining the initial gap detection image, the image processing terminal can perform preprocessing on the initial gap detection image, including image rotation, image scaling and image cropping, to obtain a gap detection image with fixed direction and fixed size (i.e., fixed resolution). Then, the scale marks on the straight ruler in the gap detection image are determined through target detection, and the gap detection image is cut into a plurality of gap detection sub-images based on the scale marks on the straight ruler as the reference boundary.

[0057] S3. Perform gray scale processing on each gap detection sub-image to obtain a corresponding gap detection sub-smooth gray scale image, and perform image binarization processing on each gap detection sub-smooth gray scale image to obtain a corresponding gap detection sub-binary image.

[0058] In specific implementation, the image processing terminal can calculate the gray scale value H of each pixel point of the gap detection sub-image by substituting the three color channel component values R, G and B of each pixel point into the gray scale conversion formula H=0.39×R+0.50×G+0.11×B, and construct a gap detection sub-gray scale image based on the gray scale value H of each pixel point; then perform median filter processing on the gap detection sub-gray scale image to obtain a gap detection sub-smooth gray scale image; and then perform image binarization processing on the gap detection sub-smooth gray scale image using the maximum inter-class variance method to obtain a gap detection sub-binary image.

[0059] S4. Perform edge detection on the gap detection sub-binary image to determine each edge-closed gap region in the gap detection sub-binary image, and extract a corresponding anchor gap region from each edge-closed gap region.

[0060] In specific implementation, the image processing terminal can perform edge detection on the gap detection sub-binary image using the Canny edge detection operator to determine each edge-closed gap region in the gap detection sub-binary image; then determine the number of pixel points contained in each pixel column in the edge-closed gap region, and take the number of pixel points as the anchor value of the corresponding pixel column; and finally take the region composed of the pixel columns with a plurality of anchor values greater than a set reference threshold value as the anchor gap region.

[0061] S5. Take the direction perpendicular to the straight ruler in the gap detection sub-binary image as the vertical direction, and take the direction parallel to the straight ruler as the horizontal direction, and determine the vertical gap parameter and the horizontal gap parameter of each anchor gap region in the gap detection sub-binary image.

[0062] In implementation, the image processing terminal determines the vertical direction in the gap detection segmented binary image as the vertical direction (i.e. the direction perpendicular to the top boundary line of the anchor gap region as the vertical direction), and determines the direction parallel to the flat ruler as the horizontal direction (i.e. the direction parallel to the top boundary line of the anchor gap region as the horizontal direction). Then the image processing terminal determines the maximum image interval of the anchor gap region in the vertical direction, and takes the maximum image interval of the anchor gap region in the vertical direction as the vertical gap parameter of the anchor gap region; and determines the maximum image interval of the anchor gap region in the horizontal direction, and takes the maximum image interval of the anchor gap region in the horizontal direction as the horizontal gap parameter of the anchor gap region.

[0063] S6. Based on the number of anchor gap regions contained and the vertical gap parameter and the horizontal gap parameter of each anchor gap region, a three-dimensional gap feature corresponding to the gap detection segmented binary image is constructed.

[0064] In implementation, the image processing terminal determines the number L of anchor gap regions contained in the gap detection segmented binary image, determines the maximum value M of the vertical gap parameter of each anchor gap region in the gap detection segmented binary image, and determines the maximum value N of the horizontal gap parameter of each anchor gap region in the gap detection segmented binary image. Then the image processing terminal constructs the three-dimensional gap feature (L, M, N) of the gap detection segmented binary image by using L, M and N.

[0065] S7. The three-dimensional gap features of each gap detection segmented binary image are used to form a three-dimensional gap feature set, and the three-dimensional gap feature set is input into a preset flatness detection model for flatness detection to obtain the flatness detection result of the corresponding roadbed section under the flat ruler, and the flatness detection result is output.

[0066] In implementation, the image processing terminal uses the three-dimensional gap features of each gap detection segmented binary image to form a three-dimensional gap feature set { (L1, M1, N1), (L2, M2, N2),..., (L n , M n , N n) and input the three-dimensional gap feature set into a preset flatness detection model to perform flatness detection, to obtain a flatness detection result of the corresponding road subgrade section under the straight ruler. Before performing flatness detection by using the flatness detection model, the image processing terminal pre-constructs an initialized BP neural network model, and trains and optimizes the initialized BP neural network model by using a training set until a set training condition (such as a set training number, a recall rate, etc.) is met, to obtain the trained flatness detection model. The training set contains a plurality of three-dimensional gap feature set samples, and each three-dimensional gap feature set sample is labeled with a corresponding flatness label. The flatness detection result obtained after the flatness detection model performs flatness detection is a corresponding flatness value, and the image processing terminal outputs and displays the flatness detection result, so that a detection personnel can directly view the flatness value of the road subgrade surface under the straight ruler.

[0067] The method of the embodiment can efficiently detect the gap distribution feature between the road subgrade surface and the straight ruler by using machine vision detection technology, and accurately determine the flatness of the road subgrade by using the gap distribution feature. The method can replace the traditional manual length straight ruler detection method, and improve the efficiency and reliability of road subgrade flatness detection.

[0068] Embodiment 2:

[0069] The embodiment provides a road subgrade flatness detection system, as shown in Figure 2 The road subgrade flatness detection system comprises an image acquisition unit, an image segmentation unit, an image conversion unit, a gap detection unit, a parameter determination unit, a feature construction unit, and a flatness detection unit, wherein:

[0070] The image acquisition unit is configured to acquire an initial gap detection image between the road subgrade surface and the straight ruler collected by the image acquisition device along the length direction of the straight ruler. The initial gap detection image contains the road subgrade surface and the straight ruler.

[0071] The image segmentation unit is configured to perform image rotation, scaling, and cropping processing on the initial gap detection image to obtain a gap detection image with a fixed direction and a fixed size, and cut the gap detection image into a plurality of gap detection segmented images based on the scale marks of the straight ruler on the gap detection image.

[0072] The image conversion unit is configured to perform grayscale processing on each gap detection segmented image to obtain a corresponding gap detection segmented smooth grayscale image, and perform image binarization processing on each gap detection segmented smooth grayscale image to obtain a corresponding gap detection segmented binary image.

[0073] The gap detection unit is configured to perform edge detection on the gap detection segmented binary image, determine each edge closed gap region in the gap detection segmented binary image, and extract a corresponding anchor gap region from each edge closed gap region.

[0074] a parameter determination unit, configured to determine vertical gap parameters and horizontal gap parameters of each anchor gap region in the gap detection segmented binary image by taking a direction perpendicular to the flat ruler as a vertical direction and a direction parallel to the flat ruler as a horizontal direction in the gap detection segmented binary image;

[0075] a feature construction unit, configured to construct a three-dimensional gap feature corresponding to the gap detection segmented binary image based on the number of anchor gap regions and the vertical gap parameters and the horizontal gap parameters of each anchor gap region;

[0076] a flatness detection unit, configured to input a three-dimensional gap feature set composed of the three-dimensional gap features of each gap detection segmented binary image into a preset flatness detection model to perform flatness detection, to obtain a flatness detection result of a corresponding road subgrade section under the flat ruler, and to output the flatness detection result.

[0077] Embodiment 3:

[0078] The embodiment provides a road subgrade flatness detection device, as shown in the accompanying drawings, which comprises, at a hardware level: Figure 3 a data interface, configured to establish data connection between the processor and the image collector;

[0079] a memory, configured to store instructions;

[0080] a processor, configured to read the instructions stored in the memory and execute the road subgrade flatness detection method in Embodiment 1 according to the instructions.

[0081] Optionally, the device further comprises an internal bus, and the processor, the memory and the data interface are connected to each other through the internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0082]

[0083] ​The memory can include, but is not limited to, a Random Access Memory (RAM), a Read Only Memory (ROM), a Flash Memory, a First Input First Output (FIFO) memory, a First In Last Out (FILO) memory, and / or the like. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), and / or the like; and can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA) or other programmable logic device, a discrete gate or transistor logic, a discrete hardware component.

[0084] Embodiment 4:

[0085] The embodiment provides a computer readable storage medium, and instructions are stored on the computer readable storage medium. When the instructions are run on a computer, the computer executes the road subgrade flatness detection method in the embodiment 1. The computer readable storage medium is a carrier for storing data, and can include, but is not limited to, a floppy disk, a compact disc, a hard disk, a flash memory, a USB flash disk, a Memory Stick, and / or the like. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0086] The embodiment also provides a computer program, and when the computer program is run on a computer, the computer program executes the road subgrade flatness detection method in the embodiment 1. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices.

[0087] Finally, it should be noted that: the above only describes the preferred embodiments of the present application and is not used to limit the protection scope of the present application. Any modification, equivalent replacement, improvement, and / or the like made within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting the flatness of a road subgrade, characterized by, The method comprises the following steps: acquiring an initial gap detection image of a road subgrade surface and a flat scale collected by an image collector along the length direction of the flat scale, wherein the initial gap detection image contains the road subgrade surface and the flat scale; performing image rotation, scaling and cropping on the initial gap detection image to obtain a gap detection image with fixed direction and size, and cutting the gap detection image into a plurality of gap detection sub-images based on the scale marks of the flat scale on the gap detection image; performing gray processing on each gap detection sub-image to obtain a corresponding gap detection sub-smooth gray image, and performing image binarization processing on each gap detection sub-smooth gray image to obtain a corresponding gap detection sub-binary image; the gray processing comprises: substituting three color channel component values R, G and B of each pixel point of the gap detection sub-image into a gray conversion formula H=0.39×R+0.50×G+0.11×B to obtain a gray value H of each pixel point, and constructing a gap detection sub-gray image based on the gray value H of each pixel point; performing median filter processing on the gap detection sub-gray image to obtain a gap detection sub-smooth gray image; and performing image binarization processing on the gap detection sub-smooth gray image by using the maximum inter-class variance method to obtain a gap detection sub-binary image; performing edge detection on the gap detection sub-binary image to determine each edge closed gap region in the gap detection sub-binary image, which comprises: performing edge detection on the gap detection sub-binary image by using a Canny edge detection operator to determine each edge closed gap region in the gap detection sub-binary image; and extracting a corresponding anchor gap region from each edge closed gap region, which comprises: determining the number of pixel points contained in each pixel column in the edge closed gap region, taking the number of pixel points as an anchor value of the corresponding pixel column, and taking a region composed of a plurality of pixel columns with anchor values greater than a set reference threshold as an anchor gap region; taking a direction perpendicular to the flat scale as a vertical direction and a direction parallel to the flat scale as a horizontal direction in the gap detection sub-binary image; determining a vertical gap parameter and a horizontal gap parameter of each anchor gap region in the gap detection sub-binary image, which comprises: determining a maximum image distance of the anchor gap region in the vertical direction, and taking the maximum image distance of the anchor gap region in the vertical direction as the vertical gap parameter of the anchor gap region; determining a maximum image distance of the anchor gap region in the horizontal direction, and taking the maximum image distance of the anchor gap region in the horizontal direction as the horizontal gap parameter of the anchor gap region; Based on the number of anchor gap regions contained and the vertical gap parameter and the horizontal gap parameter of each anchor gap region, a three-dimensional gap feature of the corresponding gap detection segmented binary image is constructed; including: determining the number L of anchor gap regions contained in the gap detection segmented binary image, determining the maximum value M of the vertical gap parameter of each anchor gap region in the gap detection segmented binary image, and determining the maximum value N of the horizontal gap parameter of each anchor gap region in the gap detection segmented binary image; using L, M and N to construct the three-dimensional gap feature (L, M, N) of the gap detection segmented binary image; The three-dimensional gap feature set is used to form a three-dimensional gap feature set, and the three-dimensional gap feature set is input into the preset flatness detection model for flatness detection to obtain the flatness detection result of the corresponding road subgrade section under the flat ruler, and the flatness detection result is output.

2. The method for detecting the flatness of a road subgrade according to claim 1, characterized in that, Before inputting the three-dimensional gap feature set into the preset flatness detection model for flatness detection, the method further comprises: An initialized BP neural network model is constructed, and the initialized BP neural network model is trained and optimized by using a training set until a set training condition is met, to obtain a trained flatness detection model, wherein the training set contains a plurality of three-dimensional gap feature set samples, and each three-dimensional gap feature set sample is labeled with a corresponding flatness label.

3. A road subgrade flatness detection system characterized by, It comprises an image acquisition unit, an image segmentation unit, an image conversion unit, a gap detection unit, a parameter determination unit, a feature construction unit and a flatness detection unit, wherein: The image acquisition unit is used to acquire an initial gap detection image between the road subgrade surface and the flat ruler collected by the image collector along the length direction of the flat ruler, and the initial gap detection image contains the road subgrade surface and the flat ruler; The image segmentation unit is used to perform image rotation, scaling and cropping processing on the initial gap detection image to obtain a gap detection image with fixed direction and fixed size, and cut the gap detection image into a plurality of gap detection segmented images based on the scale marks of the flat ruler on the gap detection image; The image conversion unit is used to perform gray processing on each gap detection segmented image to obtain a corresponding gap detection segmented smooth gray image, and perform image binarization processing on each gap detection segmented smooth gray image to obtain a corresponding gap detection segmented binary image; including: substituting the three-color channel component values R, G and B of each pixel point of the gap detection segmented image into the gray conversion formula H=0.39×R+0.50×G+0.11×B to obtain the gray value H of each pixel point, and constructing a gap detection segmented gray image based on the gray value H of each pixel point; performing median filtering processing on the gap detection segmented gray image to obtain a gap detection segmented smooth gray image; and performing image binarization processing on the gap detection segmented smooth gray image by using the maximum inter-class variance method to obtain a gap detection segmented binary image; The gap detection unit is configured to perform edge detection on the gap detection segmented binary image, determine each edge-closed gap region in the gap detection segmented binary image, and include performing edge detection on the gap detection segmented binary image by using a Canny edge detection operator, determining each edge-closed gap region in the gap detection segmented binary image, and extracting a corresponding anchor gap region from each edge-closed gap region, including determining the number of pixel points contained in each pixel column in the edge-closed gap region, taking the number of pixel points as an anchor value of the corresponding pixel column, and taking a region formed by pixel columns with a continuous plurality of anchor values greater than a set reference threshold value in the edge-closed gap region as the anchor gap region; The parameter determination unit is configured to take a direction perpendicular to the flat ruler in the gap detection segmented binary image as a vertical direction, take a direction parallel to the flat ruler as a horizontal direction, determine a vertical gap parameter and a horizontal gap parameter of each anchor gap region in the gap detection segmented binary image, including determining a maximum image interval of the anchor gap region in the vertical direction, and taking the maximum image interval of the anchor gap region in the vertical direction as the vertical gap parameter of the anchor gap region, determining a maximum image interval of the anchor gap region in the horizontal direction, and taking the maximum image interval of the anchor gap region in the horizontal direction as the horizontal gap parameter of the anchor gap region; The feature construction unit is configured to construct a three-dimensional gap feature of the corresponding gap detection segmented binary image based on the number of anchor gap regions contained and the vertical gap parameter and the horizontal gap parameter of each anchor gap region, including determining the number L of anchor gap regions contained in the gap detection segmented binary image, determining a maximum value M of the vertical gap parameter of each anchor gap region in the gap detection segmented binary image, determining a maximum value N of the horizontal gap parameter of each anchor gap region in the gap detection segmented binary image, and constructing the three-dimensional gap feature (L, M, N) of the gap detection segmented binary image by using L, M, and N. The flatness detection unit is configured to use the three-dimensional gap features of each gap detection segmented binary image to form a three-dimensional gap feature set, input the three-dimensional gap feature set into a preset flatness detection model for flatness detection, obtain a flatness detection result of the corresponding roadbed section under the flat ruler, and output the flatness detection result.

4. A road embankment flatness detection device characterized by comprising: The memory is configured to store instructions; The processor is configured to read the instructions stored in the memory, and execute the roadbed flatness detection method of claim 1 or 2 according to the instructions. When the computer program product runs on the computer, the roadbed flatness detection method of claim 1 or 2 is executed.

5. A computer program product, characterised in that, ​

Images