Driving-assisted road surface classification and recognition method, system, and intelligent terminal

Abstract

The present invention discloses a method, system, and intelligent terminal for road surface classification and recognition based on assisted driving. The method comprises: acquiring an original image of a road scene within a target area and processing the original image to obtain a target image; labeling each corresponding attribute in the target image based on preset attribute categories and attribute classification standards to obtain a target image with multiple attribute labels; training a ground type recognition network based on the target image with multiple attribute labels, and utilizing the ground type recognition network to output attribute classification results for the road surface within the target area. This method can monitor and perceive the road surface category and road surface condition in real time, enabling the vehicle body controller to adjust the chassis control mode based on the characteristics of different road surface categories, thereby improving driving safety and comfort.

Description

Technical Field

[0001] The present invention relates to the field of automatic driving assistance technology, and in particular to a method, system and intelligent terminal for classifying and identifying road surfaces based on assisted driving. Background Art

[0002] With the advancement of autonomous driving technology, people's demands for the safety and comfort of assisted-driving vehicles are increasing. Vehicle dynamics and braking performance are core performance indicators. When facing water, gravel, ice, and snow, the vehicle chassis needs to adjust the distribution of driving and braking forces to prevent tire slippage or lateral sliding and loss of control due to the different road adhesion coefficients. Furthermore, off-road driving is an integral part of automotive culture. When off-roading, different strategies are used for all-wheel drive force distribution when navigating terrain such as grass, rocks, snow, and desert.

[0003] Therefore, real-time monitoring and perception of the type and condition of the driving road surface so that the body controller can adjust the chassis control mode according to the characteristics of different types of road surfaces, thereby improving driving safety and comfort, has become an urgent problem to be solved by technical personnel in this field. Summary of the Invention

[0004] To this end, embodiments of the present invention provide a method, system, and intelligent terminal for classifying and identifying driving surfaces based on assisted driving, in order to monitor and perceive the type and condition of the driving surface in real time, so that the vehicle body controller can adjust the chassis control mode according to the characteristics of different types of road surfaces, thereby improving driving safety and comfort.

[0005] In order to achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0006] A method for classifying and identifying a driving surface based on assisted driving, the method comprising:

[0007] Acquire an original image of a road scene in a target area, and process the original image to obtain a target image;

[0008] Labeling corresponding attributes in the target image based on preset attribute categories and attribute classification standards to obtain a target image with multiple attribute labels;

[0009] A ground type recognition network is trained based on a target image with multiple attribute labels, and the ground type recognition network is used to output attribute category results of a driving road surface in a target area.

[0010] Furthermore, the acquiring of an original image of a road scene in a target area and processing the original image to obtain a target image specifically includes:

[0011] Obtaining the original image of the road scene in the target area;

[0012] Cropping the original image to remove the sky area in the original image;

[0013] The cropped original image is downsampled to obtain the target image.

[0014] Furthermore, the attribute categories specifically include:

[0015] Ground type attributes, where the ground type attributes include at least hard road surface, desert, grassland, mud, wading, and ice and snow road surface;

[0016] Pavement material attributes, where the pavement material attributes include at least asphalt, concrete, paving bricks, and sand;

[0017] Road surface wetness and dryness condition attributes, wherein the road surface wetness and dryness condition attributes include at least dry, wet, and waterlogged.

[0018] Furthermore, the training of a ground type recognition network based on a target image with multiple attribute labels specifically includes:

[0019] Data processing: labeling massive amounts of raw image data of road scenes within the target area according to multi-attribute labels and generating a training set;

[0020] Model training: input the training set into the designed ground type recognition network and perform multiple iterative learning to obtain model parameters;

[0021] Model inference: input the test image into the trained ground type recognition network model to predict and output the results.

[0022] Furthermore, the outputting of the attribute category result of the current driving road surface also includes:

[0023] By accumulating category information of consecutive frame images, a stable attribute category result is output.

[0024] Furthermore, the step of accumulating category information of consecutive frame images to output a stable attribute category result specifically includes:

[0025] The continuous frame images of the vehicle during driving are inferred by the ground type recognition network model to obtain the multi-attribute categories of the single frame image;

[0026] Accumulate the multi-attribute categories of consecutive multi-frame images and eliminate the attribute jumps between multi-frame images to output stable attribute category results.

[0027] Furthermore, the step of accumulating category information of consecutive frame images to output a stable attribute category result specifically includes:

[0028] Obtain the attribute category of the first frame image, and obtain the attribute category result and the weight of the corresponding attribute category based on the multi-layer convolutional network;

[0029] Take a preset number of consecutive frame images and record the attribute category information of each frame image and the weight of the corresponding attribute category;

[0030] Obtain the attribute category of the x-th frame image and the weight of the corresponding attribute category;

[0031] Perform weighted summation on the attribute categories of x consecutive frames, and take the attribute category corresponding to the maximum weighted sum value as the attribute category result of the image.

[0032] The present invention also provides a driving-assisted road surface classification and recognition system, the system comprising:

[0033] An image acquisition unit, configured to acquire an original image of a road scene within a target area and process the original image to obtain a target image;

[0034] an attribute classification unit, configured to label corresponding attributes in the target image based on preset attribute categories and attribute classification standards, so as to obtain an image with multiple attribute labels;

[0035] a result output unit, which trains a ground type recognition network based on the multi-attribute labels and uses the ground type recognition network to output an attribute category result of the current driving road surface;

[0036] The category information accumulation unit is used to accumulate category information of consecutive frame images to output a stable attribute category result.

[0037] The present invention also provides an intelligent terminal, which includes: a data acquisition device, a processor and a memory;

[0038] The data acquisition device is used to acquire data; the memory is used to store one or more program instructions; and the processor is used to execute one or more program instructions to perform the method described above.

[0039] The present invention also provides a computer-readable storage medium, wherein the computer-readable storage medium includes one or more program instructions, and the one or more program instructions are used to execute the method described above.

[0040] The assisted driving-based road surface classification and recognition method provided by the present invention labels the corresponding attributes in a target image based on preset attribute categories and attribute classification standards to produce an image with multiple attribute labels. A ground type recognition network is trained based on these multiple attribute labels, and this ground type recognition network outputs the attribute classification results for the current road surface. This method classifies the ground ahead of the vehicle, including its type, wetness, and material. The classification status is maintained based on the accumulated information from consecutive frames, and a stable recognition signal is output to the vehicle controller. This method can monitor and perceive the road surface category and state in real time, enabling the vehicle body controller to adjust the chassis control mode based on the characteristics of different road types, thereby improving driving safety and comfort. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments or the description of the prior art. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other implementation drawings based on the provided drawings without inventive effort.

[0042] The structures, proportions, sizes, etc. illustrated in this specification are intended only to complement the contents disclosed herein and to facilitate understanding and reading by persons familiar with the art. They are not intended to limit the conditions under which the present invention may be implemented and therefore have no substantive technical significance. Any structural modifications, changes in proportions, or adjustments in sizes, without affecting the efficacy and objectives of the present invention, shall still fall within the scope of the technical contents disclosed herein.

[0043] Figure 1 This is a flowchart of a specific implementation of the method for classifying and identifying road surfaces based on assisted driving provided by the present invention;

[0044] Figure 2 Schematic diagram of the scene of the roi image;

[0045] Figure 3 Classification box diagram for attribute labels;

[0046] Figure 4 is a schematic diagram of the network structure;

[0047] Figure 5 This is a structural block diagram of a specific implementation of the driving-assisted road surface classification and recognition system provided by the present invention. DETAILED DESCRIPTION

[0048] The following describes the implementation of the present invention using specific embodiments. Those skilled in the art will readily understand the other advantages and benefits of the present invention from the disclosure herein. Obviously, the embodiments described are only a portion of the present invention, not all of it. All other embodiments derived by persons of ordinary skill in the art based on the embodiments of the present invention without inventive effort are intended to fall within the scope of protection of the present invention.

[0049] The method for road surface classification and recognition for assisted driving, provided by this invention, first crops and downsamples the ROI image to reduce network computational complexity. It then utilizes a multi-attribute label classification network to determine the category and confidence level of each attribute. Finally, it accumulates and maintains the network output, ultimately obtaining more stable ground type information. This method classifies the ground ahead of the vehicle based on its type, wetness, and material. The classification status is maintained based on the accumulated information from consecutive frames, and a stable recognition signal is output to the vehicle controller, allowing it to determine the most appropriate driving control strategy based on the received road attribute classification. This improves the safety and comfort of assisted driving vehicles.

[0050] In a specific embodiment, Figure 1 As shown, the driving-assisted road surface classification and recognition method provided by the present invention includes the following steps:

[0051] S1: Acquire an original image of a road scene in a target area, and process the original image to obtain a target image.

[0052] Specifically, in step S1, after obtaining the original image of the road scene in the target area, the original image is firstly cropped to remove the sky area in the original image, and then the cropped original image is downsampled to obtain the target image.

[0053] In a specific usage scenario, the original collected image data (assuming the resolution is 1280*720) generally contains non-ground areas such as the sky above the road and the background. In order to reduce the computational workload of the network and improve the real-time performance of the algorithm, the original image needs to be cropped and downsampled. The sky area in the upper half of the original image is cropped to obtain the ROI (region of interest) image, and the ROI is downsampled to further reduce the computational workload. Figure 2 The following example illustrates that the image resolution of the roi area is 640*320, and the image resolution after downsampling is 256*128. Figure 2 The middle frame area is the ROI area.

[0054] S2: Labeling corresponding attributes in the target image based on preset attribute categories and attribute classification standards to obtain a target image with multiple attribute labels.

[0055] The attribute categories specifically include:

[0056] The surface type attribute includes at least normal hard pavement, desert, grass, mud, wading, and ice and snow. The surface type attribute primarily reflects the range of surface types that vehicles (especially off-road vehicles) can travel on, including but not limited to normal hard pavement, desert, grass, mud, wading, ice and snow. In subsequent algorithms, the number of types of this surface type attribute is set to n.

[0057] Pavement material attributes, including at least asphalt, concrete, paving bricks, and sandy soil. Pavement material attributes are primarily targeted at common hard pavement and are categorized based on common road surface materials encountered during normal vehicle driving, including but not limited to asphalt, concrete, paving bricks, and sandy soil. In subsequent algorithms, the number of types of these pavement material attributes is set to m.

[0058] Road surface wetness attribute, which includes at least dry, wet, and waterlogged conditions. This attribute is primarily targeted at ordinary hard roads and is categorized based on their wetness, including but not limited to dry, wet, and waterlogged conditions. In subsequent algorithms, the number of types of this attribute is set to k.

[0059] like Figure 3 As shown, still taking the above specific usage scenario as an example, the ROI image is classified into multiple attribute labels according to the above classification criteria.

[0060] S3: Training a ground type recognition network based on the target image with multiple attribute labels, and using the ground type recognition network to output attribute category results of the driving road surface in the target area.

[0061] Specifically, the ground type recognition network based on multi-attribute labels basically includes three parts: data labeling, model training and model reasoning. The basic structure of the network is as follows: Figure 4 As shown, Figure 4 In this example, input image represents the input image, Conv represents multiple convolutional layer operations, softmax represents classification, and final result represents the final classification result.

[0062] Training a ground type recognition network based on multi-attribute labels includes:

[0063] Data processing: labeling massive amounts of raw image data of road scenes within the target area according to multi-attribute labels and generating a training set;

[0064] Model training: input the training set into the designed ground type recognition network and perform multiple iterative learning to obtain model parameters;

[0065] Model inference: input the test image into the trained ground type recognition network model to predict and output the results.

[0066] The input image undergoes multi-layer convolution operations for feature extraction, and then the classification results are output from three different convolution layers. Finally, the multi-scale classification information is fused to output the final attribute category results. The category results of multiple attributes are inferred and output according to the above model structure.

[0067] In order to improve the accuracy and output stability of the results, in this specific embodiment, the method provided by the present invention further includes the following steps before outputting the attribute category result of the current driving road surface:

[0068] By accumulating category information of consecutive frame images, a stable attribute category result is output.

[0069] Specifically, the accumulation of category information of consecutive frame images includes the following steps:

[0070] The continuous frame images of the vehicle during driving are inferred by the ground type recognition network model to obtain the multi-attribute categories of the single frame image;

[0071] Accumulate the multi-attribute categories of consecutive multi-frame images and eliminate the attribute jumps between multi-frame images to output stable attribute category results.

[0072] More specifically, the accumulation of category information through consecutive frame images includes the following steps:

[0073] Obtain the attribute category of the first frame image, and obtain the attribute category result and the weight of the corresponding attribute category based on the multi-layer convolutional network;

[0074] Take a preset number of consecutive frame images and record the attribute category of each frame image and the weight of the corresponding attribute category;

[0075] Obtain the attribute category of the x-th frame image and the weight of the corresponding attribute category;

[0076] Perform weighted summation on the attribute categories of x consecutive frames, and take the attribute category corresponding to the maximum weighted sum value as the attribute category result of the image.

[0077] In one usage scenario, the accumulation of category information from consecutive frame images includes the following steps:

[0078] S100: Obtain multi-attribute category information of the first frame image. The model inference results of ground type, road surface material, and road surface dryness and wetness are {n1, m1, k1}, and the corresponding attribute weights are {wn1, wm1, wk1};

[0079] S200: Continuously record the multi-attribute category information of each frame of image, for example, the result of the i-th frame image is {ni, mi, ki}, and the corresponding attribute weights are {wni, wmi, wki}; where i is from 1 to x;

[0080] S300: Obtain multi-attribute category information of the x-th frame image, which is {nx, mx, kx}, and the weights of the corresponding attributes are {wnx, wmx, wkx};

[0081] S400: Perform weighted summation on the multi-attribute category information of x consecutive frames. The calculation method is as follows:

[0082] Sn = n1 * wn1 + n2 * wn2 + … + ni * wni + … + nx * wnx

[0083] Sm = m1 * wm1 + m2 * wm2 + … + mi * wmi + … + mx * wmx

[0084] Sk = k1 * wk1 + k2 * wk2 + … + ki * wki + … + kx * wkx

[0085] S500: Take the type corresponding to the maximum weighted sum value as the stable category of the attribute.

[0086] In the above-mentioned specific embodiment, the present invention provides a method for assisted driving road surface classification and recognition. By labeling the corresponding attributes in the target image based on preset attribute categories and attribute classification criteria, a multi-attribute labeled image is generated. A ground type recognition network is trained based on the multi-attribute labels and utilized. This method classifies the ground type, wetness, and material in front of the vehicle. The classification status is maintained based on the accumulated information from consecutive frames, thereby outputting a stable recognition signal to the vehicle controller. This method enables real-time monitoring and perception of the road surface type and condition, enabling the vehicle body controller to adjust the chassis control mode based on the characteristics of different road types, thereby improving driving safety and comfort.

[0087] In addition to the above method, the present invention also provides a road surface classification and recognition system based on assisted driving, such as Figure 5 As shown, the system includes:

[0088] The image acquisition unit 100 is used to acquire an original image of a road scene in a target area and process the original image to obtain a target image;

[0089] The image acquisition unit 100 is specifically used for:

[0090] Obtaining the original image of the road scene in the target area;

[0091] Cropping the original image to remove the sky area in the original image;

[0092] The cropped portrait is downsampled to obtain the target image.

[0093] The attribute classification unit 200 is used to label the corresponding attributes in the target image based on the preset attribute categories and attribute classification standards to obtain a target image with multiple attribute labels;

[0094] The attribute classification unit 200 is specifically used for: ground type attributes, wherein the ground type attributes include at least hard road surface, desert, grassland, mud, wading, and ice and snow road surface;

[0095] Pavement material attributes, where the pavement material attributes include at least asphalt, concrete, paving bricks, and sand;

[0096] Road surface wetness and dryness condition attributes, wherein the road surface wetness and dryness condition attributes include at least dry, wet, and waterlogged.

[0097] The result output unit 300 is used to train a ground type recognition network based on the target image with multiple attribute labels, and output attribute category results of the driving road surface in the target area using the ground type recognition network.

[0098] The result output unit 300 is specifically used to extract features from the input image through multi-layer convolution operations, then output classification results from three different convolution layers respectively, and finally fuse multi-scale classification information to output the final attribute category result.

[0099] The category information accumulation unit 400 is configured to accumulate category information of consecutive frame images to output a stable attribute category result.

[0100] The category information accumulation unit 400 is specifically configured to:

[0101] The continuous frame images of the vehicle during driving are inferred by the ground type recognition network model to obtain the multi-attribute categories of the single frame image;

[0102] Accumulate the multi-attribute categories of consecutive multi-frame images and eliminate the attribute jumps between multi-frame images to output stable attribute category results.

[0103] The category information accumulation unit 400 is specifically configured to:

[0104] Obtain the attribute category of the first frame image, and obtain the attribute category result and the weight of the corresponding attribute category based on the multi-layer convolutional network;

[0105] Take a preset number of consecutive frame images and record the attribute category of each frame image and the weight of the corresponding attribute category;

[0106] Obtain the attribute category of the x-th frame image and the weight of the corresponding attribute category;

[0107] Perform weighted summation on the attribute categories of x consecutive frames, and take the attribute category corresponding to the maximum weighted sum value as the attribute category result of the image.

[0108] In the above-mentioned specific embodiment, the driving-assisted road surface classification and recognition system provided by the present invention labels each corresponding attribute in the target image based on preset attribute categories and attribute classification standards to obtain an image with attribute classification identification. The image with attribute classification identification is then passed through a multi-layer convolutional network, and feature extraction is performed within the multi-layer convolutional network to output the attribute classification result of the current road surface. This classifies the ground type, dryness and wetness, and road surface material in front of the vehicle, and maintains the classification status based on the accumulated information of consecutive frames, thereby outputting a stable recognition signal to the vehicle controller. This method can monitor and perceive the category and road surface status of the driving road in real time, allowing the body controller to adjust the chassis control mode according to the characteristics of different road types, thereby improving driving safety and comfort.

[0109] The present invention also provides an intelligent terminal, which includes: a data acquisition device, a processor and a memory;

[0110] The data acquisition device is used to acquire data; the memory is used to store one or more program instructions; and the processor is used to execute one or more program instructions to perform the method described above.

[0111] Corresponding to the above embodiment, an embodiment of the present invention further provides a computer-readable storage medium, which includes one or more program instructions. The one or more program instructions are used to execute the above method in a binocular camera depth calibration system.

[0112] In the embodiments of the present invention, the processor may be an integrated circuit chip with signal processing capabilities. The processor may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0113] The methods, steps, and logic diagrams disclosed in the embodiments of the present invention can be implemented or executed. A general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present invention can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules within the decoding processor. The software modules can be located in a storage medium well-established in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. The processor reads the information from the storage medium and, in conjunction with its hardware, completes the steps of the aforementioned methods.

[0114] The storage medium may be a memory and may be, for example, a volatile memory or a nonvolatile memory, or may include both volatile and nonvolatile memory.

[0115] Among them, the non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM) or flash memory.

[0116] Volatile memory can be random access memory (RAM), which is used as an external cache memory. By way of example and not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDRSDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM), and direct RAM bus random access memory (DRRAM).

[0117] The storage media described in the embodiments of the present invention are intended to include, but are not limited to, these and any other suitable types of memory.

[0118] Those skilled in the art will appreciate that in one or more of the above examples, the functions described in the present invention can be implemented using a combination of hardware and software. When software is used, the corresponding functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium. Computer-readable media include computer-readable storage media and communication media, wherein communication media includes any medium that facilitates the transmission of computer programs from one place to another. The storage medium can be any available medium that can be accessed by a general-purpose or special-purpose computer.

[0119] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for classifying and identifying driving surfaces based on assisted driving, characterized in that: The method comprises: Acquire an original image of a road scene in a target area, and process the original image to obtain a target image; Labeling each corresponding attribute in the target image based on a preset attribute category and attribute classification standard to obtain an image with multiple attribute labels; Training a ground type recognition network based on the multi-attribute labels, and using the ground type recognition network to output an attribute category result of the current driving road surface; The output of the attribute category result of the current driving road surface also includes: By accumulating the category information of consecutive frame images, a stable attribute category result is output; The method of accumulating category information of consecutive frame images to output a stable attribute category result specifically includes: The continuous frame images of the vehicle during driving are subjected to network model inference to obtain the multi-attribute category information of a single frame image; Accumulate multi-attribute category information of multiple consecutive frames and eliminate attribute jumps between multi-frame recognition results to output stable attribute category results; The method of accumulating category information of consecutive frame images to output a stable attribute category result specifically includes: Obtain the attribute category information of the first frame image, and obtain the attribute category result and the weight of the corresponding attribute category based on the multi-layer convolutional network; Take a preset number of consecutive frame images and record the attribute category information of each frame image and the weight of the corresponding attribute category; Obtain the attribute category information of the x-th frame image and the weight of the corresponding attribute category; Perform weighted summation on the attribute category information of x consecutive frames, and take the type corresponding to the maximum weighted sum value as the attribute category result of the image; The attribute categories specifically include: Ground type attributes, which include at least ordinary hard road, desert, grassland, mud, wading, and ice and snow road, are set to n; Pavement material attributes, which include at least asphalt, concrete, paving bricks, and sandy soil, are set to m; Road surface dryness and wetness condition attributes, which include at least dry, wet, and waterlogged, and are set to k; The accumulation of category information through consecutive frame images includes the following steps: S100: Obtain the multi-attribute category information of the first frame image, including the ground type, road surface material, and road surface dry and wet conditions. The result of type reasoning is {n1, m1, k1}, and the weight of the corresponding attribute is {wn1, wm1, wk1}; S200: Continuously record the multi-attribute category information of each frame image, such as the result of the i-th frame image is {ni, mi, ki}, The weights of the corresponding attributes are {wni, wmi, wki}; where i ranges from 1 to x; S300: Obtain the multi-attribute category information of the x-th frame image, which is {nx, mx, kx}, and the weight of the corresponding attribute is { wnx, wmx, wkx}; S400: Perform weighted summation on the multi-attribute category information of x consecutive frames. The calculation method is as follows: Sn=n1*wn1+n2*wn2+……+ni*wni+……+nx*wnx Sm=m1*wm1+m2*wm2+……+mi*wmi+……+mx*wmx Sk=k1*wk1+k2*wk2+……+ki*wki+……+kx*wkx S500: Take the type corresponding to the maximum weighted sum value as the stable category of the attribute.

2. The method for classifying and identifying a driving road surface according to claim 1, wherein: The acquiring of an original image of a road scene in a target area and processing the original image to obtain a target image specifically includes: Obtaining the original image of the road scene in the target area; Cropping the original image to remove the sky area in the original image; The cropped portrait is downsampled to obtain the target image.

3. The method for classifying and identifying a driving road surface according to claim 1, wherein: The method of training a ground type recognition network based on multi-attribute labels specifically includes: data processing, labeling massive image data according to multi-attribute labels, and generating a training set; Model training: input the training set into the designed ground type recognition network and perform multiple iterative learning to obtain model parameters; Model inference: input the test image into the trained network model to predict and output the results.

4. A road surface classification and recognition system based on assisted driving, characterized in that: The system comprises: An image acquisition unit, configured to acquire an original image of a road scene within a target area and process the original image to obtain a target image; an attribute classification unit, configured to label corresponding attributes in the target image based on preset attribute categories and attribute classification standards, so as to obtain an image with multiple attribute labels; a result output unit, configured to train a ground type recognition network based on multiple attribute labels, and output an attribute category result of the current driving road surface using the ground type recognition network; The result output unit is further used for: By accumulating the category information of consecutive frame images, a stable attribute category result is output; The method of accumulating category information of consecutive frame images to output a stable attribute category result specifically includes: The continuous frame images of the vehicle during driving are subjected to network model inference to obtain the multi-attribute category information of a single frame image; Accumulate multi-attribute category information of multiple consecutive frames and eliminate attribute jumps between multi-frame recognition results to output stable attribute category results; The method of accumulating category information of consecutive frame images to output a stable attribute category result specifically includes: Obtain the attribute category information of the first frame image, and obtain the attribute category result and the weight of the corresponding attribute category based on the multi-layer convolutional network; Take a preset number of consecutive frame images and record the attribute category information of each frame image and the weight of the corresponding attribute category; Obtain the attribute category information of the x-th frame image and the weight of the corresponding attribute category; Perform weighted summation on the attribute category information of x consecutive frames, and take the type corresponding to the maximum weighted sum value as the attribute category result of the image; The attribute categories specifically include: Ground type attributes, which include at least ordinary hard road, desert, grassland, mud, wading, and ice and snow road, are set to n; Pavement material attributes, which include at least asphalt, concrete, paving bricks, and sandy soil, are set to m; Road surface dryness and wetness condition attributes, which include at least dry, wet, and waterlogged, and are set to k; The accumulation of category information through consecutive frame images includes the following steps: S100: Obtain the multi-attribute category information of the first frame image, including the ground type, road surface material, and road surface dry and wet conditions. The result of type reasoning is {n1, m1, k1}, and the weight of the corresponding attribute is {wn1, wm1, wk1}; S200: Continuously record the multi-attribute category information of each frame image, such as the result of the i-th frame image is {ni, mi, ki}, The weights of the corresponding attributes are {wni, wmi, wki}; where i ranges from 1 to x; S300: Obtain the multi-attribute category information of the x-th frame image, which is {nx, mx, kx}, and the weight of the corresponding attribute is { wnx, wmx, wkx}; S400: Perform weighted summation on the multi-attribute category information of x consecutive frames. The calculation method is as follows: Sn=n1*wn1+n2*wn2+……+ni*wni+……+nx*wnx Sm=m1*wm1+m2*wm2+……+mi*wmi+……+mx*wmx Sk=k1*wk1+k2*wk2+……+ki*wki+……+kx*wkx S500: Take the type corresponding to the maximum weighted sum value as the stable category of the attribute.

5. An intelligent terminal, characterized in that: The intelligent terminal includes: a data acquisition device, a processor and a memory; The data acquisition device is used to acquire data; the memory is used to store one or more program instructions; and the processor is used to execute one or more program instructions to perform the method according to any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that The computer storage medium includes one or more program instructions, and the one or more program instructions are used to execute the method according to any one of claims 1 to 3.

Images