A method and system for online automatic detection of field soil mechanical properties

Abstract

The present application relates to the field of vehicle ground mechanics and field vehicle driving auxiliary detection technology, and discloses a kind of field soil mechanics property on-line automatic detection method and system.The method is non-contact acquisition soil surface image of soil area in front of vehicle or current contact area, based on inertial measurement unit clock synchronization vehicle CAN data, combined with vehicle speed, wheel speed difference, suspension displacement compensation and tire slip rate complete area mapping, obtain in-situ detection data;Image feature data and vehicle state data are extracted from in-situ detection data, and a mechanical parameter array is obtained by a pre-trained mechanical performance inversion model, generating shear evaluation data, bearing evaluation data, passability evaluation data and vehicle control interface data.The present application is used to realize non-contact soil mechanics detection and vehicle control linkage.

Description

Technical Field

[0001] This invention relates to the field of vehicle ground mechanics and field vehicle driving assistance detection technology, and in particular to an online automatic detection method and system for field soil mechanical properties. Background Technology

[0002] In the field of vehicle ground mechanics and off-road vehicle driving assistance detection technology, existing detection schemes for off-road vehicles to detect the soil area in front of the vehicle or the current ground contact area usually rely on contact soil detection, vehicle status acquisition, soil surface image recognition, or passability experience judgment. These schemes have limitations such as difficulty in synchronizing soil surface images with vehicle status data, difficulty in converting image feature data into mechanical parameter arrays, and difficulty in inputting detection results into vehicle control interface data. Existing methods often rely on parking sampling, contact measurement, manual interpretation, or separate vehicle status judgment, which can easily lead to the separation of soil detection results from vehicle driving status.

[0003] During continuous driving in the field, there are differences in the time of data collection, the data collection area, and the stress state of the vehicle corresponding to the soil area in front of the vehicle and the current grounding area. Existing solutions are prone to problems such as the inability to generate passability evaluation data in advance for the soil area in front of the vehicle, and the inability to timely associate the current wheel end demand force and the current calculated grounding pressure of the vehicle for the current grounding area. It is difficult to meet the requirement of stable generation of passability evaluation data and vehicle control interface data.

[0004] For the joint processing of image feature data and vehicle status data, existing technologies generally lack a continuous processing link from soil surface images, vehicle status data, mechanical parameter arrays, shear evaluation data, bearing capacity evaluation data to passability evaluation data and vehicle control interface data. This makes it difficult to form a consistent process of acquisition, time alignment, grounding area mapping, mechanical performance inversion, passability evaluation and control output in field vehicle driving scenarios, resulting in a lack of stable correspondence between online soil mechanical property detection results and vehicle control systems. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides an online automatic detection method for the mechanical properties of field soil, comprising: S100: Non-contact acquisition of soil surface images of the soil area in front of the vehicle or the current grounding area; based on the inertial measurement unit clock to synchronize vehicle CAN data; combined with vehicle speed, wheel speed difference, tire pressure, suspension displacement compensation, vehicle attitude, wheel end torque command, current calculated grounding pressure and tire slip ratio to perform front area projection correction and grounding area mapping to obtain in-situ detection data. S200. Based on the in-situ detection data, perform image correction, effective area extraction, particle, texture, particle size, sphericity, roughness, porosity and vehicle status extraction to obtain image feature data and vehicle status data. S300. Input the image feature data into the pre-trained mechanical performance inversion model to obtain a mechanical parameter array, and establish a corresponding relationship with the vehicle state data according to the collection time and collection area. S400. Based on the mechanical parameter array and the vehicle state data, perform traction, braking, pressure and subsidence risk assessment to obtain shear evaluation data and pressure evaluation data; S500: Calculate the passability coefficient based on the shear-type evaluation data and the pressure-type evaluation data, and generate passability evaluation data according to the preset grading threshold. S600. Based on the passability evaluation data, generate vehicle control interface data. If there are evaluation records to be reviewed, invalid local area records, or temporary data identifiers, filter and smooth the vehicle control interface data and then output it to the vehicle control system.

[0006] Furthermore, in S100, the synchronization of vehicle CAN data based on the inertial measurement unit clock, the forward area projection correction, and the ground area mapping include: Using the inertial measurement unit clock as a synchronization reference, a unified timestamp is written to the soil surface image, the data read from the vehicle CAN bus, and the attitude change data output by the inertial measurement unit.

[0007] Tire slip ratio is generated based on vehicle speed and wheel speed, and suspension displacement compensation is generated based on suspension height changes within adjacent acquisition periods.

[0008] When the soil surface image corresponds to the soil area in front of the vehicle, the image acquisition module installation position, vehicle driving direction, vehicle speed, attitude change data output by the inertial measurement unit, suspension displacement compensation amount and tire slip ratio are combined to perform front area projection correction on the soil area in front of the vehicle, and the corrected soil area in front of the vehicle is mapped to the soil area that the vehicle is about to reach.

[0009] When the soil surface image corresponds to the current grounding area, the wheel center coordinates are determined according to the preset vehicle geometric parameters, and the current grounding area is correlated with the vehicle status data at the same acquisition time by combining the tire pressure, suspension height, vehicle posture and ground imprint model.

[0010] When two consecutive soil surface images cover the same soil area, the image with higher clarity is retained, and the image acquisition time of the replaced image is recorded.

[0011] Furthermore, in S200, the image feature data includes distribution data, particle shape data, surface texture data, and pore structure data.

[0012] The image correction includes distortion correction based on camera intrinsic parameters and region mapping based on camera extrinsic parameters.

[0013] The effective region extraction includes removing vehicle component areas, water accumulation areas, rock areas, and vegetation areas from the corrected soil surface image to obtain the soil area to be analyzed.

[0014] The particle segmentation is used to extract particle profile data from the soil region to be analyzed.

[0015] The particle size distribution statistics are used to generate distribution data based on the particle profile data.

[0016] The sphericity calculation is used to generate particle shape data based on the particle contour data.

[0017] The texture analysis and surface roughness extraction are used to generate surface texture data.

[0018] The porosity estimation is used to generate pore structure data based on the interparticle void region.

[0019] The vehicle state extraction process includes reading vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel-end torque command, and current calculated ground pressure from the in-situ detection data, and generating the current wheel-end demand force of the vehicle based on the wheel-end torque command to obtain vehicle state data.

[0020] Furthermore, in S300, the pre-trained mechanical performance inversion model is generated by the offline training subsystem.

[0021] The offline training subsystem acquires direct shear test data, bearing plate test data, penetration test data, and triaxial test data, and performs calibration processing on the discrete element simulation module to obtain a set of reference parameters.

[0022] The discrete element simulation module generates a virtual soil surface image and corresponding mechanical parameter labels based on the benchmark parameter set.

[0023] Based on the virtual soil surface image, the image feature data extracted from the virtual soil surface image, and the mechanical parameter labels, a paired training set is constructed, and the neural network model is trained to obtain the pre-trained mechanical performance inversion model.

[0024] The array of mechanical parameters includes cohesion, internal friction angle, shear modulus, and compressive strength.

[0025] Furthermore, in S300, a correspondence is established between the mechanical parameter array and the vehicle state data according to the acquisition time and acquisition area, including: Establish a correlation record between the cohesion, internal friction angle, shear deformation modulus, and bearing capacity at the same collection time and in the same collection area, and the vehicle's current wheel end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and vehicle body attitude.

[0026] When the data collection area is the soil area in front of the vehicle, the mechanical parameter array is attached to the soil area that the vehicle is about to reach, while retaining the vehicle speed and driving direction.

[0027] When the data collection area is the current grounding area, the mechanical parameter array is attached to the wheel contact position, and the vehicle's current wheel end demand force, current calculated ground pressure, tire pressure and suspension status are retained.

[0028] Furthermore, in S400, the traction capability evaluation and braking capability evaluation constitute a shear-type evaluation.

[0029] The traction capability evaluation includes: mapping cohesion, internal friction angle, and shear deformation modulus to the current wheel-end force demand of the vehicle to generate traction capability data. The traction capability data includes the maximum supportable traction force, traction support margin, data collection time, data collection area, and traction-limited state.

[0030] The braking capacity evaluation includes: extracting the braking force demand in the braking direction from the current wheel end demand force of the vehicle, and generating braking capacity data by combining vehicle speed, wheel speed, tire slip ratio and the traction capacity data.

[0031] The shear-type evaluation data includes traction capacity data, braking capacity data, acquisition time, acquisition area, traction-limited state, and braking-limited state.

[0032] Furthermore, in S400, the pressure-bearing capacity assessment and the subsidence risk assessment constitute a pressure-bearing assessment.

[0033] The pressure bearing capacity evaluation includes: mapping the pressure bearing capacity to the current calculated ground pressure, tire pressure, suspension height, suspension attitude and ground imprint model to generate pressure bearing capacity data. The pressure bearing capacity data includes the maximum pressure bearing, pressure bearing support margin, collection time, collection area and pressure-limited state.

[0034] The subsidence risk assessment includes: generating subsidence risk data based on the pressure bearing capacity data, current calculated ground pressure, tire pressure, suspension status, and ground imprint model, and distinguishing between subsidence risk ahead and current subsidence risk according to the collection area.

[0035] The pressure-bearing evaluation data includes pressure-bearing capacity data, subsidence risk data, collection time, collection area, pressure-limited state, and subsidence risk state.

[0036] Further, in S500, the calculation of the passability coefficient and the generation of passability evaluation data according to a preset grading threshold includes: The shear-type evaluation data and the pressure-type evaluation data from the same collection area are merged to generate shear-side results and pressure-side results.

[0037] Read the maximum supportable traction force from the traction capacity data, the maximum bearable pressure from the pressure bearing capacity data, the current wheel end demand force of the vehicle, and the current calculated ground pressure.

[0038] The traction support margin is generated based on the ratio of the maximum supportable traction force to the current wheel end demand force of the vehicle, and the pressure support margin is generated based on the ratio of the maximum bearable pressure to the current calculated ground pressure. The traction support margin and the pressure support margin are weighted with the same weight to obtain the passability coefficient. The same weight is 0.5 for each of them, and the same weight is calibrated through a real vehicle orthogonal test.

[0039] When the passability coefficient is greater than or equal to 1.5, the passability evaluation data is recorded as excellent passability evaluation data; when the passability coefficient is greater than or equal to 1.0 and less than 1.5, the passability evaluation data is recorded as good passability evaluation data; when the passability coefficient is greater than or equal to 0.7 and less than 1.0, the passability evaluation data is recorded as intermediate passability evaluation data; when the passability coefficient is less than 0.7, the passability evaluation data is recorded as poor passability evaluation data; the preset grading threshold is obtained by calibrating the off-road driving test data.

[0040] When the data collection area is the soil area in front of the vehicle, the passability evaluation data is recorded as forward passability evaluation data, and the vehicle speed, vehicle direction of travel, and forward subsidence risk status are retained.

[0041] When the data collection area is the current grounding area, the passability evaluation data is recorded as the current passability evaluation data, and the vehicle's current wheel end demand force, current calculated grounding pressure, tire pressure and suspension status are retained.

[0042] When the shear-type evaluation data or the pressure-type evaluation data contains evaluation records pending review, invalid local area records, or temporary data identifiers, the corresponding abnormal identifiers are written into the passability evaluation data.

[0043] Furthermore, in S600, the vehicle control interface data includes torque limiting data, brake pressure adjustment data, suspension adjustment data, tire pressure adjustment data, speed limit control data, and path replanning data.

[0044] When the shear-side results include a traction-limited state, torque-limiting data is generated.

[0045] When the shear-side results include a braking-limited state, braking pressure adjustment data is generated.

[0046] When the pressure-bearing side results include a pressure-limited state, suspension adjustment data and tire pressure adjustment data are generated.

[0047] When the pressure-bearing side results include a subsidence risk state, speed limit control data and path replanning data are generated.

[0048] For the soil area in front of the vehicle, speed limit control data and path replanning data are generated first.

[0049] For the current grounding area, torque limit data, brake pressure adjustment data, suspension adjustment data, and tire pressure adjustment data are generated first.

[0050] When the soil surface image is invalid, the vehicle CAN bus read data is abnormal, or the inertial measurement unit output is abnormal, causing the passability evaluation data to contain evaluation records to be reviewed, invalid local area records, or temporary data identifiers, filtering and smoothing processing is performed based on the valid vehicle control interface data of the previous acquisition cycle and the current vehicle control interface data. The filtering and smoothing processing includes limiting the change range of torque limit, brake pressure adjustment, suspension adjustment, tire pressure adjustment, speed limit control, or path replanning trigger state within adjacent acquisition cycles.

[0051] Furthermore, an online automatic detection system for field soil mechanical properties is characterized by comprising: The in-situ detection module is used to acquire soil surface images of the soil area in front of the vehicle or the current grounding area without contacting the soil. It establishes a unified acquisition time reference based on the inertial measurement unit clock, and reads vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel end torque command and current calculated grounding pressure through the vehicle CAN bus. It combines vehicle speed, wheel speed difference, suspension displacement compensation and tire slip ratio to perform position projection correction on the soil area in front of the vehicle, and maps the current grounding area based on the wheel center coordinates and grounding imprint model to obtain in-situ detection data.

[0052] The feature extraction module, connected to the in-situ detection module, is used to perform image correction, effective region extraction, particle segmentation, texture analysis, particle size distribution statistics, sphericity calculation, surface roughness extraction, porosity estimation, and vehicle state extraction processing based on the in-situ detection data, to obtain image feature data and vehicle state data.

[0053] The mechanical inversion module, connected to the feature extraction module, is used to input the image feature data into the pre-trained mechanical performance inversion model to obtain a mechanical parameter array, and to establish a correspondence between the mechanical parameter array and the vehicle state data according to the acquisition time and acquisition area.

[0054] The dynamic evaluation module, connected to the mechanical inversion module, is used to perform traction capacity evaluation, braking capacity evaluation, pressure bearing capacity evaluation, and subsidence risk evaluation based on the mechanical parameter array and the vehicle state data, to obtain shear evaluation data and pressure bearing evaluation data.

[0055] The passability evaluation module, connected to the dynamic evaluation module, is used to calculate the passability coefficient of the soil area in front of the vehicle or the current ground contact area based on the shear evaluation data and the pressure evaluation data, and generate passability evaluation data according to a preset grading threshold.

[0056] The control interface module, connected to the passability evaluation module, is used to generate vehicle control interface data based on the passability evaluation data. When the passability evaluation data contains evaluation records to be reviewed, invalid local area records, or temporary data identifiers, the vehicle control interface data is filtered and smoothed, and the processed vehicle control interface data is output to the vehicle control system.

[0057] The key innovations of this invention include: (1) Input the image feature data into the pre-trained mechanical property inversion model to obtain the mechanical parameter array, and establish a correspondence between the mechanical parameter array and the vehicle status data according to the collection time and collection area, so that the soil surface image is no longer used only for surface recognition, but enters the correlation processing link of cohesion, internal friction angle, shear deformation modulus and bearing capacity.

[0058] (2) Based on the mechanical parameter array and vehicle status data, traction capacity evaluation, braking capacity evaluation, pressure bearing capacity evaluation and subsidence risk evaluation are performed to obtain shear evaluation data and pressure bearing evaluation data, so that the current wheel end demand force, current calculated ground pressure, tire pressure and suspension status of the vehicle are respectively entered into the evaluation links of the shear side and the pressure side.

[0059] (3) Based on the shear evaluation data and the pressure evaluation data, perform the passability evaluation of the soil area ahead or the current ground contact area to obtain the passability evaluation data, and generate vehicle control interface data based on the passability evaluation data, so that the soil area ahead and the current ground contact area correspond to speed limit control data, path replanning data, torque limit data, brake pressure adjustment data, suspension adjustment data and tire pressure adjustment data, respectively.

[0060] The following are its main beneficial effects: (1) In view of the problem that image feature data is difficult to be converted into mechanical parameter array in the prior art, the present invention processes the image feature data by pre-training mechanical performance inversion model, and establishes a correspondence between mechanical parameter array and vehicle status data according to the collection time and collection area, so that a traceable correspondence link is formed between soil surface image, mechanical parameter array and vehicle status data, reducing the separation of soil detection results and vehicle driving status.

[0061] (2) In view of the problem that it is difficult to evaluate the soil area in front of the vehicle and the current grounding area separately in the prior art, the present invention generates shear-type evaluation data and pressure-bearing evaluation data based on the mechanical parameter array and vehicle state data, so that the traction capacity evaluation, braking capacity evaluation, pressure-bearing capacity evaluation and subsidence risk evaluation respectively take into account the current wheel end demand force, current calculated grounding pressure, tire pressure and suspension status of the vehicle, reducing the disconnect of evaluation results caused by relying solely on experience judgment.

[0062] (3) In view of the problem that the detection results are difficult to enter the vehicle control interface data in the prior art, the present invention generates passability evaluation data based on shear evaluation data and pressure evaluation data, and generates vehicle control interface data from the passability evaluation data, so that the soil area ahead corresponds to speed limit control data and path replanning data, and the current ground contact area corresponds to torque limit data, brake pressure adjustment data, suspension adjustment data and tire pressure adjustment data, forming a continuous processing link from online detection to vehicle control system output. Attached Figure Description

[0063] Figure 1 This is a flowchart illustrating an online automatic detection method for the mechanical properties of soil in the field, provided as an embodiment of this application.

[0064] Figure 2 This is a structural block diagram of an online automatic detection system for field soil mechanical properties provided in an embodiment of this application. Detailed Implementation

[0065] Example 1: Refer to Figure 1 This is a flowchart illustrating an online automatic detection method for field soil mechanical properties provided in an embodiment of the present invention. The process includes steps S100-S600: An online automatic detection method for the mechanical properties of soil in the field, characterized by comprising: S100: Non-contact acquisition of soil surface images of the soil area in front of the vehicle or the current grounding area; based on the inertial measurement unit clock to synchronize vehicle CAN data; combined with vehicle speed, wheel speed difference, tire pressure, suspension displacement compensation, vehicle attitude, wheel end torque command, current calculated grounding pressure and tire slip ratio to perform front area projection correction and grounding area mapping to obtain in-situ detection data. S200. Based on the in-situ detection data, perform image correction, effective area extraction, particle, texture, particle size, sphericity, roughness, porosity and vehicle status extraction to obtain image feature data and vehicle status data. S300. Input the image feature data into the pre-trained mechanical performance inversion model to obtain a mechanical parameter array, and establish a corresponding relationship with the vehicle state data according to the collection time and collection area. S400. Based on the mechanical parameter array and the vehicle state data, perform traction, braking, pressure and subsidence risk assessment to obtain shear evaluation data and pressure evaluation data; S500: Calculate the passability coefficient based on the shear-type evaluation data and the pressure-type evaluation data, and generate passability evaluation data according to the preset grading threshold. S600. Based on the passability evaluation data, generate vehicle control interface data. If there are evaluation records to be reviewed, invalid local area records, or temporary data identifiers, filter and smooth the vehicle control interface data and then output it to the vehicle control system.

[0066] S100: Non-contact acquisition of soil surface images of the soil area in front of the vehicle or the current grounding area; based on the inertial measurement unit clock to synchronize vehicle CAN data; combined with vehicle speed, wheel speed difference, tire pressure, suspension displacement compensation, vehicle attitude, wheel end torque command, current calculated grounding pressure and tire slip ratio to perform front area projection correction and grounding area mapping to obtain in-situ detection data. Specifically, this step is executed collaboratively by the image acquisition module, the vehicle CAN acquisition unit, the inertial measurement and synchronization module, and the grounding area mapping module; Controller Area Network (CAN). The CAN bus is a communication bus inside the vehicle used to transmit vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel-end torque commands, and current calculated ground pressure. The image acquisition module is located near the ground contact area at the front of the vehicle or under the vehicle, facing the soil area in front of the vehicle or the current ground contact area. The soil area in front of the vehicle is the soil surface that the vehicle will pass over along the driving direction. The current ground contact area is the soil surface where the wheels come into contact with the unpaved road surface. The image acquisition module does not insert into the soil, press into the soil, or contact the soil through a probe, but instead acquires images of the soil surface. The vehicle CAN acquisition unit reads vehicle status data from the vehicle CAN bus. The inertial measurement and synchronization module receives the soil surface image, vehicle status data, and attitude change data output by the inertial measurement unit, and writes them to a unified timestamp. The ground area mapping module performs position projection correction and current ground area mapping processing based on the data after writing the unified timestamp, forming in-situ detection data, which serves as the input source for S200.

[0067] Specifically, the image acquisition module enters a ready-to-acquire state when the vehicle enters an unpaved road surface in the wild; the ready-to-acquire state is triggered by a road surface detection command issued by the vehicle control system; the road surface detection command is generated when the vehicle speed changes, the wheel speed difference increases, the tire pressure is adjusted, the suspension height changes, the vehicle body tilts, the wheel end torque command increases, or the currently calculated ground pressure increases; the image acquisition module outputs a soil surface image according to the acquisition frequency given by the vehicle control system; the acquisition frequency is determined according to the vehicle speed, with a lower frequency when the vehicle is traveling at low speed, and an increased frequency when the vehicle approaches soft soil or the wheel speed fluctuation increases; the soil surface image includes the image acquisition time, the installation position of the image acquisition module, the acquisition direction, and the acquisition area; the acquisition area is written as the soil area in front of the vehicle or the current grounding area.

[0068] Furthermore, the vehicle CAN acquisition unit reads vehicle status data in the same acquisition cycle; the vehicle status data includes vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel torque command, and currently calculated ground pressure; vehicle speed represents the vehicle's speed at the acquisition time; wheel speed represents the rotational speed of each wheel at the acquisition time; tire pressure represents the internal air pressure of each tire; suspension height represents the height between the body and the wheels; suspension attitude represents the compression state of the suspension in the left-right and front-back directions; body attitude represents the vehicle's yaw, pitch, and roll states; wheel torque command represents the driving torque or braking torque distributed to the wheel ends by the vehicle control system; currently calculated ground pressure represents the tire ground pressure calculated by the vehicle control system based on load, tire pressure, contact area, and suspension status; the vehicle CAN acquisition unit writes the above data into the status acquisition time and sends it to the in-situ detection module.

[0069] Specifically, the in-situ detection module first reads the image acquisition time of the soil surface image, then reads the status acquisition time of the vehicle status data, and writes both into a unified time reference based on the inertial measurement unit clock. The in-situ detection module aligns the soil surface image and vehicle status data according to the same time reference. This time alignment is not a simple merging, but rather selecting the status acquisition time with the smallest time difference before and after the image acquisition time. If the time difference is within a set time window, the soil surface image and vehicle status data are written into the same in-situ detection data. If the time difference exceeds the set time window, the in-situ detection module retains the soil surface image and writes the corresponding vehicle status as a pending state. After the vehicle CAN acquisition unit reads the same data in the next acquisition cycle, the in-situ detection module replaces the pending state with the read data and records the replacement time. Thus, the image-side information and vehicle-side information in the in-situ detection data have the same acquisition time basis.

[0070] Furthermore, the in-situ detection module performs grounding area mapping processing. For the soil area in front of the vehicle, the in-situ detection module reads the vehicle speed, the installation position of the image acquisition module, and the vehicle's driving direction, mapping the soil surface in front of the vehicle in the image to the soil area the vehicle is about to reach. The installation position of the image acquisition module includes the installation height, longitudinal position, and orientation. The in-situ detection module determines the near and far regions in the image based on the installation position, and then estimates the time it takes for the vehicle to reach the soil area based on the vehicle speed. For the current grounding area, the in-situ detection module establishes a correspondence between the soil area near the wheel contact in the image and the vehicle status data at the same acquisition time. When two consecutive frames of soil surface images cover the same soil area, the in-situ detection module compares the clarity, occlusion area, and brightness uniformity of the two frames and retains the frame with higher clarity. The image acquisition time of the replaced image is written into the replacement record of the in-situ detection data.

[0071] Furthermore, this step includes anomaly handling: if the soil surface image is obscured by mud or water, too dark, overexposed, obscured by the vehicle body, or partially blurred, the in-situ detection module will write the corresponding image area as the image area to be verified and retain the acquisition time; if the vehicle CAN bus does not return tire pressure, suspension height, or current calculated ground pressure within the acquisition cycle, the vehicle CAN acquisition unit will read the same data from the previous acquisition cycle and write it into a temporary data identifier; if the vehicle speed is 0 and the vehicle is in a stationary rescue state, the in-situ detection module will not perform the forward arrival time conversion, but will write the current acquisition area as the current ground area; if the vehicle is reversing, the vehicle's driving direction will be rewritten as the reversing direction, and the image acquisition module's installation position will re-participate in the ground area mapping processing according to the reversing direction.

[0072] In a specific operational scenario, a field rescue vehicle is traveling at low speed along the edge of a mudflat. The vehicle control system detects that the speed of the left front wheel is lower than that of the right front wheel, and that the current calculated ground pressure is increasing, and then issues a road surface detection command. The image acquisition module acquires an image of the soil surface approximately 3 meters in front of the vehicle. The image acquisition time is 10:25:16.420, and the acquisition area is the soil area in front of the vehicle. At 10:25:16.435, the vehicle's CAN acquisition unit reads the vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, vehicle attitude, wheel end torque command, and current calculated ground pressure. The in-situ detection module determines that the image acquisition time and the status acquisition time are within the same time window, and writes them into the same in-situ detection data. The in-situ detection module then, based on the vehicle speed and the installation position of the image acquisition module, maps the soil surface in front to the soil area that the vehicle is about to reach. If the same soil area is clearer in the next frame image, the in-situ detection module retains the next frame image and records the acquisition time of the previous frame image.

[0073] Understandably, the minimum set of core parameters in this step includes the soil surface image, image acquisition time, acquisition area, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, vehicle attitude, wheel-end torque command, and current calculated ground pressure. The installation position of the image acquisition module and the vehicle's driving direction are involved in mapping the soil area in front of the vehicle. Mud and water occlusion records, temporary data identifiers, and replacement records are supplementary information for abnormal scenarios. The in-situ detection data is generated in this step, and its image field is used by S200 for image correction, effective area extraction, and particle segmentation, while its vehicle status field is used by S200 for vehicle status extraction processing.

[0074] The technical effects of this step can be summarized as follows: this step processes soil surface images and vehicle CAN bus data under the same time reference; the soil area in front of the vehicle and the current grounding area are distinguished during the acquisition phase; and the in-situ detection data provides common input for subsequent image feature extraction and vehicle status extraction.

[0075] S200. Based on the in-situ detection data, perform image correction, effective area extraction, particle, texture, particle size, sphericity, roughness, porosity and vehicle status extraction to obtain image feature data and vehicle status data. Specifically, this step is executed collaboratively by the image processing and feature extraction module and the vehicle CAN acquisition unit. The image processing and feature extraction module receives the in-situ detection data output by S100 and reads the soil surface image, image acquisition time, acquisition area, image acquisition module installation location, and replacement record. The vehicle CAN acquisition unit reads the vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, vehicle body attitude, wheel end torque command, and current calculated ground pressure from the in-situ detection data. The image processing and feature extraction module performs image processing on the soil surface image to obtain image feature data. The vehicle CAN acquisition unit performs vehicle state extraction processing on the vehicle-side data to obtain vehicle state data. The image feature data and the vehicle state data serve as the input sources for S300.

[0076] Specifically, image correction is performed by the image processing and feature extraction module. Image correction includes distortion correction based on camera intrinsic parameters and region mapping based on camera extrinsic parameters. Camera intrinsic parameters represent the imaging parameters of the image acquisition module itself, including focal length, principal point position, and distortion parameters. Camera extrinsic parameters represent the installation position and orientation of the image acquisition module relative to the vehicle. The image processing and feature extraction module first corrects the image edge curvature and scale distortion based on the camera intrinsic parameters, and then maps the soil region in the image to the vehicle coordinate system based on the camera extrinsic parameters. The vehicle coordinate system uses the vehicle's driving direction as the longitudinal direction and the vehicle's lateral direction as the lateral direction. The corrected soil surface image retains the acquisition time and acquisition area and is then used for effective region extraction.

[0077] Further, effective region extraction is used to obtain the soil region to be analyzed from the corrected soil surface image; the soil region to be analyzed is the soil image region that participates in particle segmentation, texture analysis, and porosity estimation; the image processing and feature extraction module first determines the candidate soil region based on the grounding region mapping result, and then removes the vehicle body component region, water accumulation region, stone region, and vegetation region; the vehicle body component region is represented by fixed edges and fixed positions; the water accumulation region is represented by continuous highlights or specular reflections; the stone region is represented by large boundaries and obvious shadows; the vegetation region is represented by continuous strip-shaped or sheet-shaped regions with different colors from the soil; the removed regions are written into the invalid local region record, and the remaining regions are written into the soil region to be analyzed.

[0078] Specifically, particle segmentation is performed on the soil area to be analyzed; particle segmentation is used to determine particle contour data; the image processing and feature extraction module distinguishes the particle area from the interparticle void area based on brightness differences, boundary changes, and texture breakage locations; for sandy soil and gravel pavement, particle segmentation mainly reads the outer edge of the particles and the gaps between particles; for clay and tidal flat pavement, particle segmentation mainly reads the block texture, crack boundaries, and areas of fine particle aggregation; if the soil area to be analyzed is obscured by mud and water, the image processing and feature extraction module does not extract particle contours from the obscured area, but instead writes the obscured area into the invalid local area record; the particle contour data is then used for particle size distribution statistics and sphericity calculation.

[0079] Furthermore, texture analysis is used to read the surface texture changes of the soil area to be analyzed; the image processing and feature extraction module identifies fine texture, coarse texture, cracked texture, and compacted texture based on brightness fluctuations, texture direction, and local boundary continuity; fine texture corresponds to soil surfaces with densely arranged fine particles; coarse texture corresponds to areas with exposed particles or obvious surface undulations; cracked texture corresponds to cracked areas on the surface of clay or silty soil; compacted texture corresponds to flat areas formed after wheel compaction; the texture analysis results and the collected area are written together into the initial record of surface texture data and then proceed to surface roughness extraction.

[0080] Specifically, particle size distribution statistics are based on particle outline data; the image processing and feature extraction module converts the pixel scale of the particle outline in the image into the actual size; the conversion is based on the camera extrinsic parameters and grounding area mapping results; the image processing and feature extraction module then classifies the particle size into coarse particle range, medium particle range, and fine particle range; the coarse particle range corresponds to particles with a diameter greater than 2 mm; the medium particle range corresponds to particles with a diameter from 0.075 mm to 2 mm; the fine particle range corresponds to particles with a diameter less than 0.075 mm; the image processing and feature extraction module counts the number of particles, area ratio, and regional distribution location in each particle size range to obtain distribution data; occluded areas are not included in the statistics.

[0081] Furthermore, sphericity calculation is based on particle contour data; sphericity represents the degree to which the particle contour is close to a circle; the image processing and feature extraction module reads the length direction, contour curvature, and edge integrity of the particle contour to generate particle shape data; for sandy soil scenes, particle shape data records the proportion of round particles; for gravel scenes, particle shape data records the proportion of irregular particles; for clay and tidal flat scenes, particle shape data records the shape state of blocky texture areas; sphericity calculation does not depend on vehicle status data, but retains the acquisition time and acquisition area, making it easy for the S300 to associate particle shape data with vehicle status data.

[0082] Furthermore, surface roughness extraction is based on texture analysis results and particle contour data; surface roughness represents the undulation of the soil surface, particle exposure, and texture change state; the image processing and feature extraction module reads the texture brightness changes, particle edge density, and local shadow distribution to generate surface texture data; when the soil surface image of the current grounding area enters the surface roughness extraction, the image processing and feature extraction module also reads the vehicle speed and suspension attitude to record the vehicle's running state during image acquisition; this vehicle running state does not change the image feature extraction results, but only enters S300 along with the surface texture data.

[0083] Specifically, porosity estimation is based on particle contour data and surface texture data; porosity represents the area of ​​the inter-particle void region relative to the soil region being analyzed; the image processing and feature extraction module first determines the particle region based on the particle contour, and then determines the inter-particle void region based on the dark areas and gap boundaries between particles; subsequently, the image processing and feature extraction module statistically analyzes the pore region area, pore distribution location, and pore continuity state to obtain pore structure data; for areas continuously covered by water accumulation, the image processing and feature extraction module does not include them in the pore structure data, but retains them as invalid local regions.

[0084] Furthermore, the vehicle status extraction and processing is performed by the vehicle CAN acquisition unit. The vehicle CAN acquisition unit reads vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel torque command, and current calculated ground pressure from the in-situ detection data. The vehicle CAN acquisition unit first determines whether each data point was acquired at the same time, and then determines whether each data point has a temporary data identifier. When the wheel torque command is in the driving direction, the vehicle CAN acquisition unit converts it into driving demand. When the wheel torque command is in the braking direction, the vehicle CAN acquisition unit converts it into braking demand. The driving demand and braking demand together form the current wheel end demand force of the vehicle. The vehicle CAN acquisition unit writes the current wheel end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and body attitude into the vehicle status data.

[0085] In the aforementioned driving scenario at the edge of the mudflats, the image processing and feature extraction module reads the image of the soil surface in front of the vehicle generated by S100; after image correction, the module removes the shadow area of ​​the vehicle body at the lower edge of the image and the water accumulation area on the right, retaining the muddy soil surface in the middle as the soil area to be analyzed; particle segmentation obtains the fine particle aggregation area and the crack boundary; texture analysis determines that the middle of the image is a compacted texture and the left side of the image is a cracked texture; particle size distribution statistics obtain the distribution data with a high proportion of fine particles; sphericity calculation obtains the particle shape data of the blocky texture area; porosity estimation, after excluding the water accumulation area, obtains the pore structure data; the vehicle CAN acquisition unit simultaneously reads the tire pressure and suspension attitude, and generates the current wheel end demand force of the vehicle based on the wheel end torque command; the above image-side results and vehicle-side results are jointly entered into S300.

[0086] Understandably, the image feature data in this step consists of distribution data, particle shape data, surface texture data, and pore structure data; the vehicle state data consists of the vehicle's current wheel end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and vehicle body attitude; the distribution data, particle shape data, surface texture data, and pore structure data in the image feature data serve as inputs to the pre-trained mechanical performance inversion model in S300; the vehicle state data serves as the vehicle-side input for establishing corresponding relationships in S300; invalid local area records and temporary data identifiers are also transmitted to S300.

[0087] The technical effect of this step can be summarized as follows: this step splits the in-situ detection data into image-side features and vehicle-side states; the image-side features cover particles, textures, roughness, and pore structure; the vehicle-side states retain wheel end requirements, ground pressure, tire pressure, and suspension states, providing input for subsequent mechanical inversion and vehicle state association.

[0088] S300. Input the image feature data into the pre-trained mechanical performance inversion model to obtain a mechanical parameter array, and establish a corresponding relationship with the vehicle state data according to the collection time and collection area. Specifically, this step is executed collaboratively by the mechanical performance inversion module and the offline training subsystem. The mechanical performance inversion module receives image feature data and vehicle state data output by S200. The image feature data includes distribution data, particle shape data, surface texture data, and pore structure data. The vehicle state data includes the vehicle's current wheel-end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and vehicle body attitude. The pre-trained mechanical performance inversion model is stored in the mechanical performance inversion module. The pre-trained mechanical performance inversion model is a model pre-trained by the offline training subsystem. The mechanical performance inversion module initiates model calculation when the image feature data is updated, the vehicle's current wheel-end demand force increases, the current calculated ground pressure increases, the vehicle speed changes, the tire pressure changes, or the suspension height changes.

[0089] Specifically, the offline training subsystem includes a physical test calibration module, a discrete element method (DEM) simulation module, and a model training module. The physical test calibration module receives typical soil samples and acquires direct shear test data, bearing plate test data, penetration test data, and triaxial test data. The direct shear test data reflects the shear state of the soil under shear stress; the bearing plate test data reflects the settlement state of the soil under vertical pressure; the penetration test data reflects the resistance change when the probe is pressed into the soil; and the triaxial test data reflects the stress-strain state of the soil under confining pressure. The discrete element method (DEM) simulation module receives the above test data and calibrates the particle contact parameters to obtain a reference parameter set. The reference parameter set includes particle size distribution, density, particle shape, water content, and particle contact parameters.

[0090] Furthermore, the discrete element simulation module generates virtual soil surface images and corresponding mechanical parameter labels based on a benchmark parameter set. The virtual soil surface images are soil surface images generated by the discrete element simulation module under different particle size distributions, densities, particle shapes, and water content conditions. The mechanical parameter labels are cohesion, internal friction angle, shear deformation modulus, and bearing capacity corresponding to the virtual soil surface images. Cohesion represents the bonding state between soil particles; internal friction angle represents the frictional resistance state of soil particles under shear action; shear deformation modulus represents the deformation state of soil under shear action; bearing capacity represents the soil's ability to withstand tire ground pressure. The model training module reads the virtual soil surface images and extracts distribution data, particle shape data, surface texture data, and pore structure data in the S200 manner. The above image feature data and mechanical parameter labels constitute a paired training set.

[0091] Specifically, the model training module trains the neural network model to obtain a pre-trained mechanical property inversion model. The neural network model receives distribution data, particle shape data, surface texture data, and pore structure data, and outputs cohesion, internal friction angle, shear deformation modulus, and bearing capacity. After training, the model training module writes the model parameter file into the mechanical property inversion module. The model parameter file contains the model version number, training sample source, training completion time, and applicable soil type. When the vehicle is running online, the mechanical property inversion module only calls the model parameter file for forward calculation and does not re-perform direct shear tests, bearing plate tests, penetration tests, triaxial tests, or discrete element simulations at the vehicle end.

[0092] Furthermore, after receiving the image feature data output by S200, the mechanical performance inversion module first checks whether the distribution data, particle shape data, surface texture data, and pore structure data are complete. If all four types of data exist, the mechanical performance inversion module organizes them into the same model input record and calls the pre-trained mechanical performance inversion model. If a certain type of data is incomplete due to invalid local area records, the mechanical performance inversion module reads the image feature data of the remaining valid areas and retains the verification mark in the output result. If the invalid area coverage exceeds a set proportion, the mechanical performance inversion module does not output the final mechanical parameter array, but generates a verification evaluation record for S400 to read. The set proportion can be set according to half of the area of ​​the soil to be analyzed, or it can be configured by the vehicle control system according to the road condition level.

[0093] Specifically, the pre-trained mechanical performance inversion model outputs an array of mechanical parameters, including cohesion, internal friction angle, shear modulus, and bearing capacity. The mechanical performance inversion module establishes a correspondence between these mechanical parameters and vehicle state data according to the acquisition time and acquisition area. For the soil area in front of the vehicle, the mechanical performance inversion module attaches the mechanical parameter array to the soil area the vehicle is about to reach, while retaining the vehicle speed and driving direction. For the current ground contact area, the mechanical performance inversion module attaches the mechanical parameter array to the wheel contact position, while retaining the vehicle's current wheel end demand force, current calculated ground contact pressure, tire pressure, and suspension status. After the attachment is completed, the mechanical parameter array and vehicle state data together form an association record. This association record is used by the S400 to perform traction capacity evaluation, braking capacity evaluation, bearing capacity evaluation, and subsidence risk evaluation.

[0094] In the aforementioned driving scenario at the edge of the tidal flats, the mechanical performance inversion module receives the distribution data with a high proportion of fine particles generated by S200, the particle shape data of the blocky texture area, the surface texture data corresponding to the crack texture and compaction texture, and the pore structure data after excluding the water accumulation area. The mechanical performance inversion module reads the currently stored pre-trained mechanical performance inversion model version. This version is trained from the direct shear test, bearing plate test, penetration test and triaxial test data of tidal flat soil, clay and sand samples. After the model is calculated, the mechanical performance inversion module outputs cohesion, internal friction angle, shear deformation modulus and bearing capacity. Subsequently, the mechanical performance inversion module establishes a correspondence between this mechanical parameter array and the vehicle speed, wheel speed, tire pressure, suspension attitude and the currently calculated ground pressure at the same acquisition time, and writes it as an association record of the soil area in front of the vehicle.

[0095] Furthermore, this step sets up model version management; when the offline training subsystem updates the pre-trained mechanical performance inversion model, it does not directly overwrite the model parameter file currently running on the vehicle; the model training module first generates a new model version number and writes the training sample source and applicable soil type; the mechanical performance inversion module receives the updated file when the vehicle is stationary or running under low load; after the update is completed, the mechanical performance inversion module retains the previous version model parameter file; if the new version model outputs more records to be reviewed and evaluated during the continuous acquisition period, the mechanical performance inversion module switches back to the previous version model parameter file and records the switching time; this version management process ensures that the S300 model output remains traceable during vehicle operation.

[0096] Understandably, the minimum set of core parameters in this step consists of distribution data, particle shape data, surface texture data, pore structure data, pre-trained mechanical property inversion model, acquisition time, acquisition area, and vehicle status data; direct shear test data, bearing plate test data, penetration test data, and triaxial test data are used for offline training; model version number and applicable soil type are used for vehicle-side model invocation; the mechanical parameter array is formed in this step and serves as the input for shear and bearing evaluations in S400; vehicle status data is entered into S400 along with the mechanical parameter array.

[0097] The technical effects of this step can be summarized as follows: this step transforms image feature data into an array of mechanical parameters that can be used for vehicle ground mechanics evaluation; the offline training subsystem provides the model source, and the vehicle-mounted terminal performs online inversion; the mechanical parameter array and vehicle state data are correlated according to the collection time and collection area, which facilitates subsequent evaluation and processing.

[0098] S400. Based on the mechanical parameter array and the vehicle state data, perform traction, braking, pressure and subsidence risk assessment to obtain shear evaluation data and pressure evaluation data; Specifically, this step is executed by the dynamics evaluation module; the dynamics evaluation module receives the mechanical parameter array, vehicle state data, and associated records output by S300; the mechanical parameter array includes cohesion, internal friction angle, shear deformation modulus, and bearing capacity; the vehicle state data includes the vehicle's current wheel-end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and vehicle body attitude; the dynamics evaluation module initiates evaluation processing when the mechanical parameter array is updated, the vehicle's current wheel-end demand force increases, the current calculated ground pressure increases, the wheel speed difference increases, the vehicle speed changes, or a sinking risk state is triggered; the evaluation processing is performed according to the collection time and collection area, without mixing data from different areas.

[0099] Specifically, the traction capacity evaluation reads cohesion, internal friction angle, shear modulus, and the vehicle's current wheel-end demand force. Cohesion, internal friction angle, and shear modulus together represent the soil shear-side state. The vehicle's current wheel-end demand force represents the driving or braking demand made by the vehicle at the wheel end at the current data collection time. The dynamic evaluation module maps the soil shear-side state in the same data collection area to the vehicle's current wheel-end demand force. When the current wheel-end demand force is higher than the traction state that the data collection area can support, the dynamic evaluation module writes the traction-limited state. When the current wheel-end demand force is within the range that the data collection area can support, the dynamic evaluation module writes the traction-available state. If the mechanical parameter array has a "to be verified" flag, the traction capacity evaluation does not output the final state but writes the "to be verified" evaluation record.

[0100] Furthermore, braking capacity evaluation is based on traction capacity data; the dynamics evaluation module extracts the braking direction demand force from the current wheel-end demand force of the vehicle and reads the vehicle speed and wheel speed; the braking direction demand force represents the braking torque demand distributed by the vehicle control system to the wheel ends; the dynamics evaluation module maps the braking direction demand force to the traction capacity data; when the vehicle speed is high, the wheel speed difference is large, and the traction capacity data is traction-limited, the dynamics evaluation module writes the braking-limited state; when the vehicle speed is low, the wheel speed consistency is good, and the traction capacity data is traction-available, the dynamics evaluation module writes the braking-available state; when the vehicle status data has a temporary data identifier, the braking capacity evaluation retains the identifier and writes the braking capacity data as pending review.

[0101] Specifically, traction capacity data and braking capacity data together constitute shear-type evaluation data; the shear-type evaluation data retains the acquisition time, acquisition area, traction capacity data, braking capacity data, traction-limited state, and braking-limited state; for the soil area in front of the vehicle, the shear-type evaluation data represents the support state of the driving and braking in the area the vehicle is about to reach; for the current ground contact area, the shear-type evaluation data represents the support state of the driving and braking at the wheel contact position; the shear-type evaluation data is generated in this step and enters the shear-side result generation process of S500.

[0102] Furthermore, the pressure-bearing capacity evaluation reads the pressure-bearing capacity, current calculated ground pressure, tire pressure, suspension height, and suspension attitude. Pressure-bearing capacity represents the state of the soil under tire ground pressure; current calculated ground pressure represents the pressure applied by the vehicle to the soil surface; tire pressure affects the ground contact area; suspension height and suspension attitude reflect the distribution of vehicle load between the wheels; the dynamic evaluation module correlates the pressure-bearing capacity with the current calculated ground pressure in the same data collection area; when the current calculated ground pressure is higher than the pressure-bearing capacity, the dynamic evaluation module writes a pressure-limited state; when the current calculated ground pressure is lower than the pressure-bearing capacity, the dynamic evaluation module writes a pressure-available state; when tire pressure or suspension attitude has a temporary data identifier, the pressure-bearing capacity evaluation retains this identifier.

[0103] Specifically, the subsidence risk assessment is based on the bearing capacity data; the dynamic evaluation module reads the bearing capacity limitation status, the current calculated ground pressure, tire pressure, suspension height, and suspension attitude; for the soil area in front of the vehicle, if the bearing capacity data is written as bearing capacity limitation, the dynamic evaluation module writes the subsidence risk data as subsidence risk ahead; for the current ground area, if the bearing capacity data is written as bearing capacity limitation, the dynamic evaluation module writes the subsidence risk data as current subsidence risk; if the bearing capacity data is written as bearing capacity available, the dynamic evaluation module writes the subsidence risk data as subsidence risk not triggered; if there are invalid local area records in the image feature data, the dynamic evaluation module retains the invalid local area records and writes the subsidence risk data to the pending review status.

[0104] Furthermore, the pressure bearing capacity data and the subsidence risk data together constitute the pressure bearing assessment data; the pressure bearing assessment data retains the collection time, collection area, pressure bearing capacity data, subsidence risk data, pressure-limited state, and subsidence risk state; for the soil area in front of the vehicle, the pressure bearing assessment data participates in the generation of the forward passability assessment data after entering S500; for the current ground contact area, the pressure bearing assessment data participates in the generation of the current passability assessment data after entering S500; if the shear assessment data or the pressure bearing assessment data has an assessment record to be reviewed, S500 retains the record when merging and does not rewrite it to a usable state.

[0105] In the aforementioned driving scenario at the edge of the tidal flat, the dynamic evaluation module reads the low bearing capacity, low shear deformation modulus, and low internal friction angle output by S300; the vehicle status data shows that the left front wheel speed decreases, the current calculated ground pressure increases, and the wheel-end torque command is in the driving direction; the dynamic evaluation module first performs a traction capacity evaluation, corresponding the low shear side state to the vehicle's current wheel-end force demand, obtaining a traction-limited state; subsequently, the dynamic evaluation module reads the vehicle speed and wheel speed to generate braking capacity data; in the bearing capacity evaluation stage, the dynamic evaluation module corresponds the bearing capacity to the current calculated ground pressure, obtaining a bearing capacity-limited state; since the collection area is the soil area in front of the vehicle, the subsidence risk evaluation is written into the forward subsidence risk; the above processing results form shear-type evaluation data and bearing capacity evaluation data, and are sent to S500.

[0106] Understandably, the minimum set of core parameters in this step includes cohesion, internal friction angle, shear deformation modulus, bearing capacity, current wheel end demand force of the vehicle, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, and suspension status; vehicle attitude serves as supplementary data for judging vehicle load status; evaluation records to be reviewed, temporary data identifiers, and invalid local area records are transmitted along with the evaluation data as anomaly handling results; shear-type evaluation data and bearing-type evaluation data are formed in this step and serve as inputs to the shear-side results and bearing-side results in S500, respectively.

[0107] The technical effects of this step can be summarized as follows: this step splits the mechanical parameter array into two evaluation paths: shear side and bearing side; traction capacity evaluation and braking capacity evaluation form shear-type evaluation data; bearing capacity evaluation and settlement risk evaluation form bearing-type evaluation data, providing direct input for passability evaluation.

[0108] S500: Calculate the passability coefficient based on the shear-type evaluation data and the pressure-type evaluation data, and generate passability evaluation data according to the preset grading threshold. Specifically, this step is performed by the passability evaluation module; the passability evaluation module receives shear-type evaluation data and pressure-type evaluation data output by S400; the shear-type evaluation data includes traction capacity data, braking capacity data, collection time, collection area, traction-limited state, and braking-limited state; the pressure-type evaluation data includes pressure-type capacity data, subsidence risk data, collection time, collection area, pressure-limited state, and subsidence risk state; the passability evaluation module initiates passability evaluation processing when shear-type evaluation data is updated, pressure-type evaluation data is updated, the soil area in front of the vehicle changes, the current ground contact area status changes, or a pending evaluation record appears; the passability evaluation processing merges data according to the same collection time and the same collection area.

[0109] Specifically, the passability evaluation module first reads the traction-limited and braking-limited states from the shear-type evaluation data and generates shear-side results. The shear-side results represent the driving and braking states of the vehicle in the corresponding soil area. If the traction capacity data is traction-limited, the shear-side results record traction-limited; if the braking capacity data is braking-limited, the shear-side results record braking-limited; if both the traction capacity data and the braking capacity data are in a usable state, the shear-side results record shear-side usable; if the traction capacity data or the braking capacity data has evaluation records pending verification, the shear-side results retain the evaluation records pending verification and retain the corresponding collection time.

[0110] Furthermore, the pressure-bearing capacity module reads the pressure-limited state and settlement risk state from the pressure-bearing assessment data and generates pressure-bearing side results. The pressure-bearing side results represent the pressure and settlement risk state of the vehicle in the corresponding soil area. If the pressure-bearing capacity data is pressure-limited, the pressure-bearing side results record pressure-limited. If the settlement risk data is forward settlement risk or current settlement risk, the pressure-bearing side results record settlement risk state. If the pressure-bearing capacity data is pressure-available and the settlement risk has not been triggered, the pressure-bearing side results record pressure-bearing available. If the pressure-bearing capacity data or settlement risk data contains invalid local area records, the pressure-bearing side results retain the invalid local area records.

[0111] Specifically, the passability evaluation module merges the shear side and bearing side results from the same data collection area to generate passability evaluation data. For the soil area in front of the vehicle, the passability evaluation module records the passability evaluation data as forward passability evaluation data, and retains the vehicle speed, vehicle direction, and forward subsidence risk status. The forward passability evaluation data reflects the passability status of the area the vehicle is about to reach. For the current ground contact area, the passability evaluation module records the passability evaluation data as current passability evaluation data, and retains the vehicle's current wheel end demand force, current calculated ground contact pressure, tire pressure, and suspension status. The current passability evaluation data reflects the passability status of the wheel contact position. The forward passability evaluation data and the current passability evaluation data correspond to different vehicle control interface data after entering the S600.

[0112] Furthermore, the passability evaluation module reads the maximum supportable traction force from the traction capacity data, the maximum bearable pressure from the pressure bearing capacity data, the current wheel-end demand force of the vehicle, and the current calculated ground pressure. It generates a traction support margin based on the ratio of the maximum supportable traction force to the current wheel-end demand force, and a pressure bearing support margin based on the ratio of the maximum bearable pressure to the current calculated ground pressure. The traction support margin and the pressure bearing support margin are then weighted equally to obtain a passability coefficient; each weight is 0.5 and is calibrated through orthogonal tests on actual vehicles; the preset grading threshold is obtained from field driving test data.

[0113] Furthermore, the passability evaluation module sets a region priority rule; when there is a risk of subsidence in the soil area ahead, the passability evaluation module writes the area into the forward passability evaluation data and retains the vehicle speed and driving direction; when the current ground contact area has traction limitation, braking limitation, or pressure limitation, the passability evaluation module writes the current passability evaluation data as the current limited state and retains the vehicle's current wheel end demand force, current calculated ground contact pressure, tire pressure, and suspension status; when both the soil area ahead and the current ground contact area have limited states, the passability evaluation module first generates the current passability evaluation data and then generates the forward passability evaluation data; this order allows the S600 to process the vehicle control interface data of the current ground contact area first.

[0114] Furthermore, the passability evaluation module sets anomaly merging rules; if the shear side result has a record pending review, while the pressure side result is in an available state, the passability evaluation module writes the passability evaluation data as passability evaluation data pending review; if the pressure side result has an invalid local area record, while the shear side result is in an available state, the passability evaluation module retains the invalid local area record and writes it into the review trigger record; if the temporary data identifier comes from the vehicle status data, the passability evaluation module transmits it to S600 along with the passability evaluation data; the passability evaluation module does not delete abnormal records, nor does it directly rewrite abnormal records to an available state.

[0115] In the aforementioned driving scenario at the edge of the mudflats, the passability evaluation module receives the traction-limited state, pressure-limited state, and forward subsidence risk output by S400. The passability evaluation module first merges shear-type evaluation data and pressure-type evaluation data in the soil area in front of the vehicle to form shear-side results and pressure-side results. The shear-side results record traction limitation; the pressure-side results record pressure limitation and forward subsidence risk. Since this area is the soil area that the vehicle is about to reach, the passability evaluation module generates forward passability evaluation data and retains the vehicle speed and driving direction. If the vehicle subsequently enters this area, the image acquisition module acquires the image of the current ground contact area, and S100 to S400 regenerate the evaluation data of the current ground contact area. The passability evaluation module then generates the current passability evaluation data.

[0116] Understandably, the minimum set of core parameters in this step includes shear side results, bearing side results, acquisition time, acquisition area, vehicle speed, vehicle direction of travel, current wheel end demand force, current calculated ground pressure, tire pressure, and suspension status. Forward passability evaluation data is used by S600 to generate speed limit control data and path replanning data. Current passability evaluation data is used by S600 to generate torque limit data, brake pressure adjustment data, suspension adjustment data, and tire pressure adjustment data. Evaluation records pending review, temporary data identifiers, and invalid local area records are entered into S600 along with the passability evaluation data.

[0117] The technical effects of this step can be summarized as follows: this step merges shear-type evaluation data and pressure-type evaluation data into passability evaluation data; corresponding evaluation results are generated for the soil area ahead and the current grounding area; the passability evaluation data converts the evaluation results of S400 into vehicle control interface inputs that can be called by S600.

[0118] S600. Based on the passability evaluation data, generate vehicle control interface data. If there are evaluation records to be reviewed, invalid local area records, or temporary data identifiers, filter and smooth the vehicle control interface data and output it to the vehicle control system. Specifically, this step is executed by the control interface module; the control interface module receives the passability evaluation data output by S500; the passability evaluation data includes forward passability evaluation data or current passability evaluation data, and retains the shear side result, pressure side result, collection time, collection area, vehicle speed, vehicle driving direction, current wheel end demand force, current calculated ground pressure, tire pressure and suspension status; the control interface module starts interface processing when the passability evaluation data is updated, traction limitation state occurs, braking limitation state occurs, pressure limitation state occurs, sinking risk state occurs, or review trigger record occurs; the control interface module does not directly change the vehicle actuators, but generates vehicle control interface data and outputs it to the vehicle control system.

[0119] Specifically, the vehicle control interface data includes torque limiting data, brake pressure adjustment data, suspension adjustment data, tire pressure adjustment data, speed limit control data, and path replanning data. Torque limiting data is the interface data received by the vehicle control system to limit the wheel-end drive torque; brake pressure adjustment data is the interface data received by the vehicle control system to adjust the brake pressure; suspension adjustment data is the interface data received by the vehicle control system to adjust the suspension height or suspension attitude; tire pressure adjustment data is the interface data received by the vehicle control system to adjust the tire pressure; speed limit control data is the interface data received by the vehicle control system to limit the vehicle's speed; and path replanning data is the interface data received by the vehicle control system or path planning unit to adjust the driving path.

[0120] Furthermore, the control interface module first reads the shear-side results; if the shear-side results include a traction-limited state, the control interface module generates torque-limiting data; the torque-limiting data reads the vehicle's current wheel-end demand force, vehicle speed, wheel speed, and the data collection area; for the current ground contact area, the torque-limiting data directly corresponds to the current wheel-end drive demand; for the soil area in front of the vehicle, the torque-limiting data is not immediately issued as an execution command, but enters the vehicle control system along with the speed limit control data; if the shear-side results include a braking-limited state, the control interface module generates brake pressure adjustment data; the brake pressure adjustment data reads the braking direction demand force, vehicle speed, and wheel speed, and writes the braking-limited source.

[0121] Furthermore, the control interface module reads the pressure-bearing side results; if the pressure-bearing side results include a pressure-limited state, the control interface module generates suspension adjustment data and tire pressure adjustment data; the suspension adjustment data reads the suspension height, suspension attitude, and current calculated ground pressure; the tire pressure adjustment data reads the tire pressure and current calculated ground pressure; if the pressure-bearing side results include a subsidence risk state, the control interface module generates speed limit control data and path replanning data; the speed limit control data reads the vehicle speed, collection area, and subsidence risk state; the path replanning data reads the vehicle's driving direction, collection area, and forward subsidence risk state; speed limit control data and path replanning data are prioritized for the soil area in front of the vehicle; torque limit data, brake pressure adjustment data, suspension adjustment data, and tire pressure adjustment data are prioritized for the current ground contact area.

[0122] Specifically, the control interface module sets interface priorities; when the current ground contact area has traction limitations, braking limitations, or pressure limitations, the control interface module first outputs the torque limit data, brake pressure adjustment data, suspension adjustment data, and tire pressure adjustment data corresponding to the current ground contact area; when there is a risk of subsidence in the soil area in front of the vehicle, the control interface module outputs speed limit control data and path replanning data; if the current ground contact area and the soil area in front of the vehicle are triggered simultaneously, the control interface module first outputs the interface data corresponding to the current ground contact area, and then outputs the interface data corresponding to the soil area in front of the vehicle; if the passability evaluation data has a record of evaluation to be reviewed, the control interface module generates review prompt data and outputs it to the vehicle control system along with the vehicle control interface data.

[0123] Furthermore, when the soil surface image is invalid, the vehicle CAN bus read data is abnormal, or the inertial measurement unit output is abnormal, the control interface module retains the evaluation record to be reviewed, the invalid local area record, or the temporary data identifier in the passability evaluation data, and performs filtering and smoothing processing on the current vehicle control interface data based on the valid vehicle control interface data of the previous acquisition cycle; the filtering and smoothing processing is used to limit the change range of torque limit, brake pressure adjustment, suspension adjustment, tire pressure adjustment, speed limit control, or path replanning trigger state in adjacent acquisition cycles.

[0124] Furthermore, the control interface module sets boundary constraints and exception handling; if the vehicle status data contains a temporary data identifier, the control interface module does not delete the identifier, but writes a temporary source record into the vehicle control interface data; if the image feature data contains an invalid local region record, the control interface module writes the invalid local region record into the vehicle control interface data; if the vehicle CAN bus is temporarily unavailable, the control interface module writes the vehicle control interface data into the interface buffer and records the transmission time; after the vehicle CAN bus is restored, the control interface module sends the data in the interface buffer according to the acquisition time sequence; if the data in the interface buffer has exceeded the time range allowed by the vehicle control system, the control interface module does not send the data, but writes an expired record.

[0125] In the aforementioned driving scenario at the edge of the mudflats, the control interface module receives forward passability evaluation data generated by S500. This data includes traction-limited status, pressure-limited status, and forward subsidence risk. Since the collection area is the soil area in front of the vehicle, the control interface module first generates speed limit control data and path replanning data. The speed limit control data includes the current vehicle speed, forward subsidence risk status, and collection area. The path replanning data includes the vehicle's driving direction and the location of the soil area ahead. If the vehicle subsequently enters this area, S500 generates current passability evaluation data, and the control interface module then generates torque limit data based on the traction-limited status of the current ground contact area, and generates suspension adjustment data and tire pressure adjustment data based on the pressure-limited status. After the above vehicle control interface data is sent to the vehicle control system, the control interface module records the sending time, collection area, and interface data type.

[0126] Understandably, the minimum set of core parameters in this step includes passability evaluation data, shear side results, bearing side results, acquisition area, current wheel end demand force of the vehicle, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, and suspension status. Vehicle control interface data is generated in this step and output to the vehicle control system. After the output is completed, the changes in the operating status of the vehicle control system are entered into S100 through the vehicle CAN bus, thereby generating new in-situ detection data in the next acquisition cycle.

[0127] The technical effects of this step can be summarized as follows: this step transforms the passability evaluation data into vehicle control interface data; the soil area ahead corresponds to speed limit control and path replanning, while the current ground contact area corresponds to torque, braking, suspension, and tire pressure adjustment; after the vehicle control interface data is output, the vehicle status changes enter the next acquisition cycle, forming a continuous operation link.

[0128] Example 2: Figure 2 This diagram illustrates a structural block diagram of an online automatic detection system for field soil mechanical properties according to an embodiment of the present invention; as shown. Figure 2 As shown, the structure may include: The in-situ detection module 01 is used to acquire soil surface images of the soil area in front of the vehicle or the current ground contact area without contacting the soil. It establishes a unified acquisition time reference based on the inertial measurement unit clock, and reads vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel end torque command, and current calculated ground contact pressure via the vehicle CAN bus. Combining vehicle speed, wheel speed difference, suspension displacement compensation, and tire slip ratio, it performs position projection correction on the soil area in front of the vehicle, and maps the current ground contact area based on wheel center coordinates and a ground contact imprint model to obtain in-situ detection data. This module receives soil surface images output by the image acquisition unit, clock signals and attitude change data output by the inertial measurement unit, and vehicle status data returned by the vehicle CAN bus. When the soil area in front of the vehicle enters the processing phase, this module... The vehicle's driving direction and the image acquisition unit's installation position determine the image area's location in the vehicle coordinate system. This is then combined with vehicle speed, wheel speed difference, suspension displacement compensation, and tire slip ratio to correct the mapping relationship between the vehicle and the corresponding soil area. When the current grounding area enters processing, this module reads the wheel center coordinates, tire pressure, and suspension attitude, calls the grounding imprint model to determine the contact range between the wheel and the soil surface, and binds this contact range to the currently calculated grounding pressure and wheel-end torque commands at the same acquisition moment, forming in-situ detection data with the acquisition time and area. When the soil surface image is obscured by mud or water, too dark, or partially blurred, this module writes a record to be reviewed and evaluated. When vehicle CAN bus return data is missing, this module retains valid data with the same name from the previous acquisition cycle and writes a temporary data identifier. The in-situ detection data then enters the feature extraction module.

[0129] Feature extraction module 02, connected to the in-situ detection module, is used to perform image correction, effective region extraction, particle segmentation, texture analysis, particle size distribution statistics, sphericity calculation, surface roughness extraction, porosity estimation, and vehicle state extraction processing based on the in-situ detection data, to obtain image feature data and vehicle state data. This module reads soil surface image, acquisition time, acquisition area, image acquisition unit installation position, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, vehicle body attitude, wheel end torque command, and current calculated ground pressure from the in-situ detection data. During image processing, this module first corrects distortion based on camera intrinsic parameters, then completes region mapping based on camera extrinsic parameters and the acquisition area, and subsequently removes vehicle component areas and accumulated water from the corrected soil surface image. The module identifies the soil region to be analyzed by dividing it into three areas: a rocky area, a vegetation area, and a soil area. Within this soil region, it extracts particle outlines, texture variations, particle size distribution, particle shape, surface undulations, and interparticle porosity, writing these data into distribution data, particle shape data, surface texture data, and pore structure data, respectively. During vehicle-side processing, the module distinguishes between driving and braking directions based on wheel-end torque commands, generates the vehicle's current wheel-end demand force, and writes the current wheel-end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and vehicle attitude into the vehicle status data. Removed image regions form invalid local region records, and vehicle status fields with temporary data identifiers are retained along with the vehicle status data for subsequent modules to determine whether to participate in filtering and smoothing processing.

[0130] The mechanical inversion module 03, connected to the feature extraction module, is used to input the image feature data into the pre-trained mechanical performance inversion model to obtain a mechanical parameter array, and establish a correspondence between the mechanical parameter array and the vehicle state data according to the acquisition time and acquisition area. After receiving the image feature data and the vehicle state data, this module first verifies the acquisition time and acquisition area of ​​the two, and establishes an inversion task for data with the same acquisition time and acquisition area. The pre-trained mechanical performance inversion model calls distribution data, particle shape data, surface texture data, and pore structure data to output cohesion, internal friction angle, shear deformation modulus, and bearing capacity. The module then converts the mechanical parameter array... The module establishes a correlation record with the vehicle's current wheel end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and vehicle body attitude. For the soil area in front of the vehicle, the mechanical parameter array is linked to the soil area the vehicle is about to reach. For the current ground contact area, the mechanical parameter array is linked to the wheel contact position. If the image feature data contains invalid local area records, this module only performs inversion on the feature data corresponding to the valid image area and passes the record to be reviewed and evaluated to the dynamic evaluation module along with the mechanical parameter array. If the model output lacks any mechanical parameter, this module retains the inversion anomaly mark for that acquisition area and does not participate in the complete dynamic evaluation of the current acquisition cycle.

[0131] The dynamic evaluation module 04, connected to the mechanical inversion module, is used to perform traction capacity evaluation, braking capacity evaluation, pressure bearing capacity evaluation, and subsidence risk evaluation based on the mechanical parameter array and the vehicle state data, obtaining shear-type evaluation data and pressure bearing type evaluation data. This module reads the cohesion, internal friction angle, shear deformation modulus, pressure bearing capacity, current wheel-end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, and suspension state at the same acquisition time and in the same acquisition area. In the traction capacity evaluation, this module correlates the cohesion, internal friction angle, and shear deformation modulus with the current wheel-end demand force of the vehicle to form the maximum supportable traction force, traction support margin, and traction-limited state. In the braking capacity evaluation, this module obtains the current wheel-end demand force from the vehicle... The module extracts the braking force in the braking direction and combines it with vehicle speed, wheel speed, and tire slip ratio to form braking capacity data. In the pressure bearing capacity evaluation, the module correlates the pressure bearing capacity with the current calculated ground pressure, tire pressure, suspension height, suspension attitude, and ground imprint model to form the maximum pressure bearing capacity, pressure bearing support margin, and pressure bearing limited state. In the subsidence risk evaluation, the module distinguishes between forward subsidence risk and current subsidence risk based on the pressure bearing capacity data, current calculated ground pressure, tire pressure, and suspension status. Both shear-type and pressure-type evaluation data retain the collection area, collection time, and anomaly markers. When the mechanical parameter array or vehicle status data contains evaluation records pending verification, the module retains the evaluation results but marks them as pending verification and passes the marker to the passability evaluation module.

[0132] The passability evaluation module 05, connected to the dynamic evaluation module, is used to calculate the passability coefficient of the soil area in front of the vehicle or the current ground contact area based on the shear-type evaluation data and the pressure-type evaluation data, and generate passability evaluation data according to a preset grading threshold. After receiving the shear-type evaluation data and the pressure-type evaluation data, this module first merges the shear-side results and the pressure-side results according to the collection area, and then reads the maximum supportable traction force, the maximum supportable pressure, the current wheel-end demand force of the vehicle, and the current calculated ground contact pressure. This module generates a traction support margin based on the ratio of the maximum supportable traction force to the current wheel-end demand force of the vehicle, and generates a pressure support margin based on the ratio of the maximum supportable pressure to the current calculated ground contact pressure. The traction support margin and the pressure support margin are weighted with equal weights of 0.5 to obtain the passability coefficient. The passability coefficient; the same weight is calibrated through orthogonal tests on actual vehicles, and the preset grading threshold is obtained from field driving test data; after comparing the passability coefficient with the preset grading threshold, excellent passability evaluation data, good passability evaluation data, medium passability evaluation data, or poor passability evaluation data are generated; the passability evaluation data corresponding to the soil area in front of the vehicle retains the vehicle speed, vehicle driving direction, and forward subsidence risk status; the passability evaluation data corresponding to the current ground contact area retains the vehicle's current wheel end demand force, current calculated ground contact pressure, tire pressure, and suspension status; if the shear-type evaluation data or the bearing-type evaluation data contains evaluation records to be reviewed, invalid local area records, or temporary data identifiers, this module writes the corresponding abnormal identifier into the passability evaluation data and sends the abnormal identifier along with the passability evaluation data to the control interface module.

[0133] Control interface module 06, connected to the passability evaluation module, is used to generate vehicle control interface data based on the passability evaluation data. When the passability evaluation data contains evaluation records to be reviewed, invalid local area records, or temporary data identifiers, the vehicle control interface data is filtered and smoothed, and the processed vehicle control interface data is output to the vehicle control system. This module reads the collection area, passability coefficient, grading result, traction-limited state, braking-limited state, pressure-limited state, and subsidence risk state from the passability evaluation data. When the current collection area is the soil area in front of the vehicle, this module prioritizes generating speed limit control data and path replanning data. When the current collection area is the current ground contact area, this module prioritizes generating torque limit data, brake pressure adjustment data, suspension adjustment data, and tire pressure adjustment data. Traction-limited state corresponds to torque limit data, braking-limited state corresponds to brake pressure adjustment data, and pressure-limited state corresponds to suspension adjustment data. Data and tire pressure regulation data, speed limit control data and path replanning data corresponding to the sinking risk state; after the vehicle control interface data is output to the vehicle control system, the control interface module receives the execution receipt of the vehicle control system, which is used to record whether the controlled object has received the corresponding data; when the soil surface image is invalid, the vehicle CAN bus read data is abnormal, or the inertial measurement unit output is abnormal, causing the passability evaluation data to contain anomaly indicators, this module reads the valid vehicle control interface data of the previous acquisition cycle and limits the change range of torque limit, brake pressure regulation, suspension adjustment, tire pressure regulation, speed limit control, or path replanning trigger state in adjacent acquisition cycles; if multiple consecutive acquisition cycles have anomaly indicators, this module maintains the valid vehicle control interface data of the previous acquisition cycle and marks the current acquisition area as the area to be reviewed, and waits for the in-situ detection module of the next acquisition cycle to regenerate the in-situ detection data before entering the new processing link.

Claims

1. A method for online automatic detection of the mechanical properties of soil in the field, characterized in that, include: S100: Non-contact acquisition of soil surface images of the soil area in front of the vehicle or the current grounding area; based on the inertial measurement unit clock to synchronize vehicle CAN data; combined with vehicle speed, wheel speed difference, tire pressure, suspension displacement compensation, vehicle attitude, wheel end torque command, current calculated grounding pressure and tire slip ratio to perform front area projection correction and grounding area mapping to obtain in-situ detection data. S200. Based on the in-situ detection data, perform image correction, effective area extraction, particle, texture, particle size, sphericity, roughness, porosity and vehicle status extraction to obtain image feature data and vehicle status data. S300. Input the image feature data into the pre-trained mechanical performance inversion model to obtain a mechanical parameter array, and establish a corresponding relationship with the vehicle state data according to the collection time and collection area. S400. Based on the mechanical parameter array and the vehicle state data, perform traction, braking, pressure and subsidence risk assessment to obtain shear evaluation data and pressure evaluation data; S500: Calculate the passability coefficient based on the shear-type evaluation data and the pressure-type evaluation data, and generate passability evaluation data according to the preset grading threshold. S600. Based on the passability evaluation data, generate vehicle control interface data. If there are evaluation records to be reviewed, invalid local area records, or temporary data identifiers, filter and smooth the vehicle control interface data and then output it to the vehicle control system.

2. The online automatic detection method for field soil mechanical properties according to claim 1, characterized in that, In S100, the synchronization of vehicle CAN data based on the inertial measurement unit clock, the forward area projection correction, and the ground area mapping include: Using the inertial measurement unit clock as a synchronization reference, a unified timestamp is written to the soil surface image, the data read from the vehicle CAN bus, and the attitude change data output by the inertial measurement unit. Tire slip ratio is generated based on vehicle speed and wheel speed, and suspension displacement compensation is generated based on suspension height changes within adjacent acquisition periods. When the soil surface image corresponds to the soil area in front of the vehicle, the front area projection correction is performed on the soil area in front of the vehicle by combining the installation position of the image acquisition module, the vehicle's driving direction, vehicle speed, attitude change data output by the inertial measurement unit, suspension displacement compensation amount and tire slip ratio, and the corrected soil area in front of the vehicle is mapped to the soil area that the vehicle is about to reach. When the soil surface image corresponds to the current grounding area, the wheel center coordinates are determined according to the preset vehicle geometric parameters, and the current grounding area is established in correspondence with the vehicle status data at the same acquisition time by combining the tire pressure, suspension height, vehicle posture and ground imprint model. When two consecutive soil surface images cover the same soil area, the image with higher clarity is retained, and the image acquisition time of the replaced image is recorded.

3. The online automatic detection method for field soil mechanical properties according to claim 2, characterized in that, In S200, the image feature data includes distribution data, particle shape data, surface texture data, and pore structure data; The image correction includes distortion correction based on camera intrinsic parameters and region mapping based on camera extrinsic parameters; The effective region extraction includes removing vehicle component areas, water accumulation areas, rock areas, and vegetation areas from the corrected soil surface image to obtain the soil area to be analyzed; The particle segmentation is used to extract particle profile data from the soil region to be analyzed. The particle size distribution statistics are used to generate distribution data based on the particle profile data; The sphericity calculation is used to generate particle shape data based on the particle contour data; The texture analysis and surface roughness extraction are used to generate surface texture data; The porosity estimation is used to generate pore structure data based on the interparticle void region. The vehicle state extraction process includes reading vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel-end torque command, and current calculated ground pressure from the in-situ detection data, and generating the current wheel-end demand force of the vehicle based on the wheel-end torque command to obtain vehicle state data.

4. The online automatic detection method for field soil mechanical properties according to claim 3, characterized in that, In S300, the pre-trained mechanical performance inversion model is generated by the offline training subsystem; The offline training subsystem acquires direct shear test data, bearing plate test data, penetration test data and triaxial test data, performs calibration processing on the discrete element simulation module, and obtains a set of reference parameters. The discrete element simulation module generates a virtual soil surface image and corresponding mechanical parameter labels based on the benchmark parameter set; Based on the virtual soil surface image, the image feature data extracted from the virtual soil surface image, and the mechanical parameter labels, a paired training set is constructed, and the neural network model is trained to obtain the pre-trained mechanical performance inversion model. The array of mechanical parameters includes cohesion, internal friction angle, shear modulus, and compressive strength.

5. The online automatic detection method for field soil mechanical properties according to claim 4, characterized in that, In S300, a correspondence is established between the mechanical parameter array and the vehicle state data according to the acquisition time and acquisition area, including: Establish a correlation record between the cohesion, internal friction angle, shear deformation modulus, and bearing capacity at the same collection time and in the same collection area and the vehicle's current wheel end demand force, current calculated ground pressure, vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, and vehicle body attitude. When the data collection area is the soil area in front of the vehicle, the mechanical parameter array is attached to the soil area that the vehicle is about to reach, while retaining the vehicle speed and vehicle direction. When the data collection area is the current grounding area, the mechanical parameter array is attached to the wheel contact position, and the vehicle's current wheel end demand force, current calculated ground pressure, tire pressure and suspension status are retained.

6. The online automatic detection method for field soil mechanical properties according to claim 5, characterized in that, In S400, the traction capacity evaluation and braking capacity evaluation constitute a shear-type evaluation; The traction capacity evaluation includes: mapping the cohesion, internal friction angle and shear deformation modulus to the current wheel end demand force of the vehicle to generate traction capacity data. The traction capacity data includes the maximum supportable traction force, traction support margin, collection time, collection area and traction-limited state. The braking capacity evaluation includes: extracting the braking direction demand force from the current wheel-end demand force of the vehicle, and generating braking capacity data by combining vehicle speed, wheel speed, tire slip ratio and the traction capacity data; The shear-type evaluation data includes traction capacity data, braking capacity data, acquisition time, acquisition area, traction-limited state, and braking-limited state.

7. The online automatic detection method for field soil mechanical properties according to claim 6, characterized in that, In S400, the pressure-bearing capacity assessment and the subsidence risk assessment constitute a pressure-bearing assessment. The pressure bearing capacity evaluation includes: mapping the pressure bearing capacity to the current calculated ground pressure, tire pressure, suspension height, suspension attitude and ground imprint model to generate pressure bearing capacity data. The pressure bearing capacity data includes the maximum pressure bearing, pressure bearing support margin, collection time, collection area and pressure-limited state. The subsidence risk assessment includes: generating subsidence risk data based on the pressure bearing capacity data, current calculated ground pressure, tire pressure, suspension status, and ground imprint model, and distinguishing between subsidence risk ahead and current subsidence risk according to the collection area; The pressure-bearing evaluation data includes pressure-bearing capacity data, subsidence risk data, collection time, collection area, pressure-limited state, and subsidence risk state.

8. The online automatic detection method for field soil mechanical properties according to claim 7, characterized in that, In S500, the calculation of the passability coefficient and the generation of passability evaluation data according to the preset grading threshold include: The shear-type evaluation data and the pressure-type evaluation data under the same collection area are merged to generate shear-side results and pressure-side results; Read the maximum supportable traction force from the traction capacity data, the maximum bearable pressure from the pressure bearing capacity data, the current wheel end demand force of the vehicle, and the current calculated ground pressure; The traction support margin is generated based on the ratio of the maximum supportable traction force to the current wheel end demand force of the vehicle, and the pressure support margin is generated based on the ratio of the maximum bearable pressure to the current calculated ground pressure. The traction support margin and the pressure support margin are weighted with the same weight to obtain the passability coefficient. The same weight is 0.5 for each of them, and the same weight is calibrated through a real vehicle orthogonal test. When the passability coefficient is greater than or equal to 1.5, the passability evaluation data is recorded as excellent passability evaluation data; when the passability coefficient is greater than or equal to 1.0 and less than 1.5, the passability evaluation data is recorded as good passability evaluation data; when the passability coefficient is greater than or equal to 0.7 and less than 1.0, the passability evaluation data is recorded as intermediate passability evaluation data; when the passability coefficient is less than 0.7, the passability evaluation data is recorded as poor passability evaluation data; the preset grading threshold is obtained by calibrating the off-road driving test data; When the data collection area is the soil area in front of the vehicle, the passability evaluation data is recorded as forward passability evaluation data, and the vehicle speed, vehicle direction of travel and forward subsidence risk status are retained. When the data collection area is the current grounding area, the passability evaluation data is recorded as the current passability evaluation data, and the vehicle's current wheel end demand force, current calculated ground pressure, tire pressure and suspension status are retained; When the shear-type evaluation data or the pressure-type evaluation data contains evaluation records pending review, invalid local area records, or temporary data identifiers, the corresponding abnormal identifiers are written into the passability evaluation data.

9. The online automatic detection method for field soil mechanical properties according to claim 8, characterized in that, In S600, the vehicle control interface data includes torque limiting data, brake pressure adjustment data, suspension adjustment data, tire pressure adjustment data, speed limit control data, and path replanning data. When the shear-side results include a traction-limited state, torque-limiting data is generated; When the shear-side results include a brake-limited state, brake pressure adjustment data is generated; When the pressure-bearing side results include a pressure-limited state, suspension adjustment data and tire pressure adjustment data are generated. When the pressure-bearing side results include a subsidence risk state, speed limit control data and path replanning data are generated. For the soil area in front of the vehicle, speed limit control data and path replanning data are generated first. For the current ground contact area, prioritize generating torque limit data, brake pressure adjustment data, suspension adjustment data, and tire pressure adjustment data; When the soil surface image is invalid, the vehicle CAN bus read data is abnormal, or the inertial measurement unit output is abnormal, causing the passability evaluation data to contain evaluation records to be reviewed, invalid local area records, or temporary data identifiers, filtering and smoothing processing is performed based on the valid vehicle control interface data of the previous acquisition cycle and the current vehicle control interface data. The filtering and smoothing processing includes limiting the change range of torque limit, brake pressure adjustment, suspension adjustment, tire pressure adjustment, speed limit control, or path replanning trigger state within adjacent acquisition cycles.

10. An online automatic detection system for field soil mechanical properties, applied to the method described in any one of claims 1 to 9, characterized in that, include: The in-situ detection module is used to acquire soil surface images of the soil area in front of the vehicle or the current grounding area in a non-contact manner. It establishes a unified acquisition time reference based on the inertial measurement unit clock, and reads vehicle speed, wheel speed, tire pressure, suspension height, suspension attitude, body attitude, wheel end torque command and current calculated grounding pressure through the vehicle CAN bus. It combines vehicle speed, wheel speed difference, suspension displacement compensation and tire slip ratio to perform position projection correction on the soil area in front of the vehicle, and maps the current grounding area based on the wheel center coordinates and grounding imprint model to obtain in-situ detection data. The feature extraction module, connected to the in-situ detection module, is used to perform image correction, effective region extraction, particle segmentation, texture analysis, particle size distribution statistics, sphericity calculation, surface roughness extraction, porosity estimation, and vehicle state extraction processing based on the in-situ detection data, to obtain image feature data and vehicle state data. The mechanical inversion module, connected to the feature extraction module, is used to input the image feature data into the pre-trained mechanical performance inversion model to obtain a mechanical parameter array, and to establish a correspondence between the mechanical parameter array and the vehicle state data according to the collection time and collection area. The dynamic evaluation module, connected to the mechanical inversion module, is used to perform traction capacity evaluation, braking capacity evaluation, pressure bearing capacity evaluation and subsidence risk evaluation based on the mechanical parameter array and the vehicle state data, and obtain shear evaluation data and pressure bearing evaluation data. The passability evaluation module is connected to the dynamic evaluation module and is used to calculate the passability coefficient of the soil area in front of the vehicle or the current ground contact area based on the shear evaluation data and the pressure evaluation data, and generate passability evaluation data according to the preset grading threshold. The control interface module, connected to the passability evaluation module, is used to generate vehicle control interface data based on the passability evaluation data. When the passability evaluation data contains evaluation records to be reviewed, invalid local area records, or temporary data identifiers, the vehicle control interface data is filtered and smoothed, and the processed vehicle control interface data is output to the vehicle control system.

Images