Vehicle terrain automatic identification method and system and automobile

Abstract

The invention belongs to the field of vehicle terrain automatic identification, and discloses a vehicle terrain automatic identification method and system and an automobile, and the method comprises the steps: collecting topographic image data, chassis dynamic data and vehicle state data, inputting the topographic image data into a pre-trained terrain identification model, and obtaining a first terrain identification result; identifying the chassis dynamic data and the vehicle state data by adopting a second terrain identification model to obtain a second terrain identification result; arbitrating the first terrain recognition result and the second terrain recognition result through dual-path recognition in combination with an arbitration mechanism to obtain a final terrain recognition result; the topographic image data provides visual topographic features, the chassis dynamic data and the vehicle state data reflect the interaction condition of the vehicle and the topography, and the recognition accuracy is improved through complementation of the chassis dynamic data and the vehicle state data. The arbitration mechanism ensures that a reliable terrain recognition result can be obtained even under a complex meteorological condition, so that the accuracy and the real-time performance of terrain recognition are improved, and the safety of vehicle driving is guaranteed.

Description

Technical Field

[0001] The present invention belongs to the technical field of vehicle terrain automatic recognition, and particularly relates to a vehicle terrain automatic recognition method, system and vehicle. Background Art

[0002] With the rapid development of intelligent driving technology, the demand for vehicle adaptability to complex terrain environments has become increasingly prominent. Under harsh weather conditions such as night, heavy rain, and thick fog, vehicles may encounter special terrains such as snow, mud, and wading roads. It is difficult for drivers to accurately evaluate the road conditions through subjective judgment, and the limitations of traditional terrain recognition technologies relying on lidar or cameras are gradually exposed in extreme environments. Existing technologies mainly obtain road surface information through vision sensors or lidar, and combine multi-sensor fusion algorithms to achieve terrain classification, so as to assist drivers in switching driving modes or automatically adjusting vehicle control strategies. However, although such technologies can achieve a high recognition accuracy under normal climate conditions, in extreme weather, the performance of sensors will decline significantly, resulting in a greatly reduced recognition accuracy or even complete failure, unable to meet the safety requirements under complex working conditions.

[0003] Existing terrain recognition means rely too much on the external environment. For example, lidar is easily interfered in heavy rain or thick fog, and the imaging quality of cameras deteriorates under low light conditions, resulting in the inability to effectively extract key terrain features. In addition, although multi-sensor fusion algorithms can make up for the defects of single sensors to a certain extent, the robustness of the fusion strategy is insufficient, and it is still difficult to ensure the reliability of recognition results in extreme weather. When a vehicle enters a wading section or soft sand, if the system fails to timely remind the driver to switch modes or automatically adjust control parameters, it may lead to serious accidents such as vehicle out of control and wading submergence.

[0004] It can be seen that under complex meteorological conditions, using traditional vehicle terrain road surface automatic recognition technology cannot guarantee the accuracy and real-time performance of terrain recognition. Summary of the Invention

[0005] The present invention provides a vehicle terrain automatic recognition method, system and vehicle to solve the technical problem that using traditional vehicle terrain road surface automatic recognition technology cannot guarantee the accuracy and real-time performance of terrain recognition under complex meteorological conditions.

[0006] To achieve the above object, the present invention adopts the following technical solutions: A vehicle terrain automatic recognition method, comprising: Collecting terrain image data, chassis dynamic data and vehicle state data; Inputting the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; Input the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; Arbitrate between the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result.

[0007] Further, The collection of terrain image data, chassis dynamic data, and vehicle status data includes: Collect terrain image data; collect chassis dynamic data, where the chassis dynamic data includes xyz three-axis acceleration information, yaw angular velocity information, wheel-end braking torque, wheel speed, steering wheel angle signal, wheel-end torque, vehicle speed, vehicle mass, wind resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio; collect vehicle status data, where the vehicle status data includes wheel noise signal, ambient temperature signal, wading and water depth signal, ramp signal, and wiper status.

[0008] Further, The inputting of the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result includes: Input the terrain image data into a pre-trained terrain recognition model to obtain a first terrain recognition result; among them, the pre-trained terrain recognition model uses a depth graph convolutional neural network and is trained based on the data of the cloud vehicle networking; Upload the first terrain recognition result to the artificial intelligence terrain database of the advanced driver assistance computing center through over-the-air (OTA) technology and a telematics box.

[0009] Further, The inputting of the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result includes: Train a pre-constructed second terrain recognition model based on historical chassis dynamic data and historical vehicle status data to obtain a pre-trained second terrain recognition model; the second terrain recognition model uses a convolutional neural network, a recurrent neural network, or a long short-term memory network; Clean, integrate, and extract the collected chassis dynamic data and vehicle status data to obtain features to be recognized; Input the features to be recognized into the pre-trained second terrain recognition model to obtain a second terrain recognition result.

[0010] Further, The arbitration between the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result includes: Calculate the confidence of the first terrain recognition result by combining the initial weight of the first terrain recognition result with the environmental dynamic factor, and calculate the confidence of the second terrain recognition result by combining the initial weight of the second terrain recognition result with the vehicle state dynamic factor; Compare the absolute value of the difference between the confidence of the first terrain recognition result and the confidence of the second terrain recognition result with a preset threshold; If the absolute value of the difference is greater than the preset threshold, output the larger of the confidence of the first terrain recognition result and the confidence of the second terrain recognition result as the final terrain recognition result; Otherwise, output the terrain recognition result output after the previous confidence comparison as the current final terrain recognition result.

[0011] Furthermore, After arbitrating the first terrain recognition result and the second terrain recognition result to obtain the final terrain recognition result, it further includes: Render the 3D scene of the terrain in real time through the human-machine interface with the final terrain recognition result to remind the passengers in the vehicle of the terrain category ahead.

[0012] A vehicle terrain automatic recognition system includes: An acquisition module for acquiring terrain image data, chassis dynamic data, and vehicle state data; A first recognition module for inputting the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; A second recognition module for inputting the chassis dynamic data and vehicle state data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; An arbitration module for arbitrating the first terrain recognition result and the second terrain recognition result to obtain the final terrain recognition result.

[0013] Furthermore, The acquisition of terrain image data, chassis dynamic data, and vehicle state data includes: Acquire terrain image data; acquire chassis dynamic data, where the chassis dynamic data includes xyz three-axis acceleration information, yaw angular velocity information, wheel-side braking torque, wheel speed, steering wheel angle signal, wheel-side torque, vehicle speed, vehicle mass, wind resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio; acquire vehicle state data, where the vehicle state data includes wheel noise signal, environmental temperature signal, wading and water depth signal, ramp signal, and wiper state; The inputting of the terrain image data into a pre-trained terrain recognition model to obtain a first terrain recognition result includes: Input the topographic image data into the pre-trained topographic recognition model to obtain the first topographic recognition result; among them, the pre-trained topographic recognition model uses a depth graph convolutional neural network and is trained based on the data of the cloud vehicle network; Upload the first topographic recognition result to the artificial intelligence topographic database of the advanced driver assistance computing center through over-the-air (OTA) technology and the telematics box.

[0014] Furthermore, The step of inputting the chassis dynamic data and vehicle status data into the pre-trained second topographic recognition model to obtain the second topographic recognition result includes: Train the pre-constructed second topographic recognition model based on the historical chassis dynamic data and historical vehicle status data to obtain the pre-trained second topographic recognition model; the second topographic recognition model uses a convolutional neural network, a recurrent neural network, or a long short-term memory network; Clean, integrate, and extract the collected chassis dynamic data and vehicle status data to obtain the features to be recognized; Input the features to be recognized into the pre-trained second topographic recognition model to obtain the second topographic recognition result; The step of arbitrating the first topographic recognition result and the second topographic recognition result to obtain the final topographic recognition result includes: Calculate the confidence of the first topographic recognition result by combining the initial weight of the first topographic recognition result with the environmental dynamic factor, and calculate the confidence of the second topographic recognition result by combining the initial weight of the second topographic recognition result with the vehicle status dynamic factor; Compare the absolute value of the difference between the confidence of the first topographic recognition result and the confidence of the second topographic recognition result with a preset threshold; If the absolute value of the difference is greater than the preset threshold, output the one with the higher confidence among the first topographic recognition result and the second topographic recognition result as the final topographic recognition result; Otherwise, output the topographic recognition result output after the previous confidence comparison as the final topographic recognition result this time.

[0015] A vehicle, including a vehicle body; The vehicle body is integrated with a terrain automatic recognition control system, and the terrain automatic recognition control system is used to implement the steps of the above vehicle terrain automatic recognition method.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for automatically identifying vehicle terrain. First, terrain image data, chassis dynamic data, and vehicle state data are collected. The terrain image data is input into a pre-trained first terrain recognition model to obtain a first terrain recognition result. At the same time, a second terrain recognition model is used to recognize the chassis dynamic data and vehicle state data to obtain a second terrain recognition result. Through dual-path recognition and combined with an arbitration mechanism, the first terrain recognition result and the second terrain recognition result are arbitrated to obtain a final terrain recognition result, effectively reducing the dependence on a single sensor or data source. The terrain image data provides intuitive terrain features, while the chassis dynamic data and vehicle state data reflect the interaction between the vehicle and the terrain. The two complement each other to improve the accuracy of recognition. Finally, the arbitration mechanism ensures that a reliable terrain recognition result can be obtained even under complex meteorological conditions, thereby improving the accuracy and real-time performance of terrain recognition and ensuring the safety of vehicle driving. Preferably, in the present invention, the chassis dynamic data includes xyz three-axis acceleration information, yaw angular velocity information, wheel-end braking torque, wheel speed, steering wheel angle signal, wheel-end torque, vehicle speed, vehicle mass, wind resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio; the vehicle state data includes wheel noise signal, ambient temperature signal, wading and water depth signal, ramp signal, and wiper state. Such comprehensive data collection ensures that the terrain recognition model can obtain sufficient information, thereby improving the recognition accuracy. At the same time, the specific data items also provide a clear direction for subsequent data processing and model training.

[0017] Preferably, in the present invention, a depth graph convolutional neural network is used as the first terrain recognition model and trained based on the data of the cloud vehicle networking, which can utilize a large amount of actual terrain data to improve the recognition ability of the model. Through the over-the-air download technology and the upload of the recognition result by the telematics box, the real-time update and sharing of data are realized, which helps to build a more intelligent and efficient driving assistance system. Preferably, in the present invention, the second terrain recognition model adopts advanced neural network structures such as convolutional neural network, recurrent neural network, or long short-term memory network, which can process complex chassis dynamic data and vehicle state data. Through data cleaning, integration, and extraction processing, the quality of the input data is improved, thereby enhancing the accuracy of terrain recognition. Preferably, in the present invention, an arbitration mechanism combining confidence and dynamic factors is proposed, which can dynamically adjust the weights of the two terrain recognition results according to the environmental dynamic factor and the vehicle state dynamic factor, so as to more accurately select the final terrain recognition result. This mechanism takes into account the real-time changing environment and vehicle state, improving the adaptability and robustness of terrain recognition. Preferably, in the present invention, the final terrain recognition result is used to perform real-time display of the terrain 3D scene rendering through the human-machine interface, which not only provides intuitive terrain information but also enhances the perception and understanding of the terrain ahead by the vehicle occupants. This visualization method helps to improve the driving experience and safety of the driver, especially under complex or dangerous terrain conditions. The present invention also provides an automobile, including an automobile body, on which a terrain automatic recognition control system is integrated, and an innovative vehicle terrain recognition control method is adopted. The terrain automatic recognition control system collects terrain images, chassis dynamics, and vehicle status data, processes the image data using a pre-trained terrain recognition model to obtain a first recognition result, and at the same time analyzes the chassis and vehicle status data using a terrain recognition dynamic algorithm to obtain a second recognition result. Subsequently, the system arbitrates these two results to determine the final terrain recognition result. The terrain automatic recognition control system of this automobile reduces the dependence on a single sensor by integrating multiple data sources, improving the accuracy and robustness of recognition. Under complex meteorological conditions, such as heavy rain, thick fog, or low-light environments, this system can still maintain high recognition performance, ensuring that the vehicle can timely and accurately recognize the terrain, remind the driver to switch modes or automatically adjust control parameters, effectively avoiding safety accidents such as vehicle out of control and wading submergence, and improving the safety and reliability of vehicle driving. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow block diagram executed by a vehicle terrain automatic recognition system provided by an embodiment of the present invention; Figure 2 It is an overall vehicle layout and detection distance diagram of an automobile integrated with a terrain automatic recognition control system provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of sensor arrangement provided by an embodiment of the present invention; where (a) is a side view; (b) is a K-direction view of (a); Figure 4 It is a flowchart of a vehicle terrain automatic recognition method provided by an embodiment of the present invention; Figure 5 It is a schematic structural diagram of a vehicle terrain automatic recognition system provided by an embodiment of the present invention.

[0019] Reference Signs: 1, inertial measurement unit; 2, noise sensor; 3, temperature sensor; 4, water level sensor; 5, ramp sensor; 6, chassis domain control ECU; 7, lidar; 8, camera; 9, acquisition range; 10, vehicle front distance; 11, 3D perception information effective area. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] To further understand the content of the present invention, the following provides a detailed description of the present invention in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are merely for explaining the present invention rather than limiting it.

[0021] The following explains the technical terms related to the present invention: ADCC is fully called Advanced Driver Assistance Computing Center, an advanced driver assistance computing center, which is responsible for processing perception information and terrain recognition algorithms.

[0022] TSP is fully called Telematics Service Provider, a telematics service provider, which is responsible for data interaction and AI training in the cloud vehicle network.

[0023] OTA is fully called Over-The-Air, an over-the-air technology, which is used to remotely update vehicle software (such as the AI database).

[0024] TBOX is fully called Telematics BOX, a telematics box, which realizes communication between the vehicle and the cloud (such as OTA transmission).

[0025] Lidar is fully called Light Detection and Ranging, a lidar, which obtains three-dimensional point cloud data of the environment through laser beams.

[0026] CAN is fully called Controller Area Network, a controller area network, which is used for communication between internal electronic modules of the vehicle.

[0027] ETH is fully called Ethernet, an Ethernet, a high-speed data transmission protocol (such as the vehicle internal network).

[0028] ICC is fully called Intelligent Computing Center, an intelligent computing center, which is responsible for terrain arbitration algorithms and information integration.

[0029] ECU is fully called Electronic Control Unit, an electronic control unit (here it refers to the chassis domain control ECU, which processes chassis-related signals).

[0030] HMI is fully called Human-Machine Interface, a human-machine interface, which is used to display vehicle information and interaction (such as terrain 3D rendering).

[0031] CANFD stands for CAN with Flexible Data Rate, a CAN bus with flexible data rate that supports higher transmission rates and larger amounts of data.

[0032] As described in the background art, currently mass-produced vehicles on the market identify the terrain type of the road surface in front of the vehicle through lidar or cameras. Under normal conditions, the accuracy of identifying the road surface type by the vehicle is about 95%. In extreme weather (night, cloudy, foggy, heavy rain, waterlogging, etc.), the accuracy of identifying the road surface type is only 80% or cannot be identified (accuracy 0%). The low accuracy of vehicle terrain type identification may lead to vehicle safety accidents and cause property losses to vehicle owners.

[0033] To solve the above problems, this embodiment provides a method for automatic vehicle terrain identification. This method is a method that comprehensively collects and processes terrain images, chassis dynamics, and vehicle status data, and obtains the final terrain identification result through dual-path recognition and arbitration by models and algorithms. Using this method can solve the problem of low accuracy of terrain type identification on the current market. Moreover, it can ensure that the accuracy of identifying terrain road surface categories reaches 100% under normal climate conditions, and the accuracy of identifying terrain road surface categories in extreme weather (night, cloudy, foggy, heavy rain, waterlogging, etc.) reaches 95%, which can ensure the safety of passengers and avoid property losses to vehicle owners.

[0034] The following further explains the control method provided in this embodiment with reference to the accompanying drawings: This embodiment provides a method for automatic vehicle terrain identification, and the specific implementation steps are as follows: First step, deploy the perception information algorithm and the AI terrain database to the ADCC intelligent driving computing center controller.

[0035] Exemplarily, in this step, the deployment work of the perception information algorithm and the AI terrain database, that is, the pre-training process of the first terrain recognition model, is to train the pre-constructed depth graph convolutional neural network with the terrain image data in the historical AI terrain database, and output the trained terrain recognition model; the terrain recognition model is used to input the terrain image data collected in real time into the trained terrain recognition model and output the first terrain recognition result.

[0036] In a specific implementation manner, the AI terrain training algorithm of the first terrain recognition model can be deployed in the cloud vehicle networking TSP vehicle-cloud interconnection. The terrain category training is carried out by collecting images offline, and then the trained images are transmitted to the AI terrain database of the ADCC intelligent driving computing center through OTA and TBOX.

[0037] Step 2: With the help of a camera with night vision function and lidar, continuously collect the image and point cloud data of at least 25 meters in front of the vehicle to the ADCC intelligent driving computing center controller. After being analyzed and processed by the perception information algorithm and the AI terrain database in the ADCC intelligent driving computing center controller, the recognized road surface type (i.e., the first terrain recognition result) is sent to the ICC information computing center in the form of CAN or ETH signals.

[0038] Step 3: Deploy the terrain recognition AI algorithm on the chassis domain control ECU.

[0039] Exemplarily, in this step, the deployment process of the terrain recognition AI algorithm, that is, the pre-training process of the first terrain recognition model, is used to dynamically recognize the current terrain according to the chassis dynamic data and vehicle state data. The specific steps are as follows: Train the pre-constructed second terrain recognition model based on the historical chassis dynamic data and historical vehicle state data to obtain the pre-trained second terrain recognition model; the second terrain recognition model uses a convolutional neural network, a recurrent neural network, or a long short-term memory network.

[0040] In a specific embodiment, the collected chassis dynamic data and vehicle state data are cleaned, integrated, and extracted to obtain the features to be recognized; the features to be recognized are input into the pre-trained second terrain recognition model to obtain the second terrain recognition result.

[0041] The more specific recognition steps are as follows: Step 1: Complete data collection through the multi-source data collection module: Collect wheel noise signals, ambient temperature signals, wading and water depth signals, ramp signals, output wiper status, xyz three-axis acceleration information, yaw angular velocity information, wheel-side braking torque, wheel speed, steering wheel angle signal, wheel-side torque, vehicle speed, vehicle mass, wind resistance coefficient, transmission efficiency, wheelbase, track width, steering ratio and other signals through vehicle sensors and control units (noise sensors, temperature sensors, water level sensors, ramp sensors, rain brake controllers, 6D-IMU (inertial measurement unit, integrated in the chassis domain control ECU), IPB / onebox (brake controller), EPS (steering controller), VCU (vehicle control unit), FZCU (front area controller), etc.).

[0042] Step 2: Data Processing: Clean, integrate, and analyze the collected data to extract relevant features to be recognized. The features to be recognized include: the wheel slip rate and IMU vibration spectrum to be recorded for the sandy road surface type, the wheel speed fluctuation and water level sensor data to be collected for the muddy road surface type, the high-frequency vibration and steering torque changes to be captured for the rock road surface type, the relationship between temperature and braking distance to be recorded for the snow road surface type, the correlation between water level depth and power output to be analyzed for the wading road surface type, etc.

[0043] Step 3: Feature Recognition: Input the features to be recognized into the pre-trained second terrain recognition model to obtain the second terrain recognition result, which is the current road surface type of the vehicle (such as sandy, muddy, rocky, snowy, wading, etc. road surface types).

[0044] Among them, the vehicle status data includes wheel noise, ambient temperature, wading and water depth, slope, wiper status signal, and the corresponding sensors send the above signals to the chassis domain control ECU respectively.

[0045] The chassis dynamic data includes xyz three-axis acceleration information, yaw angular velocity information, wheel-side braking torque, wheel speed, steering wheel angle signal, wheel-side torque, vehicle speed, vehicle mass, air resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio; the above chassis dynamic data is also sent to the chassis domain control ECU through CAN or CANFD signals.

[0046] Fourth step: The chassis domain control ECU identifies the second terrain recognition result based on the chassis dynamic data and vehicle status data through the terrain recognition AI algorithm (the second terrain recognition model); it should be noted that both the first terrain recognition result and the second terrain recognition result represent the road surface terrain categories, including snow, mud, wading road surface, and sand; both the first terrain recognition result and the second terrain recognition result are sent to the ICC information calculation center through CAN or CANFD signals.

[0047] Fifth step: Through the terrain arbitration algorithm deployed in the ICC information calculation center, arbitrate the terrain types output by the ADCC intelligent driving calculation center and the chassis domain control ECU, that is, complete the arbitration of the first terrain recognition result and the second terrain recognition result, and the specific steps for outputting the final terrain recognition result are as follows: Calculate the confidence of the first terrain recognition result by combining the initial weight of the first terrain recognition result and the environmental dynamic factor, and calculate the confidence of the second terrain recognition result by combining the initial weight of the second terrain recognition result and the vehicle status dynamic factor; Compare the absolute value of the difference between the confidence of the first terrain recognition result and the confidence of the second terrain recognition result with a preset threshold; If the absolute value of the difference is greater than the preset threshold, output the terrain recognition result with the higher confidence between the first terrain recognition result and the second terrain recognition result as the final terrain recognition result; Otherwise, output the terrain recognition result output after the previous confidence comparison as the final terrain recognition result this time.

[0048] In a specific embodiment, the arbitration steps are more specific as follows: Step 1: Initialize the basic weights of the first terrain recognition result and the second terrain recognition result: The basic weight value is defined as: The basic weight of the road surface type recognized by the chassis AI that fuses the three-modal vibration spectrum, acoustic features, and wheel speed dynamics is 0.4; that is, the basic weight of the first terrain recognition result; The basic weight value of the terrain type judged by the perception information (image + point cloud data) is 0.6; that is, the basic weight of the second terrain recognition result.

[0049] The dynamic factor is defined as: The dynamic factor range of the road surface type recognized by the chassis AI that fuses the three-modal vibration spectrum, acoustic features, and wheel speed dynamics is 0.5 - 1.3, that is, the vehicle state dynamic factor; The dynamic factor range of the terrain type judged by the perception information (image + point cloud data) is 0.5 - 1.5, that is, the environmental dynamic factor.

[0050] Step 2: Arbitration decision-making process: Calculate the weighted confidence (basic weight × dynamic factor) through weighted fusion, and the decision-making logic: Select the terrain with the highest weighted confidence as the main candidate result and output the terrain road surface type after arbitration.

[0051] In this step, if the absolute value of the difference is greater than the preset threshold, output the terrain recognition result with the higher confidence between the first terrain recognition result and the second terrain recognition result as the final terrain recognition result; Otherwise, output the terrain recognition result output after the previous confidence comparison as the final terrain recognition result this time.

[0052] When the absolute value of the difference is less than or equal to the preset threshold, that is, when the arbitration result cannot reach a clear conclusion: Keep the current mode and increase the system vigilance, send a request confirmation signal to the driver, limit the maximum vehicle speed to the safe range, and record the fault code for subsequent diagnosis.

[0053] Exemplarily, the environmental dynamic factor is dynamically adjusted according to light, weather, and obstacle complexity; the vehicle state dynamic factor is adjusted according to the sensor signal quality and vehicle state.

[0054] Exemplarily, the local terrain recognition method regularly updates the terrain recognition algorithm, adjusts the weights according to the system learning situation, and optimizes the arbitration logic through OTA updates.

[0055] Step 6: Output the final terrain recognition result to the HMI human-machine interface and the chassis domain control ECU.

[0056] Exemplarily, in this step, after receiving the final terrain recognition result, the HMI human-machine interface displays the 3D terrain rendering scene in real time to remind the vehicle occupants (driver or passengers) of the terrain category ahead.

[0057] In a specific embodiment, the final terrain recognition result can also be transmitted to the vehicle owner through the in-vehicle system. The vehicle owner can understand the terrain environment where the current vehicle is located in real time and inform the vehicle occupants of the terrain environment where the current vehicle is located by communicating with the vehicle occupants. This situation is usually applied to the scenario where the vehicle owner is not in the vehicle, providing a multiple safeguard mechanism.

[0058] Finally, after receiving the arbitration result, that is, after receiving the final terrain recognition result, the chassis domain control ECU performs vehicle all-terrain control. In this embodiment, the specific implementation process is as follows: The chassis domain control ECU performs power system control: Torque distribution: Adjust the engine output torque to each wheel according to different terrains. For example, when off-road climbing, more torque is distributed to the rear wheels to increase the climbing ability; when driving on a slippery road surface, the torque output is reduced to prevent wheel slippage.

[0059] Shift strategy: Change the shift timing of the automatic transmission according to the terrain. When driving on a rough mountain road, delay shifting to keep the engine at a higher speed to provide greater power; when driving on a flat highway, shift gears in a timely manner to reduce the engine speed and improve fuel economy.

[0060] The chassis domain control ECU performs suspension system control: Height adjustment: When encountering a terrain with large potholes or high bumps, raise the vehicle suspension to increase the ground clearance and prevent the chassis from scraping; when driving at high speed, lower the suspension height to lower the vehicle center of gravity and improve driving stability and aerodynamic performance.

[0061] Damping adjustment: On off-road surfaces, increase the suspension damping to reduce body swaying and bouncing, improve vehicle passability and ride comfort; when driving on a flat road surface, reduce the damping to make the suspension softer and improve ride comfort.

[0062] The chassis domain control ECU performs steering system control: Steering assist adjustment: When off-roading or driving at low speeds, increase the steering assist to make steering easier for the driver to control; when driving at high speeds, reduce the steering assist to increase the stability of the steering feel and improve driving safety.

[0063] Steering angle limit: Limit the maximum steering angle of the vehicle according to different terrains and driving speeds. For example, on narrow mountain roads, reduce the maximum steering angle to prevent the vehicle from getting out of control; in open areas or when making a U-turn at low speeds, allow a larger steering angle to improve the vehicle's maneuverability.

[0064] The chassis domain control ECU controls the braking system: Braking force distribution: Reasonably distribute the braking force of each wheel under different terrains. For example, on icy or snowy roads, reduce the braking force on the rear wheels to prevent the vehicle from skidding; when climbing a slope, appropriately increase the braking force on the front wheels to prevent the vehicle from tipping forward.

[0065] Braking assist function: Enable corresponding braking assist functions according to different terrains. For example, when descending a steep slope, enable the hill descent control system to automatically control the vehicle's downhill speed; when driving on a slippery road surface, enable the anti-lock braking system (ABS) and electronic stability program (ESP) to prevent wheel lock-up and vehicle skidding.

[0066] The chassis domain control ECU controls the tire pressure: Tire pressure adjustment: Adjust the tire pressure in real time according to the terrain and driving conditions. When off-roading, appropriately reduce the tire pressure to increase the contact area between the tire and the ground, improving traction and passability; when driving at high speeds, ensure that the tire pressure is normal to reduce rolling resistance, improve fuel economy and driving safety.

[0067] Tire pressure monitoring: Continuously monitor the tire pressure and temperature. Once an abnormality is detected, promptly alert the driver so that appropriate measures can be taken to avoid safety accidents caused by tire problems.

[0068] This embodiment also provides a vehicle terrain automatic recognition system for implementing the steps of the above vehicle terrain automatic recognition method. Using this system can ensure that the accuracy of identifying terrain road types reaches 100% under normal climate conditions, and the accuracy of identifying terrain road types in extreme weather (night, cloudy, foggy, heavy rain, waterlogging, etc.) reaches 95%. The industrialization level is very high, ensuring the safety of passengers and avoiding property losses of vehicle owners; the specific structure and execution process are as Figure 1 shown, including: Step 1: The vehicle driving speed ≥ a km / h (1 ≤ a ≤ 100 can be calibrated), and the road surface driving time ≥ b ms (100 ≤ b ≤ 500 can be calibrated).

[0069] Driving under normal climate conditions or extreme weather (night, cloudy day, fog, heavy rain, waterlogging, etc.).

[0070] The perception information algorithm and the AI terrain database are deployed in the ADCC intelligent driving computing center controller.

[0071] The AI terrain training algorithm is deployed in the cloud vehicle networking TSP vehicle-cloud interconnection. Images are collected offline, and continuous terrain category training is carried out. The trained images are sent to the AI terrain database in the ADCC intelligent driving computing center through the OTA and TBOX remote / on-vehicle communication modules.

[0072] The camera with night vision function preferably uses a binocular or trinocular camera to continuously collect image data c meters (c≥25) in front of the vehicle. The lidar continuously collects point cloud data c meters (c≥25) in front of the vehicle to the ADCC intelligent driving computing center controller. The ADCC intelligent driving computing center controller analyzes and processes the perception information through the perception information algorithm and the AI terrain database, and sends the recognized road surface types to the ICC information computing center in the form of CAN signals or ETH Ethernet signals.

[0073] Step 2: The terrain recognition AI algorithm is deployed in the chassis domain control ECU.

[0074] The noise sensor collects the wheel noise signal, the temperature sensor collects the ambient temperature signal, the water level sensor collects the wading and water depth signals, the ramp sensor collects the ramp signal, and the rain brake controller collects the wiper status to the chassis domain control ECU.

[0075] The 6D-IMU (inertial measurement unit, integrated in the chassis domain control ECU) collects the acceleration information of the xyz three axes and the yaw angular velocity information. The IPB / onebox (brake controller) collects the wheel-side braking torque and the wheel speed. The EPS (steering controller) collects the steering wheel angle signal. The VCU (vehicle control unit) collects the wheel-side torque and the vehicle speed.

[0076] The FZCU (front area controller) collects signals such as vehicle mass, air resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio. The signals output by the above controllers are sent to the chassis domain control ECU through CAN or CANFD signals.

[0077] Based on the vehicle chassis signals, the chassis domain control ECU intelligently recognizes the current road surface terrain category through the terrain recognition AI algorithm deployed in the chassis domain control ECU, and sends it to the ICC information computing center through CAN or CANFD signals.

[0078] Step 3: The terrain arbitration algorithm is deployed in the ICC information computing center.

[0079] In the ICC information calculation center, through the terrain arbitration algorithm, the terrain types output by the ADCC intelligent driving calculation center controller and the terrain types output by the chassis domain control ECU are arbitrated, and the arbitrated road surface type is output to the HMI human-machine interface and the chassis domain control ECU.

[0080] Step 4: The HMI human-machine interface receives the arbitrated road surface type, and the HMI human-machine interface displays the real-time terrain 3D scene rendering to remind the vehicle owner of the terrain category ahead.

[0081] After receiving the arbitrated road surface type, the chassis domain control ECU performs full-terrain control of the vehicle.

[0082] As Figure 2 shown, this embodiment also provides a vehicle, including a vehicle body integrated with a terrain automatic recognition control system. The terrain automatic recognition control system can implement the steps of the above vehicle terrain automatic recognition method or can execute the functions of the vehicle terrain automatic recognition system. The specific sensor layout and detection distance are designed as follows: This system includes a lidar 7, with a quantity of 1, arranged on the roof near the front windshield. It can continuously collect the point cloud data in the area 9 with a range of c meters (c≥25), that is, a distance of 10 in front of the vehicle.

[0083] In this embodiment, a camera 8 with night vision function (preferably a binocular or trinocular camera with a built-in computing module) is used, with a quantity of 1, arranged at the front of the vehicle. It can detect the image data at a distance of c meters (c≥25) in front of the vehicle, that is, a distance of 10 in front of the vehicle, and uses the perspective overlap area of two co-focal cameras to achieve stereo parallax, thereby obtaining the effective area 11 of 3D perception information in the perspective overlap area.

[0084] The chassis domain control ECU 6 involved in the present invention integrates a 6D-IMU inertial measurement unit 1, and the chassis domain control ECU 6 is arranged on the body floor of the vehicle body.

[0085] As Figure 3 shown, specifically as shown in (a) and (b) in Figure 3 , the sensor installation positions involved in this embodiment are as follows: The noise sensor 2 involved in the present invention, with 1≤quantity≤4, is arranged near the tire position. The wheel-side noise is collected through the noise sensor 2 and the wheel speed, and output to the chassis domain control ECU 6 for input into the terrain type recognition algorithm.

[0086] The temperature sensor 3 involved in the present invention, with 1≤quantity≤2, is arranged on the lower vehicle body and away from the heat source position, and outputs the ambient temperature signal to the chassis domain control ECU 6 for input into the terrain type recognition algorithm.

[0087] The water level sensors 4 involved in the present invention, with the quantity satisfying 1 ≤ quantity ≤ 4, are arranged on the lower vehicle body and at positions far from heat sources, and output wading and water depth signals to the chassis domain control ECU 6 for input into the terrain type recognition algorithm.

[0088] The ramp sensors 5 involved in the present invention, with the quantity satisfying 1 ≤ quantity ≤ 2, are arranged on the frame of the lower vehicle body and at positions far from heat sources, and output ramp signals to the chassis domain control ECU 6 for input into the terrain type recognition algorithm.

[0089] Exemplarily, as Figure 4 shown, this embodiment also provides a method for automatically identifying vehicle terrain, including the following steps: Collect terrain image data, chassis dynamic data, and vehicle status data; Input the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; Input the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; Arbitrate between the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result.

[0090] In this embodiment, the collecting of terrain image data, chassis dynamic data, and vehicle status data includes: Collect terrain image data; collect chassis dynamic data, where the chassis dynamic data includes xyz three-axis acceleration information, yaw angular velocity information, wheel-end braking torque, wheel speed, steering wheel angle signal, wheel-end torque, vehicle speed, vehicle mass, air resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio; collect vehicle status data, where the vehicle status data includes wheel noise signal, ambient temperature signal, wading and water depth signal, ramp signal, and wiper status.

[0091] In this embodiment, the inputting of the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result includes: Input the terrain image data into a pre-trained terrain recognition model to obtain a first terrain recognition result; wherein, the pre-trained terrain recognition model adopts a depth graph convolutional neural network and is trained based on data from the cloud vehicle networking. Upload the first terrain recognition result to the artificial intelligence terrain database of the advanced driver assistance computing center through over-the-air technology and the telematics box.

[0092] In this embodiment, the inputting of the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result includes: Train the pre - constructed second terrain recognition model based on historical chassis dynamic data and historical vehicle state data to obtain a pre - trained second terrain recognition model; the second terrain recognition model uses a convolutional neural network, a recurrent neural network, or a long short - term memory network; Clean, integrate, and extract the collected chassis dynamic data and vehicle state data to obtain features to be recognized; Input the features to be recognized into the pre - trained second terrain recognition model to obtain a second terrain recognition result.

[0093] In this embodiment, the arbitration of the first terrain recognition result and the second terrain recognition result to obtain the final terrain recognition result includes: Calculate the confidence of the first terrain recognition result by combining the initial weight of the first terrain recognition result and the environmental dynamic factor, and calculate the confidence of the second terrain recognition result by combining the initial weight of the second terrain recognition result and the vehicle state dynamic factor; Compare the absolute value of the difference between the confidence of the first terrain recognition result and the confidence of the second terrain recognition result with a preset threshold; If the absolute value of the difference is greater than the preset threshold, output the one with the larger confidence among the first terrain recognition result and the second terrain recognition result as the final terrain recognition result; Otherwise, output the terrain recognition result output after the previous confidence comparison as the current final terrain recognition result.

[0094] In this embodiment, after the arbitration of the first terrain recognition result and the second terrain recognition result to obtain the final terrain recognition result, it further includes: Render the 3D terrain scene of the final terrain recognition result in real - time through the human - machine interface to remind the passengers in the vehicle of the terrain category ahead.

[0095] As Figure 5 shown, this embodiment also provides a vehicle terrain automatic recognition system, including: a collection module for collecting terrain image data, chassis dynamic data, and vehicle state data; a first recognition module for inputting the terrain image data into a pre - trained first terrain recognition model to obtain a first terrain recognition result; a second recognition module for inputting the chassis dynamic data and vehicle state data into a pre - trained second terrain recognition model to obtain a second terrain recognition result; an arbitration module for arbitrating the first terrain recognition result and the second terrain recognition result to obtain the final terrain recognition result.

[0096] The present invention also provides an electronic device, including: a memory for storing a computer program; a processor for implementing the steps of the vehicle terrain automatic recognition method when executing the computer program.

[0097] When the processor executes the computer program, it implements the steps of the above-mentioned automatic vehicle terrain recognition control, for example: collecting terrain image data, chassis dynamic data, and vehicle status data; inputting the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; inputting the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; arbitrating the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result.

[0098] Alternatively, when the processor executes the computer program, it implements the functions of each module in the above system, for example: a collection module for collecting terrain image data, chassis dynamic data, and vehicle status data; a first recognition module for inputting the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; a second recognition module for inputting the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; an arbitration module for arbitrating the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result.

[0099] Exemplarily, the computer program can be divided into one or more modules / units, and the one or more modules / units are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing preset functions, and the instruction segments are used to describe the execution process of the computer program in the automatic vehicle terrain recognition control device. For example, the computer program can be divided into a collection module, a first recognition module, a second recognition module, and an arbitration module; the specific functions of each module are as follows: the collection module is used to collect terrain image data, chassis dynamic data, and vehicle status data; the first recognition module is used to input the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; the second recognition module is used to input the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; the arbitration module is used to arbitrate the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result.

[0100] The vehicle terrain automatic recognition and control device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The vehicle terrain automatic recognition and control device may include, but is not limited to, a processor and a memory. Those skilled in the art can understand that the above are examples of the vehicle terrain automatic recognition and control device, which do not constitute a limitation on the vehicle terrain automatic recognition and control device. It may include more components than the above, or combine some components, or different components. For example, the vehicle terrain automatic recognition and control device may also include input / output devices, network access devices, buses, etc.

[0101] The so-called processor may be a central processing unit (CPU), or may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The processor is the control center of the vehicle terrain automatic recognition and control, and uses various interfaces and lines to connect all parts of the vehicle terrain automatic recognition and control device.

[0102] The memory can be used to store the computer programs and / or modules. The processor realizes various functions of the vehicle terrain automatic recognition and control device by running or executing the computer programs and / or modules stored in the memory, and by calling the data stored in the memory.

[0103] The memory may mainly include a program storage area and a data storage area. Among them, the program storage area may store an operating system, application programs required for at least one function (such as a sound playback function, an image playback function, etc.); the data storage area may store data created according to the use of the mobile phone (such as audio data, phone book, etc.). In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, at least one magnetic disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0104] The present invention also provides a computer-readable storage medium storing a computer program, which when executed by a processor, implements the steps of the vehicle terrain automatic recognition method described above.

[0105] If the modules / units integrated in the vehicle terrain automatic recognition system are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0106] Based on such understanding, all or part of the processes in the vehicle terrain automatic recognition method of the present invention can also be completed by instructing relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, the steps of the vehicle terrain automatic recognition method described above can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, executable file or preset intermediate form, etc.

[0107] The computer-readable storage medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disc, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0108] It should be noted that the content included in the computer-readable storage medium can be appropriately increased or decreased according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable storage medium does not include electrical carrier signals and telecommunication signals.

[0109] The above embodiments are only one of the implementation manners capable of implementing the technical solution of the present invention. The scope of protection required by the present invention is not only limited by this embodiment, but also includes any changes, substitutions and other implementation manners that are easily conceivable by those skilled in the art within the technical scope disclosed by the present invention.

[0110] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: the specific implementation manners of the present invention can still be modified or equivalently replaced, and any modification or equivalent replacement without departing from the spirit and scope of the present invention should be covered by the protection scope of the claims of the present invention.

Claims

1. An automatic vehicle terrain recognition method, characterized in that, Including: Collecting terrain image data, chassis dynamic data, and vehicle status data; Inputting the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; Inputting the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; Arbitrating between the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result.

2. The vehicle terrain automatic recognition method according to claim 1, wherein the collecting of the terrain image data, chassis dynamic data, and vehicle status data includes: Collecting terrain image data; collecting chassis dynamic data, where the chassis dynamic data includes xyz three-axis acceleration information, yaw angular velocity information, wheel-end braking torque, wheel speed, steering wheel angle signal, wheel-end torque, vehicle speed, vehicle mass, wind resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio; collecting vehicle status data, where the vehicle status data includes wheel noise signal, ambient temperature signal, wading and water depth signal, ramp signal, and wiper status.

3. The vehicle terrain automatic recognition method according to claim 1, wherein the inputting of the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result includes: Inputting the terrain image data into a pre-trained terrain recognition model to obtain a first terrain recognition result; wherein, the pre-trained terrain recognition model uses a depth graph convolutional neural network and is trained based on data from a cloud vehicle network; Uploading the first terrain recognition result to the artificial intelligence terrain database of the advanced driver assistance computing center through over-the-air (OTA) technology and a telematics box.

4. The vehicle terrain automatic recognition method according to claim 1, wherein the inputting of the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result includes: Training a pre-constructed second terrain recognition model based on historical chassis dynamic data and historical vehicle status data to obtain a pre-trained second terrain recognition model; the second terrain recognition model uses a convolutional neural network, a recurrent neural network, or a long short-term memory network; Cleaning, integrating, and extracting the collected chassis dynamic data and vehicle status data to obtain features to be recognized; Inputting the features to be recognized into the pre-trained second terrain recognition model to obtain a second terrain recognition result.

5. The vehicle terrain automatic recognition method according to claim 1, wherein the arbitrating between the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result includes: Calculating the confidence of the first terrain recognition result by combining the initial weight of the first terrain recognition result and the environmental dynamic factor, and calculating the confidence of the second terrain recognition result by combining the initial weight of the second terrain recognition result and the vehicle status dynamic factor; Comparing the absolute value of the difference between the confidence of the first terrain recognition result and the confidence of the second terrain recognition result with a preset threshold; If the absolute value of the difference is greater than the preset threshold, output the terrain recognition result with the higher confidence between the first terrain recognition result and the second terrain recognition result as the final terrain recognition result; Otherwise, output the terrain recognition result output after the previous confidence comparison as the final terrain recognition result this time.

6. The vehicle terrain automatic recognition method according to claim 1, characterized in that After arbitrating the first terrain recognition result and the second terrain recognition result to obtain the final terrain recognition result, it further includes: Render the 3D terrain scene of the final terrain recognition result in real time through the human-machine interface to remind the vehicle occupants of the terrain category ahead.

7. An automatic vehicle terrain recognition system, characterized in that, It includes: A collection module for collecting terrain image data, chassis dynamic data, and vehicle status data; A first recognition module for inputting the terrain image data into a pre-trained first terrain recognition model to obtain a first terrain recognition result; A second recognition module for inputting the chassis dynamic data and vehicle status data into a pre-trained second terrain recognition model to obtain a second terrain recognition result; An arbitration module for arbitrating the first terrain recognition result and the second terrain recognition result to obtain the final terrain recognition result.

8. The vehicle terrain automatic recognition system according to claim 7, characterized in that The collection of terrain image data, chassis dynamic data, and vehicle status data includes: Collect terrain image data; collect chassis dynamic data, where the chassis dynamic data includes xyz three-axis acceleration information, yaw angular velocity information, wheel-side braking torque, wheel speed, steering wheel angle signal, wheel-side torque, vehicle speed, vehicle mass, wind resistance coefficient, transmission efficiency, wheelbase, track width, and steering ratio; collect vehicle status data, where the vehicle status data includes wheel noise signal, ambient temperature signal, wading and water depth signal, ramp signal, and wiper status; The input of the terrain image data into the pre-trained terrain recognition model to obtain the first terrain recognition result includes: Input the terrain image data into the pre-trained terrain recognition model to obtain the first terrain recognition result; among them, the pre-trained terrain recognition model uses a depth graph convolutional neural network and is trained based on the data of the cloud vehicle network; Upload the first terrain recognition result to the artificial intelligence terrain database of the advanced driver assistance computing center through over-the-air technology and the telematics box.

9. The vehicle terrain automatic recognition system according to claim 7, characterized in that The input of the chassis dynamic data and vehicle status data into the pre-trained second terrain recognition model to obtain the second terrain recognition result includes: Train the pre-constructed second terrain recognition model based on historical chassis dynamic data and historical vehicle status data to obtain the pre-trained second terrain recognition model; the second terrain recognition model uses a convolutional neural network, a recurrent neural network, or a long short-term memory network; Clean, integrate, and extract the collected chassis dynamic data and vehicle status data to obtain the features to be recognized; Input the features to be recognized into the pre-trained second terrain recognition model to obtain the second terrain recognition result; Arbitrating the first terrain recognition result and the second terrain recognition result to obtain a final terrain recognition result includes: Calculating the confidence of the first terrain recognition result by combining the initial weight of the first terrain recognition result with the environmental dynamic factor, and calculating the confidence of the second terrain recognition result by combining the initial weight of the second terrain recognition result with the vehicle state dynamic factor; Comparing the absolute value of the difference between the confidence of the first terrain recognition result and the confidence of the second terrain recognition result with a preset threshold; If the absolute value of the difference is greater than the preset threshold, outputting the one with the greater confidence among the first terrain recognition result and the second terrain recognition result as the final terrain recognition result; Otherwise, outputting the terrain recognition result output after the previous confidence comparison as the current final terrain recognition result.

10. A vehicle, characterized in that, Including an automobile body; The automobile body is integrated with a terrain automatic recognition control system, and the terrain automatic recognition control system is used to implement the steps of the vehicle terrain automatic recognition method according to any one of claims 1-6.

Images