All-terrain pavement type identification method and system

Abstract

The invention discloses an all-terrain pavement type identification method and system. The method comprises the following steps: identifying a pixel type label of each pixel point in a road image; determining a cell label of each cell in the road scene distribution grid chart based on the pixel category label; dividing road ROI regions of various preset road types in the road scene distribution grid chart based on the preset road types, and determining region tags of the road ROI regions based on cell tags of cells in the road ROI regions; on the basis of the region labels, accumulating the values of the effective counts of the various preset road types; and determining the preset road type of which the value of the effective count is greater than or equal to the effective threshold value as the current road type based on the preset priority. According to the technical scheme of the invention, different road ROI regions are obtained through division according to different preset road types, and then the final road type is determined based on accumulation of effective counts, so that the accuracy of all-terrain road surface type recognition is effectively improved.

Description

Technical Field

[0001] The present application relates to the technical field of vehicle assisted driving, and in particular to an all-terrain road surface type recognition method and an all-terrain road surface type recognition system. Background Art

[0002] The domestic automobile intelligence is developing rapidly. After active safety and cockpit intelligence have developed to a certain stage, the ride comfort experience has become another focus of vehicle differentiation. Through binocular vision sensors to detect the road type ahead of the vehicle, combined with active suspension chassis, intelligent adaptive suspension and other functions can be realized, effectively improving the vehicle driving experience and reducing the control requirements for off-road sections.

[0003] However, such existing technologies are still immature, which is mainly reflected in two aspects. First, it is impossible to accurately identify the road type under complex road conditions, especially mixed road scenes such as ice, snow, water, mud, gravel, etc.; second, there is misidentification of road types in some scenarios. For example, when a vehicle turns and identifies grass on the side of the road, the road type ahead may be modified to "grass". Summary of the invention

[0004] The purpose of this application is to accurately identify the road type ahead of a vehicle.

[0005] The technical solution of the first aspect of the present application is: a method for identifying all-terrain road types is provided, the method comprising: step 1: identifying the pixel category label of each pixel point in the road image; step 2: determining the cell label of each cell in the road scene distribution grid map based on the pixel category label; step 3: dividing the road ROI area of ​​each preset road type in the road scene distribution grid map based on the preset road type, and determining the regional label of the road ROI area based on the cell label of the cell in the road ROI area; step 4: based on the regional label, respectively accumulating the values ​​of the valid counts of each preset road type; step 5: based on the preset priority, determining the preset road type whose valid count value is greater than or equal to the valid threshold as the current road type.

[0006] In some embodiments, the method further includes: step 6: based on the area label, respectively accumulating the values ​​of the invalid counts of each type of preset road type; step 7: when it is determined that the value of the invalid count is greater than or equal to the invalid threshold, setting the value of the valid count of the preset road type corresponding to the invalid count to 0.

[0007] In some embodiments, step 6 specifically includes: when it is determined that the region labels of any road ROI region corresponding to the preset road type do not belong to the preset road type, adding one to the invalid count value of the preset road type.

[0008] In some embodiments, step 4 specifically includes: counting the number of road ROI areas whose area labels belong to the preset road type in any preset road type, recorded as the accumulated number; and accumulating the values ​​of the valid counts of the preset road type according to the accumulated number.

[0009] In some embodiments, a binocular camera is provided on the vehicle, and the binocular camera is used to obtain an image in front of the vehicle. Step 2 specifically includes: based on a preset unit length and with the vehicle as the center, dividing a road scene distribution grid map on the road in front of the vehicle; based on the road image and the disparity map corresponding to the road image, performing a disparity operation on the pixel points in the road image to transform the coordinates and determine the cell corresponding to the pixel point in the road scene distribution grid map; counting the number of pixels of each type of pixel category label in the cell, and recording the pixel category label with the largest number as the cell label of the cell.

[0010] In some embodiments, before step 3, the method also includes: determining a coordinate transformation matrix based on the driving information of the vehicle on the road; aligning the coordinates of each cell in the road scene distribution grid map at the previous moment with each cell in the road scene distribution grid map at the current moment based on the coordinate transformation matrix; weighting the pixel category label ratio of each cell at the previous moment to the pixel category label ratio of each cell at the current moment in a weighted manner according to the aligned cells; and updating the cell label of each cell in the road scene distribution grid map at the current moment according to the weighted pixel category label ratio of each cell at the current moment.

[0011] The technical solution of the second aspect of the present application is: a system for identifying all-terrain road types is provided, the system comprising: an identification unit, the identification unit being configured to identify the pixel category label of each pixel point in the road image; a first label determination unit, the first label determination unit being configured to determine the cell label of each cell in the road scene distribution grid map based on the pixel category label; a second label determination unit, the second label determination unit being configured to divide the road ROI area of ​​each preset road type in the road scene distribution grid map based on the preset road type, and determine the regional label of the road ROI area based on the cell labels of the cells in the road ROI area; an accumulation unit, the accumulation unit being configured to accumulate the values ​​of the valid counts of each preset road type based on the regional label; and a decision unit, the decision unit being configured to determine the preset road type with a valid count value greater than or equal to the valid threshold as the current road type based on the preset priority.

[0012] In some embodiments, the system is installed on a vehicle, and a state sensor is provided on the vehicle. The state sensor is used to detect whether the vehicle is in a driving pending state, wherein the driving pending state includes a stationary state, a turning state, a crab state, and a parking state. The decision unit is also configured to keep the road type determined at the last moment unchanged when it is determined that the vehicle is in the driving pending state.

[0013] In some embodiments, the decision unit is further configured to keep the road type determined at the last moment unchanged when it is determined that the rotation speed of the vehicle's wiper is greater than or equal to a preset rotation speed.

[0014] In some embodiments, the vehicle is also provided with a temperature sensor for detecting the ambient temperature outside the vehicle. The decision unit is also configured to keep the road type determined at the last moment unchanged when it is determined that the ambient temperature is greater than a preset temperature and the current road type is determined to be an ice and snow type.

[0015] The beneficial effects of this application are:

[0016] The technical solution in the present application sets preset road types with different priorities in combination with the actual situation during vehicle driving, and divides different road ROI areas for each preset road type in the same road scene distribution grid map, and then determines the regional label of the road ROI area based on the identified pixel category label, and finally determines the preset road type with a higher priority and a valid count reaching a threshold as the current road type by accumulating effective counts, which can accurately identify the road type ahead of the vehicle and improve the driving experience of the vehicle while ensuring driving safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The advantages of the above and / or additional aspects of the present application will become apparent and easily understood in the description of the embodiments in conjunction with the following drawings, in which:

[0018] Figure 1 is a schematic flow chart of an all-terrain road type identification method according to an embodiment of the present application;

[0019] Figure 2 is a schematic diagram of a road scene distribution grid map according to an embodiment of the present application;

[0020] Figure 3 is a schematic diagram of road ROI area division according to an embodiment of the present application;

[0021] Figure 4 is a schematic block diagram of an all-terrain road type recognition system according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the present application is further described in detail below in conjunction with the accompanying drawings and specific implementation methods. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.

[0023] In the following description, many specific details are elaborated to facilitate a full understanding of the present application. However, the present application may also be implemented in other ways different from those described herein. Therefore, the protection scope of the present application is not limited to the specific embodiments disclosed below.

[0024] Embodiment 1:

[0025] like Figure 1 As shown, this embodiment provides a method for identifying all-terrain road types, the method comprising:

[0026] Step 1: Identify the pixel category label of each pixel in the road image;

[0027] Step 2: Based on the pixel category label, determine the cell label of each cell in the road scene distribution grid map;

[0028] Step 3: Based on the preset road types, the road ROI areas of various preset road types are divided in the road scene distribution grid map, and the area labels of the road ROI areas are determined based on the cell labels of each cell contained in the road ROI areas;

[0029] Step 4: Based on the regional labels, the valid count values ​​of each preset road type are accumulated respectively;

[0030] Step 5: Based on the preset priority, a preset road type whose valid count value is greater than or equal to the valid threshold is determined as the current road type.

[0031] Specifically, the image acquisition device in this embodiment can be a binocular camera or an RGBD camera. The image acquisition device is installed in front of the vehicle and calibrated so as to obtain an image in front of the vehicle, which includes road information in front of the vehicle.

[0032] Taking the RGBD camera as an example, the RGB image collected by the RGBD camera will be input into the AI ​​recognition model. Based on the image recognition method, each pixel in the input RGB image is identified and a pixel category label with a higher confidence level is output. The pixel category label is used to mark the road surface type to which the pixel belongs.

[0033] The pixels in the RGB image can be marked by assigning numerical values ​​as pixel category labels. The value of the pixel category label can be set manually as needed, for example:

[0034] The pixel value of the background in the image is 0, the pixel value of the ordinary road surface is 1, the pixel value of the gravel road surface is 2, the pixel value of the dirt road surface is 3, the pixel value of the grass road surface is 4, the pixel value of the mud is 5, the pixel value of the wading road surface is 6, the pixel value of the rock element is 7, the pixel value of the thick sand road surface is 8, the pixel value of the ordinary snow road surface is 9, and the pixel value of the thick snow road surface is 10.

[0035] In some embodiments, in order to ensure the consistency of the road scene distribution grid map and reduce the interference of external factors on the confirmation of cell types, step 2 specifically includes:

[0036] Based on a preset unit length and with the vehicle as the center, a road scene distribution grid is divided on the road in front of the vehicle;

[0037] Based on the disparity map corresponding to the road image and the road image, coordinate transformation is performed on the pixel points in the road image through disparity calculation to determine the cell corresponding to the pixel point in the road scene distribution grid map;

[0038] The number of pixels of each type of pixel category label in the cell is counted, and the pixel category label with the largest number is recorded as the cell label of the cell.

[0039] A coordinate system is established with the location of the vehicle equipped with the RGBD camera as the center, and a unified road scene distribution grid map is established based on the set unit length, such as Figure 2 As shown, the image collected by the RGBD camera can be mapped to the road scene distribution grid map by means of parallax calculation, so that by counting the pixel category labels of the corresponding pixels in each cell, the pixel category label with the largest number of pixels is used as the cell label of the cell.

[0040] After that, the area that needs attention can be divided in the road scene distribution grid map, recorded as the road ROI area, and the road type can be judged based on the road ROI area. In the process of determining the road ROI area, a certain area in front of the vehicle can be directly determined as a road ROI area, or a certain area in front of the vehicle can be divided into multiple road ROI areas of different numbers according to different preset road types.

[0041] However, considering factors such as tire adhesion and ease of getting out of the road during vehicle driving, preset road types with different priorities can be set. For example, the preset road types sorted by priority are:

[0042] Rocks > Mud > Deep snow / sand > Wading > Ice / snow / gravel / grass > General roads

[0043] This embodiment determines different road ROI areas based on preset road types of different priorities to improve the accuracy of determining road types and timely adjust the vehicle suspension to improve driving experience. Therefore, the road ROI area is divided as follows in this embodiment:

[0044] Step A: According to the order of road priority, the number of road ROI regions to be divided is determined in order from most to least, and the starting position of the first road ROI region is determined in order from near to far;

[0045] Step B: Divide each road ROI region according to the same area range according to the number of road ROI regions divided and the starting position of the first road ROI region.

[0046] Specifically, Figure 3 As shown in the figure, mud, grass, and general roads are used as examples to illustrate the three preset road types. In order to ensure driving experience and driving safety, when the road type is set to switch from high priority to low priority, the switch can be delayed, which is a low-sensitivity switch; when the road type is switched from low priority to high priority, in order to avoid insufficient vehicle torque, it must be switched immediately, which is a high-sensitivity switch.

[0047] Therefore, within the same area, when the preset road type corresponds to a higher priority, the determined road ROI area should cover the area ahead of the vehicle to be driven as much as possible; and when the preset road type corresponds to a lower priority, the area that has been driven before can be divided into the road ROI area, making full use of historical data and avoiding frequent switching of road types.

[0048] For preset road types with higher priorities, in order to improve the accuracy of all-terrain road type recognition, a larger number of road ROI areas should be divided within the same area to perform fine recognition of road images; otherwise, fewer road ROI areas can be set.

[0049] For example, when the preset road type is mud, the number of road ROI area divisions can be set to 6, and the starting position of the first road ROI area can be set to a certain position in front of the vehicle (such as 1 meter); when the preset road type is a general road, the number of road ROI area divisions can be set to 1, and the starting position of the first road ROI area can be set to the rear of the front wheel of the vehicle.

[0050] It should be noted that the number of road ROI regions divided and the starting position of the first road ROI region can be set according to actual conditions, but it should be ensured that each road ROI region divided in the road scene distribution grid map contains an integer number of cells.

[0051] After the road ROI areas corresponding to the preset road types are obtained by division, the cell labels of the largest type of cells in the road ROI area can be used as the area labels of the road ROI area.

[0052] In the road type decision process of this embodiment, each preset road type is set to correspond to a valid count and a valid threshold.

[0053] During the driving process of the vehicle, taking any moment as an example, it is assumed that the valid counts and valid thresholds corresponding to each preset road type at the current moment are as shown in Table 1.

[0054] Table 1

[0055] Preset Road Types Mud grassland General roads Valid count A1 B1 C1 Effective threshold A2 B2 C2

[0056] When the next sampling moment is reached, first, the binocular camera acquires the road image in front of the vehicle, and identifies the pixel category label of each pixel in the road image through methods such as AI recognition; secondly, the cell label of each cell in the road scene distribution grid map is determined based on each pixel category label; thirdly, according to the above three preset road types, the respective road ROI areas are determined respectively. It should be noted that even if the current road type is determined to be grass, it is still necessary to divide the road ROI areas corresponding to mud and general roads so that when driving out of the grass, it can be switched in time to ensure timeliness.

[0057] In some embodiments, step 4 specifically includes:

[0058] Count the number of road ROI areas whose area labels belong to the preset road type in any preset road type, and record it as the cumulative number;

[0059] According to the accumulated number, the values ​​of the valid counts of the preset road types are accumulated.

[0060] Therefore, based on the above embodiment, it is necessary to count the accumulated numbers and accumulate the valid counts of each preset road type.

[0061] Assumption: The current vehicle is traveling on a road with grass and mud. At the current sampling time, the corresponding results are as follows:

[0062] 1) One of the six road ROI areas corresponding to mud is determined to be mud, and the other five are determined to be grass, then the cumulative number corresponding to mud is "1";

[0063] 2) The three road ROI areas corresponding to the grassland are all determined to be grassland, so the cumulative number corresponding to the grassland is "3";

[0064] 3) If a single road ROI area corresponding to a general road is identified as grass, the accumulated number corresponding to the general road is "0".

[0065] Therefore, the accumulated effective count corresponding to the mud is A1+1, the effective count corresponding to the grass is B1+3, and the effective count corresponding to the general road is C1.

[0066] Finally, in the order of preset priorities (mud>grass>general road), first determine whether the effective count A1+1 of the mud is greater than or equal to the effective threshold A2. If so, it indicates that the proportion of mud in the road the vehicle has traveled has increased, and the type of the road ahead needs to be switched / maintained to "mud" to adjust the vehicle to make it suitable for driving on mud. If not, determine whether the effective count B1+3 of the grass is greater than or equal to the effective threshold B2. If so, switch / maintain the road ahead to "grass" without making a "general road" judgment. If not, determine whether the effective count C1 of the general road is greater than or equal to the effective threshold C2.

[0067] In some embodiments, in order to ensure the timeliness of each valid count and avoid interference of historical data on current road type decision, the method further includes:

[0068] Step 6: Based on the area labels, the invalid count values ​​of each preset road type are accumulated respectively; wherein, when it is determined that the area label of any road ROI area corresponding to the preset road type does not belong to the preset road type, the invalid count value of the preset road type is increased by one; if there is any road ROI area whose area label belongs to the preset road type, the invalid count value of the preset road type remains unchanged;

[0069] Step 7: When it is determined that the value of the invalid count is greater than or equal to the invalid threshold, the value of the valid count of the preset road type corresponding to the invalid count is set to 0.

[0070] Specifically, taking the scenario set in the above embodiment as an example, for "general road", since its corresponding road ROI area is identified as grass, that is, there is no area label belonging to "general road", the invalid count value of the preset road type "general road" is +1. If the value of the invalid count is greater than or equal to the invalid threshold, it indicates that there has been no driving on the general road for a period of time before. Therefore, the valid count (historical valid count) of the "general road" is set to 0 to avoid misjudgment.

[0071] As for "mud", since one of the corresponding six road ROI areas is determined to be mud, it means that there is still "mud" ahead. The invalid count value of "mud" remains unchanged, and the valid count value is +1. The "mud" that is about to pass or has passed is recorded to ensure the rationality of subsequent all-terrain road type recognition.

[0072] In some embodiments, considering that the pixel category labels of each pixel point determined by the AI ​​recognition model are pixel category labels with higher confidence, rather than real pixel category labels, therefore, in order to further improve the accuracy of all-terrain road type recognition, this embodiment also weights the pixel category label ratio (or quantity) of each cell at the current moment based on the ratio (or quantity) of the pixel category labels of each cell at the previous moment. Therefore, before step 3, the method further includes:

[0073] Determine a coordinate transformation matrix based on the driving information of the vehicle on the road;

[0074] Based on the coordinate transformation matrix, align the coordinates of each cell in the road scene distribution grid map at the previous moment with each cell in the road scene distribution grid map at the current moment;

[0075] According to the aligned cells, the pixel category label ratio of each cell at the previous moment is weighted to the pixel category label ratio of each cell at the current moment;

[0076] According to the weighted pixel category label ratio of each cell at the current moment, the cell label of each cell in the road scene distribution grid map at the current moment is updated.

[0077] Specifically, since the road scene distribution grid is set with the vehicle as the center, the coordinate transformation matrix of the vehicle at two adjacent moments can be calculated according to the vehicle's IMU information and vehicle speed, including the rotation matrix R and the translation matrix T, where:

[0078]

[0079]

[0080] β=ΔT*yaw

[0081] Where β is the rotation angle of the vehicle, ΔT is the time interval between two frames of images, yaw is the yaw angular velocity in the IMU information, V is the vehicle speed, and T x 、T z are the components in the translation matrix T.

[0082] Based on the calculated rotation matrix R and translation matrix T, the coordinates of each cell in the road scene distribution grid map at the previous moment can be aligned with the coordinates of each cell in the road scene distribution grid map at the current moment;

[0083] Then, the pixel category label ratio (or quantity) of the cells aligned at the previous and next moments is counted for weighting to improve the accuracy of the overall recognition result. The corresponding weighted calculation formula can be:

[0084] U t =U t-1 (1-α)+V t α

[0085] Where U t is the proportion of a certain type of pixel category label in a certain cell after alignment weighting, U t-1 V is the proportion of pixel category labels of the same type in the corresponding cell at the previous moment after weighted alignment. t is the ratio of the same type of pixel category labels of the corresponding cell at the current moment before the alignment weighting, and α is the preset weight, and its value can be set manually.

[0086] For example, assuming that the proportion of pixel category labels of grass in cell A at the previous moment is 60%, which corresponds to cell B at the current moment, and the proportion of pixel category labels of grass in cell B before weighting is 70%. If the preset weight α is 50%, the proportion of pixel category labels of grass in cell B after weighting is 65%, which can improve the accuracy and reliability of the "grass" pixel category label ratio.

[0087] It should be noted that if, after aligning the coordinates, a cell at the current moment does not correspond to any cell at the previous moment, the pixel category label ratio of the cell at the current moment is retained unchanged; similarly, if a cell at the previous moment does not correspond to any cell at the current moment, the pixel category label ratio of the cell at the previous moment is discarded.

[0088] In addition, the pixel category label ratio of each cell in the weighted road scene distribution grid map at the current moment will be used to weight the pixel category label ratio of each cell in the road scene distribution grid map at the next moment.

[0089] Embodiment 2:

[0090] like Figure 4 As shown, this embodiment provides an all-terrain road type recognition system 100, and the system 100 includes:

[0091] The recognition unit 10 is configured to recognize a pixel category label of each pixel point in the road image;

[0092] A first label determination unit 20, the first label determination unit 20 is configured to determine a cell label of each cell in the road scene distribution grid map based on the pixel category label;

[0093] A second label determination unit 30, the second label determination unit 30 is configured to divide the road ROI area of ​​each preset road type in the road scene distribution grid map based on the preset road type, and determine the area label of the road ROI area based on the cell label of each cell included in the road ROI area;

[0094] An accumulation unit 40, the accumulation unit 40 is configured to accumulate the values ​​of the valid counts of each type of preset road type based on the area label;

[0095] The decision unit 50 is configured to determine, based on a preset priority, a preset road type whose valid count value is greater than or equal to a valid threshold as a current road type.

[0096] In some embodiments, the system 100 is installed on a vehicle, which is provided with a state sensor, which is used to detect whether the vehicle is in a driving pending state, wherein the driving pending state includes a stationary state (vehicle speed is 0), a steering state (the vehicle steering angle is greater than a set angle threshold), a crab state, and a parking state. The decision unit 50 is also configured to keep the road type determined at the last moment unchanged when it is determined that the vehicle is in the driving pending state.

[0097] It should be noted that the system 100 in the embodiment of the present application can also be installed on a movable platform with a mobile function, such as a lawn mower, a motorcycle, a truck, etc.

[0098] Specifically, the all-terrain road type recognition system 100 in this embodiment can be integrated into the vehicle's own control system 100, connected to various status sensors on the vehicle through the CAN bus, and obtain vehicle body signals, such as vehicle speed, IMU signal, light intensity, steering wheel angle, car wash mode, OTA mode, crab mode, automatic parking mode and other signals.

[0099] In some embodiments, considering that when the rainfall is heavy, not only will there be misidentification during the image processing process, but it is also inappropriate to adjust the suspension of the vehicle, the decision unit 50 is also configured to keep the road type determined at the previous moment unchanged when it is determined that the speed of the vehicle's wiper is greater than or equal to the preset speed.

[0100] In some embodiments, the vehicle is also provided with a temperature sensor for detecting the ambient temperature outside the vehicle. The decision unit 50 is also configured to keep the road type determined at the last moment unchanged when it is determined that the ambient temperature is greater than a preset temperature and the current road type is determined to be an ice and snow type.

[0101] Specifically, when external environmental factors lead to misidentification, such as the influence of light, there is a possibility of misidentifying a flooded road as an ice or snow type. Therefore, by detecting the ambient temperature outside the vehicle, if the ambient temperature is greater than a preset temperature (such as 15°), the road type determined at the last moment remains unchanged.

[0102] In some embodiments, the accumulation unit 40 is further configured to: based on the area label, accumulate the values ​​of the invalid counts of each type of preset road type respectively; when it is determined that the value of the invalid count is greater than or equal to the invalid threshold, set the value of the valid count of the preset road type corresponding to the invalid count to 0.

[0103] In some embodiments, when it is determined that the region labels of any road ROI region corresponding to the preset road type do not belong to the preset road type, the value of the invalid count of the preset road type is increased by one.

[0104] In some embodiments, the accumulation unit 40 is further configured to: count the number of road ROI areas whose area labels belong to the preset road type in any preset road type, recorded as the accumulated number; and accumulate the values ​​of the valid counts of the preset road type according to the accumulated number.

[0105] In some embodiments, the first label determination unit 20 is further configured to: divide the road scene distribution grid map on the road in front of the vehicle based on a preset unit length and with the vehicle as the center; based on the road image and the disparity map corresponding to the road image, perform coordinate transformation on the pixel points in the road image through disparity operation to determine the cell corresponding to the pixel point in the road scene distribution grid map; count the number of pixels of each type of pixel category label in the cell, and record the pixel category label with the largest number as the cell label of the cell.

[0106] In some embodiments, the first label determination unit 20 is also configured to: determine a coordinate transformation matrix based on the driving information of the vehicle on the road; align the coordinates of each cell in the road scene distribution grid map at the previous moment with each cell in the road scene distribution grid map at the current moment based on the coordinate transformation matrix; weight the pixel category label ratio of each cell at the previous moment to the pixel category label ratio of each cell at the current moment in a weighted manner according to the aligned cells; and update the cell label of each cell in the road scene distribution grid map at the current moment according to the weighted pixel category label ratio of each cell at the current moment.

[0107] This embodiment sets preset road types with different priorities in combination with the actual situation during vehicle driving, and divides different road ROI areas for each preset road type in the same road scene distribution grid map, and then determines the regional label of the road ROI area based on the identified pixel category label. Finally, by accumulating effective counts, the preset road type with a higher priority and whose effective count reaches a threshold is determined as the current road type. This can accurately identify the road type ahead of the vehicle, and improve the driving experience of the vehicle while ensuring driving safety.

[0108] So far, various embodiments of the present application have been described in detail. In order to avoid obscuring the concept of the present application, some details known in the art are not described. Based on the above description, those skilled in the art can fully understand how to implement the technical solution disclosed herein.

[0109] Although some specific embodiments of the present application have been described in detail through examples, those skilled in the art should understand that the above examples are only for illustration rather than for limiting the scope of the present application.

[0110] The steps in this application can be adjusted in order, combined, and deleted according to actual needs.

[0111] Although the present application is disclosed in detail with reference to the accompanying drawings, it should be understood that these descriptions are merely exemplary and are not intended to limit the application of the present application. The scope of protection of the present application is defined by the appended claims and may include various modifications, alterations and equivalents made to the invention without departing from the scope and spirit of the present application.

Claims

1. A method for identifying all-terrain road types, characterized in that: The method comprises: Step 1: Identify the pixel category label of each pixel in the road image; Step 2: Based on the pixel category label, determine the cell label of each cell in the road scene distribution grid map; Step 3: Based on preset road types, dividing road ROI areas of various preset road types in the road scene distribution grid map, and determining area labels of the road ROI areas based on cell labels of cells in the road ROI areas; Step 4: Based on the area label, the valid count values ​​of each preset road type are accumulated respectively; Step 5: Based on the preset priority, the preset road type whose valid count value is greater than or equal to the valid threshold is determined as the current road type.

2. The all-terrain road type identification method according to claim 1, characterized in that: The method further comprises: Step 6: Based on the area label, the invalid count values ​​of each preset road type are accumulated respectively; Step 7: When it is determined that the value of the invalid count is greater than or equal to the invalid threshold, the value of the valid count of the preset road type corresponding to the invalid count is set to 0.

3. The all-terrain road type identification method according to claim 2, characterized in that: The step 6 specifically includes: When it is determined that the region labels of any road ROI region corresponding to a preset road type do not belong to the preset road type, the value of the invalid count of the preset road type is increased by one.

4. The all-terrain road type identification method according to claim 1, characterized in that: The step 4 specifically includes: Counting the number of road ROI areas whose area labels belong to any preset road type, and recording it as the accumulated number; According to the accumulated number, the values ​​of the valid count of the preset road type are accumulated.

5. The all-terrain road type identification method according to claim 1, characterized in that: The vehicle is provided with a binocular camera, and the binocular camera is used to obtain an image in front of the vehicle. The step 2 specifically includes: Based on a preset unit length, with the vehicle as the center, dividing the road scene distribution grid map on the road in front of the vehicle; Based on the road image and the disparity map corresponding to the road image, coordinate transformation is performed on the pixel points in the road image through disparity calculation to determine the cells corresponding to the pixel points in the road scene distribution grid map; The number of pixels of each type of pixel category label in the cell is counted, and the pixel category label with the largest number is recorded as the cell label of the cell.

6. The all-terrain road type identification method according to claim 1, characterized in that: Before step 3, the method further includes: Determine a coordinate transformation matrix based on the driving information of the vehicle on the road; Based on the coordinate transformation matrix, coordinate alignment is performed between each cell in the road scene distribution grid map at the previous moment and each cell in the road scene distribution grid map at the current moment; According to the aligned cells, the pixel category label ratio of each cell at the previous moment is weighted to the pixel category label ratio of each cell at the current moment; According to the weighted pixel category label ratio of each cell at the current moment, the cell label of each cell in the road scene distribution grid map at the current moment is updated.

7. An all-terrain road type recognition system, characterized in that: The system comprises: an identification unit, wherein the identification unit is configured to identify a pixel category label of each pixel point in the road image; A first label determination unit, the first label determination unit being configured to determine a cell label of each cell in the road scene distribution grid map based on the pixel category label; A second label determination unit, the second label determination unit is configured to divide the road ROI area of ​​each preset road type in the road scene distribution grid map based on the preset road type, and determine the area label of the road ROI area based on the cell label of the cell in the road ROI area; An accumulation unit, the accumulation unit being configured to accumulate the values ​​of the valid counts of each type of preset road type respectively based on the area label; A decision unit is configured to determine, based on a preset priority, a preset road type whose valid count value is greater than or equal to a valid threshold as a current road type.

8. The all-terrain road type recognition system according to claim 7, characterized in that: The system is installed on a vehicle, and a state sensor is provided on the vehicle, and the state sensor is used to detect whether the vehicle is in a driving pending state, wherein the driving pending state includes a stationary state, a turning state, a crab state, and a parking state. The decision unit is also configured to keep the road type determined at the last moment unchanged when it is determined that the vehicle is in the pending driving state.

9. The all-terrain road type recognition system according to claim 7, characterized in that: The decision unit is further configured to keep the road type determined at the last moment unchanged when it is determined that the rotation speed of the wiper of the vehicle is greater than or equal to a preset rotation speed.

10. The all-terrain road type recognition system according to claim 7, characterized in that: The vehicle is also provided with a temperature sensor, which is used to detect the ambient temperature outside the vehicle. The decision unit is also configured to keep the road type determined at the last moment unchanged when it is determined that the ambient temperature is greater than a preset temperature and the current road type is determined to be an ice and snow type.

Images