Narrow channel traversing method based on visual detection

Abstract

The invention provides a narrow channel traversal method based on visual detection. According to the method, on the basis of the distribution condition, obtained through visual detection, of boundary lines of an operation area in the view field range of the self-walking equipment, when more than one boundary line appears in different areas of the view field of the self-walking equipment, the algorithm step of performing traversal operation on a narrow channel is executed. According to the application, the relative position of the self-walking device relative to the boundary line of the narrow channel can be determined directly based on the pixel information provided by the visual module, and the self-walking device is driven to successively traverse the area in a reciprocating manner in a manner of being parallel to the side edge of the narrow channel based on the relative position of the boundary line, so that uniform operation in the narrow channel is realized. Sensing equipment does not need to be added, and the operation flatness and the trimming effect of the self-walking equipment are improved by regulating and controlling the traversal mode of the self-walking equipment in the narrow channel area in a software algorithm mode.

Description

Technical Field

[0001] The present application relates to the technical field of intelligent lawn mowing equipment, and in particular to a narrow channel traversal method based on visual detection. Background Art

[0002] The current boundaryless mowing solutions for automatic lawn mowers mainly use RTK technology, UWB positioning technology, LiDAR positioning technology, visual detection and other means to determine the boundaries of the lawn. The existing recognition algorithms and narrow channel judgment logic make it difficult to identify the narrow channels on the running path of the automatic walking equipment based on visual detection technology, making it difficult to traverse the narrow channel area.

[0003] Solutions based on RTK, UWB, and LiDAR can determine the distance between the lawn mower and the boundary of the grass. The premise is that a person needs to remotely control the lawn mower or take the lawn mower's mapping module to walk a circle along the trajectory where the lawn mower needs to operate. At this time, the lawn mower or mapping module will create an operation map for the lawn mower according to the positioning coordinates when it walks a circle along the grass boundary. The narrow channel in this operation map can be obtained by judging the relationship between points and lines, line segments and line segments of each boundary coordinate in the global map. However, this technology requires manual control of the automatic walking equipment to perform traversal modeling of the narrow channel area in advance, so it is difficult to implement in actual applications and there are certain operational risks.

[0004] After determining the boundary of the grassland, there are currently two solutions for the intelligent lawn mower to traverse and mow the working area within the boundary. One is planned traversal mowing: after the grassland boundary is manually delineated, the lawn mower automatically generates a traversal planning path based on the work map. The other is random traversal mowing: when the machine runs to the grassland boundary, it is confirmed to have reached the grassland boundary through the above-mentioned boundaryless positioning solution, and then the machine runs in a straight line to the next grassland boundary after a random angle, and completes the traversal of the entire grassland through the accumulation of time and running space.

[0005] Current lawn mower products only have the function of passing through narrow passages, but do not have the function of mowing the grass in narrow passages. When the current automatic walking lawn mower is doing lawn mowing operations, if there is a slender passage in the grass, such as a 5m long and 1m wide passage, when the lawn mower detects the grass boundary based on the camera, the existing boundary avoidance mechanism makes it very unlikely that the device can enter the narrow passage, and the lawn mower can only complete the full coverage mowing operation of the grass area in the narrow passage in the limited number of times it passes through the narrow passage. Summary of the invention

[0006] In view of the shortcomings of the prior art, the present application provides a narrow channel traversal method based on visual detection. The present application identifies the working area in the form of a narrow channel based on visual detection technology, and through a special driving mode for the traversal path of the narrow channel area, it can ensure the working effect of the equipment on the special area of ​​the narrow channel. The present application specifically adopts the following technical solutions.

[0007] Firstly, to achieve the above-mentioned purpose, a narrow channel traversal method based on visual detection is proposed, and the steps include: based on the working area boundary obtained by visual detection, when the distribution state of the working area boundary meets the narrow channel judgment standard, marking the current position of the self-propelled device as the starting point of the narrow channel; driving the self-propelled device to run along a single working area boundary from the starting point of the narrow channel until the distribution state of the working area boundary obtained by visual detection no longer meets the narrow channel judgment standard; driving the self-propelled device to turn around and run along the last running path by a preset distance until the distribution state of the working area boundary obtained by visual detection no longer meets the narrow channel judgment standard; repeating the previous step until the distance between the self-propelled device and another working area boundary reaches a safe distance.

[0008] Optionally, in any of the above-mentioned narrow channel traversal methods based on visual detection, the narrow channel determination criterion is: the boundaries of the working area obtained based on visual detection are located on the left and right sides of the field of view of the self-propelled device. Therefore, the present application can not trigger the narrow channel traversal when the lawn mower passes through a narrow grass channel on the left or right side. At this time, the lawn mower can rotate and move forward according to the normal random mowing plan; only when the grass boundaries appear on the left and right sides in front of the lawn mower at the same time and the lawn mower judges that the grass area in the middle can be entered by the visual image, the narrow channel determination criterion will be met, and the lawn mower will move forward to enter this area. After entering the area, when the machine runs a certain distance in the narrow channel and can detect the end of the narrow channel, it is determined that the area it has just walked through is the narrow channel range. At this time, the traversal mechanism can be triggered, and the reciprocating operation can be performed to perform safe mowing operations on this special area.

[0009] Optionally, in any of the narrow channel traversal methods based on visual detection as described above, the narrow channel determination criteria also include: the spacing distance between the working area boundaries on the left and right sides of the field of view of the self-propelled device is not less than 1.5 times the width or length of the self-propelled device body.

[0010] Optionally, in a narrow channel traversal method based on visual detection as described above, when a self-propelled device is driven to run along a single work area boundary from the starting point of a narrow channel: first adjust the walking direction of the self-propelled device so that the distance between the self-propelled device and any work area boundary on its left or right side reaches a safe distance, and keep the self-propelled device moving forward along any of the work area boundaries at a safe distance.

[0011] Optionally, a narrow channel traversal method based on visual detection as described above, wherein the preset distance for translation along the last running path is any of the following widths: the body width of the self-propelled device, the distance between the running wheels of the self-propelled device, the operating width of the operating device in the self-propelled device, 0.5m or any other preset fixed width.

[0012] Optionally, in any of the narrow channel traversal methods based on visual detection as described above, before driving the self-propelled device to turn around, the self-propelled device is driven to run at least twice a preset distance along the boundary of a single working area.

[0013] Optionally, in any of the narrow channel traversal methods based on visual detection as described above, the safe distance between the self-propelled device and the boundary of the working area is a distance that ensures that the front wheels of the self-propelled device will not exceed the boundary of the working area when the self-propelled device moves forward or rotates left or right.

[0014] Optionally, in a narrow channel traversal method based on visual detection as described in any of the above, for the interval distance, preset distance, and safety distance, the self-propelled device is verified in the following manner: the pixel coordinates of the above distances detected by a monocular camera, a binocular camera, or a depth camera in the self-propelled device are obtained; for the pixel coordinates of the monocular camera, the above distances are verified according to the viewing angle range corresponding to each pixel in the self-propelled device; for the pixel coordinates of the binocular camera, a disparity map is calculated according to a stereo matching algorithm, and the depth information corresponding to the pixel coordinates is determined based on the disparity map, and each pixel is aligned according to the depth information to verify the above distances; for the pixel coordinates of the depth camera, the above distances are verified according to the alignment relationship between the depth camera detection point and the pixel.

[0015] At the same time, in order to achieve the above-mentioned purpose, the present application also provides an automatic walking device, which includes a working device and a walking drive device. The automatic walking device drives the walking drive device to traverse the working area in a narrow channel based on visual detection according to any of the method steps described above, and synchronously drives the working device to operate.

[0016] In addition, the present application also provides a computer-readable storage medium, which is arranged in an automatic walking device and is used to control the automatic walking device to perform traversal operations in a narrow channel area according to any of the method steps described above during the operation of the automatic walking device. Beneficial Effects

[0017] The present application provides a narrow channel traversal method based on visual detection. The present application is based on the distribution of the boundary lines of the working area obtained by visual detection within the field of view of the self-propelled device. When more than one boundary line appears in different areas of the field of view of the self-propelled device, the algorithm steps of traversing the narrow channel are executed. The present application can directly determine the relative position of the self-propelled device relative to the boundary line of the narrow channel based on the pixel information provided by the visual module, and drive the self-propelled device to traverse the area back and forth in a manner parallel to the side of the narrow channel based on the relative position of the boundary line, so as to achieve uniform operation inside the narrow channel. The present application does not need to add sensing equipment, but improves its operation flatness and trimming effect by regulating the traversal method of the self-propelled device in the narrow channel area through software algorithms.

[0018] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent from the description, or may be understood by practicing the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings are used to provide a further understanding of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings: Figure 1 It is a schematic diagram of an automatic walking device passing near a narrow passage under the prior art; Figure 2 It is a schematic diagram of a narrow passage detected by the automatic walking device lock of the present application; Figure 3 This is a schematic diagram of the camera shooting range on the automatic walking device in this application; Figure 4 It is a schematic diagram of the translation mode of the running path of the automatic walking device of the present application during each traversal process; Figure 5 It is a schematic diagram of the traversal operation process of the automatic walking equipment of the present application. DETAILED DESCRIPTION

[0020] In order to make the purpose and technical solution of the embodiment of the present application clearer, the technical solution of the embodiment of the present application will be clearly and completely described in conjunction with the drawings of the embodiment of the present application. Obviously, the described embodiment is a part of the embodiment of the present application, not all of the embodiments. Based on the described embodiment of the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0021] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as generally understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with the meanings in the context of the prior art, and will not be interpreted with idealized or overly formal meanings unless defined as herein.

[0022] The meaning of "and / or" described in this application means that the situations where each exists alone or both exist at the same time are included.

[0023] The meaning of "inside" and "outside" mentioned in this application refers to the direction relative to the working area itself, that is, the direction pointing to the inside of the working area is inside, and the opposite direction is outside; it is not a specific limitation on the device mechanism of this application.

[0024] The meaning of "left and right" in this application refers to that when the user is facing the direction of travel of the self-propelled device, the user's left is left and the user's right is right, rather than a specific limitation on the device mechanism of this application.

[0025] The term “connection” as used in this application may mean a direct connection between components or an indirect connection between components via other components.

[0026] The meaning of "up and down" mentioned in this application refers to that when the user is facing the forward direction of the self-propelled device, the direction from the working ground to the top of the self-propelled device body is up, and the opposite is down, rather than a specific limitation on the device mechanism of this application.

[0027] The present invention proposes a camera-based lawn mower grass narrow channel traversal method. This application mainly implements random traversal mowing based on a visual RGB camera. Figure 1 In the prior art, this solution has the problem of difficulty in identifying narrow channels or inability to traverse and mow in narrow channels. The present application uses a camera installed on the intelligent mowing robot to perform boundary recognition based on the field of view image of the working area collected by the camera, and calculates the machine's posture in conjunction with the device's own real-time wheel speed meter and inertial navigation system. Thus, based on the determination of narrow channels in the grassland in the field of view, the machine is intelligently controlled through a fusion algorithm to complete the traversal and mowing of the local grassland in the narrow channel.

[0028] Specific reference Figure 3As shown, a visual detection device is fixedly installed on the upper part of the fuselage of the automatic walking device, which can be optionally set as a monocular camera, a binocular camera or a depth camera. Since the installation position of the visual detection device on the fuselage is fixed, the actual position of each pixel position in the field of view image taken by it relative to the self-propelled device is also relatively fixed. Therefore, the present application can use the field of view image to determine the narrow channel according to the camera's viewing angle, the camera's installation height, and the corresponding position of the grass boundary in the field of view image, so as to determine the narrow channel based on the working area boundary in the field of view image clarified by visual detection, and according to the position relationship of each pixel point in the field of view image relative to the device body, when the distribution state of the working area boundary relative to the fuselage meets the narrow channel determination standard, the current position of the self-propelled device is marked as the starting point of the narrow channel, so that the device is set to enter the narrow channel traversal mode. In this mode, the control unit of the equipment first drives the self-propelled device to run from the starting point of the narrow channel along the boundary of the working area on a single side of the channel until the device passes through the narrow channel so that the distribution state of the boundary of the working area obtained by visual detection no longer meets the narrow channel judgment standard; then, the self-propelled device is driven to turn around and return along the previous running path for a certain preset distance until the device returns to the side of the starting point in the narrow channel from the opposite direction, and then the device continues to turn around and continue to traverse the narrow channel in another direction based on the distribution state of the boundary of the working area obtained by visual detection no longer meeting the narrow channel judgment standard. Thus, the previous step is repeated until the distance between the self-propelled device and the boundary of another working area reaches a safe distance, and it can be considered that the device has completed the traversal operation of all narrow channel areas under the premise of safe operation, and thus the device can be driven to leave the narrow channel area and continue to traverse the continuous larger working area.

[0029] RGB cameras are most commonly used in visual solutions to collect market images of the work area. In more advanced models, RGB cameras can also be combined with depth cameras to obtain grassland boundary information through the combined detection of the two sensors.

[0030] The single RGB solution generally uses an RGB camera installed at the front of the lawn mower, and the camera's shooting angle is set to be diagonally downward from the front of the body. The distance of the ground position corresponding to each pixel relative to the body is calibrated in advance according to the camera installation height and the actual imaging position of the pixel in the RGB image. Through this calibration, the approximate distance of the ground position corresponding to each pixel relative to the body can be estimated. Therefore, through traditional image processing methods, such as identifying the boundary between grass and non-grass through image texture, color and other information, the approximate position relationship between the boundary line and the device body can be estimated through the pixel position corresponding to the boundary. In models with sufficient computing power, different pixel areas in the image can also be identified and segmented through deep learning solutions to identify the boundary of the grass. Therefore, the distance information between the grass boundary and the device body is determined based on the pixel position of the grass boundary detected by RGB.

[0031] For the solution of RGB camera plus depth camera, the pixels obtained by the RGB camera and the depth camera can be aligned, and each depth sampling point in the depth camera can be corresponded to a pixel area in the pixel image. After the pixel position of the boundary line of the working area is determined based on the RGB information, the depth camera can provide the depth distance of the grass boundary identified in the RGB camera relative to the device body in real time.

[0032] The above-mentioned RGB camera can be a monocular camera or a binocular camera. If a monocular camera is used, the position of the grass boundary relative to the camera on the lawn mower can be estimated by the height and angle of the camera installation, the distance of the pixels in the camera output image in the actual space, and other information. Figure 1 As shown, if the camera is installed at 101, the distance between the grass and the front of the car is L1. When the camera is installed at a certain height, the depth distance of different pixels within the viewing angle can be estimated.

[0033] If a binocular camera is used, the disparity map is calculated through a stereo matching algorithm to obtain a depth map, which is then matched with the grass boundary detection map to obtain the position of the grass boundary relative to the camera on the lawn mower body.

[0034] refer to Figure 2 Usually, there are several types of narrow grass passages in the grass: 1. Two large lawns connecting the passage 2. A narrow passage is formed between large objects and the grass or large objects in a large grassland.

[0035] 3. A narrow grassland formed by a part of the grassland due to the terrain; In these cases, the autonomous walking device can determine the boundary line of the working area corresponding to the grass through the RGB color range or the texture characteristics of the grass area, which is the light green part that appears on both sides of the detection image.

[0036] Considering that the prerequisite for the machine camera to identify narrow grass channels is that it needs to be able to observe the grass boundaries on both sides at the same time when entering the narrow channel. Therefore, the narrow channel judgment standard in the above driving strategy can be: the working area boundaries obtained based on visual detection are located on the left and right sides of the field of view of the self-propelled device. At this time, the device can enter the narrow channel area along the boundary line on either side to operate.

[0037] In contrast, when the mower passes through a narrow grass channel on the left or right side, the mower can rotate and move forward according to the normal random mowing plan. Only when the grass boundaries appear on the left and right sides in front of the mower and the mower determines through the visual image that the machine can enter the grass area in the middle, the mower will move forward and enter this area. When the machine runs a certain distance in the narrow channel and can detect the end of the narrow channel, it is judged that this area is a narrow channel, and then the traversal mechanism is triggered, triggering the automatic walking device to return backwards to operate on other positions in the narrow channel. In this way, the path spacing during the reciprocating operation of the automatic walking device can be adjusted, so as to drive it to gradually complete the traversal operation of the entire channel area during the reciprocating traversal process.

[0038] Considering the limitation of the installation height of the camera on the lawn mower and the field of view of the camera lens, the narrow passage can generally be defined as an area that meets the following requirements: the narrowest is 1.5 times the length of the lawn mower body and the widest is 2m; or the length of the area is 2 times or more than 2 times the width. For example, if the length of a lawn mower is 0.5m, the minimum width of the narrow passage is 0.75m and the minimum length is 1m.

[0039] Therefore, when the self-propelled device is in operation, when the left and right sides of its camera view simultaneously detect the grass boundaries and the grass width in the middle is wide enough for the mower to pass through, the position can be determined as the starting point of the narrow passage. At this time, the machine can be driven to adjust the walking direction of the self-propelled device so that the distance between the self-propelled device and any boundary of the working area on its left or right side reaches a safe distance, and the self-propelled device is kept in accordance with the narrow passage. Figure 4 As shown, the machine moves forward along any of the boundaries of the working area while maintaining a safe distance, and the machine's running trajectory is calculated based on the machine's wheel speed meter and IMU. After running forward for a certain distance, the camera's field of view cannot simultaneously detect the grass boundary on both sides, and it is determined to be the end point. At this time, the lawn mower will determine that the area is a narrow channel based on the running distance from the starting point to the end point when the self-propelled device is driven to run at least twice the preset distance along the boundary of a single working area.

[0040] When the distance meets the narrow channel judgment logic, the machine will start the traversal mowing control logic in the narrow channel, refer to Figure 4 In this way, the position of the boundary line of the working area determined by its image sensor device is used to add a preset translation distance along the last running path, and this path is used as the path for the next passage through the narrow channel. Thus, by successively translating the path through the narrow channel, the running track of the automatic walking device is covered with the narrow channel, so as to achieve the trimming of the grass in the narrow channel. During the above-mentioned reciprocating translation traversal process, reciprocating operation can be performed in the narrow channel area according to any of the following widths until the operating path of the equipment cannot meet the safety distance requirements: the body width of the self-propelled device, the distance between the running wheels of the self-propelled device, the operating width of the operating device in the self-propelled device, 0.5m or any other preset fixed width.

[0041] When the distance does not meet the narrow channel judgment logic, the machine will delete the narrow channel starting point mark and continue to perform the traversal mowing task.

[0042] During the above operation, in a narrow channel, the safe distance between the self-propelled device and the boundary of the working area refers to the distance that the front wheels of the machine will not go beyond the boundary of the grass when the machine moves forward or rotates left or right. It can generally be determined based on factors such as the length and width of the body of the self-propelled device and its minimum turning radius.

[0043] In order to accurately control the safe operation of the equipment, during the above operation, the equipment can verify the separation distance, preset distance, and safety distance in the following manner: For a self-propelled device equipped with a monocular camera, a binocular camera, or a depth camera, first obtain the pixel coordinates of the above distance detected by the monocular camera, the binocular camera, or the depth camera in the self-propelled device; Then, for the pixel coordinates of the monocular camera, the above distances are verified according to the viewing angle range corresponding to each pixel in the self-propelled device; Alternatively, for the pixel coordinates of the binocular camera, a disparity map is calculated according to a stereo matching algorithm, and depth information corresponding to the pixel coordinates is determined based on the disparity map, and each pixel is registered according to the depth information to verify the above distances; Alternatively, for the pixel coordinates of the depth camera, the above distances may also be verified according to the registration relationship between the depth camera detection points and the pixels.

[0044] by Figure 5Take this as an example. When the above method identifies that the machine is passing through a narrow channel, the machine will start the narrow channel traversal mowing mode at the position determined as the end of the narrow channel. In this mode, the machine will turn around and enter the narrow channel, and use the camera to maintain the distance between the mower and one of the grass boundaries. Assuming that the grass edges of the narrow channel are A and B, the control can be set to start the mowing operation from the edge A close to the narrow channel. When the mower reaches the other end, the distance between the mower and the grass edge A is increased by a blade cutting width to start the mowing operation of the next path, and the cycle continues. When the mower reaches the set safe distance from the grass edge B, the machine exits the mowing operation of the narrow channel and enters the random traversal mower operation.

[0045] In order to prevent the lawn mower from entering the narrow channel multiple times and repeatedly traversing and mowing the grass, the present application can complete the traversal and mowing operation of the narrow channel when the lawn mower passes through the narrow channel for the first time based on the above settings.

[0046] When the machine is randomly traversing the mowing operation, it may enter the narrow channel area many times. At this time, it can be judged whether the narrow channel has completed the mowing operation through vision or the load of the mower disc to determine whether the operation is needed.

[0047] For example, when the load size of the blade disc is used for judgment, when the mower enters a narrow channel that has been cut, the no-load operation state of the blade disc can be used to determine that the area does not need to continue mowing. At this time, the mower can be controlled to pass through the narrow channel directly in the center and enter the lawn at the other end of the channel, or the mower can be controlled to exit the narrow grass area. When visual judgment is made, the judgment of the completed area can be achieved based on the color range or texture characteristics of the grass after mowing, or combined with a machine learning recognition module. When it is determined that the narrow channel does not need to be mowed, the mower can also be controlled to pass through the narrow channel directly in the center and enter the lawn at the other end of the channel, or the mower can be controlled to exit the narrow grass area.

[0048] Therefore, the present application can quickly determine the area of ​​the narrow channel that requires independent planning of the working path by the distribution of the working area boundary relative to the field of view of the self-propelled device based on the visual detection, guide and position the equipment based on the boundary line of the narrow channel area, complete the grass traversal in the narrow channel, and improve the lawn maintenance efficiency of the lawn mower. The traversal operation method for the narrow channel area provided by the present application is compatible with the lawn mower vision solutions of the monocular camera, binocular camera, and depth camera currently used by the equipment.

[0049] The above is only an implementation method of the present application, and its description is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present application. It should be pointed out that for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present application, which all belong to the protection scope of the present application.

Claims

1. A narrow channel traversal method based on visual detection, characterized in that the steps include: Based on the working area boundary obtained by visual detection, when the distribution state of the working area boundary meets the narrow channel judgment standard, the current position of the self-propelled device is marked as the starting point of the narrow channel; The self-propelled device is driven to run along the boundary of a single working area from the starting point of the narrow channel until the distribution state of the boundary of the working area obtained by visual inspection no longer meets the narrow channel determination standard; The self-propelled device is driven to turn around and move along the last running path by a preset distance until the distribution state of the boundary of the working area obtained by visual inspection no longer meets the narrow channel judgment standard; Repeat the previous step until the distance between the self-propelled device and the boundary of another working area reaches a safe distance.

2. The narrow channel traversal method based on visual detection as claimed in claim 1, characterized in that: The narrow channel judgment standard is: the boundaries of the working area obtained based on visual detection are located on the left and right sides of the field of view of the self-propelled device.

3. The narrow channel traversal method based on visual detection as claimed in claim 2, characterized in that: The narrow channel determination criteria also include: the distance between the boundaries of the working area on the left and right sides of the field of view of the self-propelled device is not less than 1.5 times the width or length of the self-propelled device.

4. The narrow channel traversal method based on visual detection as claimed in claim 1, characterized in that: When the self-propelled device is driven from the starting point of a narrow channel to run along the boundary of a single working area: First, adjust the travel direction of the self-propelled device so that the distance between the self-propelled device and any working area boundary on its left or right side reaches a safe distance, and keep the self-propelled device moving forward along any working area boundary at a safe distance.

5. The narrow channel traversal method based on visual detection as claimed in claim 1, characterized in that: The preset distance of translation along the last running path is any of the following widths: the body width of the self-propelled device, the running wheel spacing of the self-propelled device, the working width of the working device in the self-propelled device, 0.5m or any other preset fixed width.

6. The narrow channel traversal method based on visual detection as described in claims 1-5, characterized in that: Before driving the self-propelled device to turn around, the self-propelled device shall be driven to run at least twice the preset distance along the boundary of the single working area.

7. The narrow channel traversal method based on visual detection as described in claims 1-5, characterized in that: The safe distance between the self-propelled device and the boundary of the working area is: When the self-propelled device moves forward or rotates left or right, ensure that its front wheels do not exceed the boundary of the working area.

8. The narrow channel traversal method based on visual detection as described in any one of claims 3 to 7, characterized in that: For the separation distance, preset distance and safety distance, the self-propelled device is verified in the following manner: Obtain pixel coordinates for the above distance detected by a monocular camera, a binocular camera or a depth camera in the self-propelled device; For the pixel coordinates of the monocular camera, verify the above distances according to the viewing angle range corresponding to each pixel in the self-propelled device; For the pixel coordinates of the binocular camera, a disparity map is calculated according to a stereo matching algorithm, and depth information corresponding to the pixel coordinates is determined based on the disparity map, and each pixel is registered according to the depth information to verify the above distances; For the pixel coordinates of the depth camera, the above distances are verified according to the registration relationship between the depth camera detection points and the pixels.

9. An automatic walking device, characterized in that: The automatic walking device comprises an operating device and a travel drive device. The automatic walking device drives the travel drive device to traverse the operating area in a narrow channel based on visual detection according to any of the above method steps, and synchronously drives the operating device to operate.

10. A computer-readable storage medium, characterized in that: It is arranged in an automatic walking device and is used to control the automatic walking device to perform traversal operations in a narrow channel area according to any of the method steps described above during its operation.

Images