Map partitioning and construction, object recognition and cleaning method, device, and storage medium

Abstract

This application provides a map partitioning and construction method, object recognition and cleaning method, device, and storage medium. In this application embodiment, because the autonomous mobile device uses multiple trajectory patterns with different main travel directions to traverse the working environment, it can cover the entire working environment as much as possible, thereby capturing as many partition reference objects in the working environment as possible, and increasing the probability of capturing frontal angle images of partition reference objects. This is beneficial for more accurately identifying partition reference objects and their locations in the working environment, and improving the accuracy of map partitioning based on partition reference objects.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a method, device and storage medium for map partitioning and construction, object recognition and cleaning. Background Technology

[0002] With the development of artificial intelligence technology, robots are becoming increasingly intelligent. For example, robotic vacuum cleaners free up users' hands, automatically cleaning floors throughout the room.

[0003] Robotic vacuum cleaners are becoming increasingly powerful, supporting a wider range of trajectory modes, such as zoned cleaning. As they navigate their environment, robotic vacuum cleaners automatically detect environmental information using onboard sensors such as LiDAR and visual sensors, constructing an environmental map based on this information and intelligently dividing the map into zones. However, existing zoning methods suffer from inaccuracies. Summary of the Invention

[0004] This application provides a map partitioning and construction method, object recognition and cleaning method, device and storage medium in several aspects to improve the accuracy of environmental map partitioning.

[0005] This application provides a map partitioning method, comprising: controlling an autonomous mobile device to traverse a work environment using at least two trajectory modes, and acquiring multiple environmental images during the traversal using at least two trajectory modes; marking the positions of partition reference objects in an environmental map corresponding to the work environment based on the frequency and position of the partition reference objects appearing in the multiple environmental images; partitioning the environmental map according to the positions of the partition reference objects in the environmental map; wherein the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0006] This application also provides a semantic map construction method, including: controlling an autonomous mobile device to traverse the working environment using at least two trajectory modes; acquiring multiple environmental images during the traversal using the at least two trajectory modes; and constructing a semantic map of the working environment based on the multiple environmental images; wherein the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0007] This application embodiment also provides an object recognition method, including: controlling an autonomous mobile device to traverse the work environment using at least two trajectory modes; acquiring multiple environmental images during the traversal using the at least two trajectory modes; and identifying target objects existing in the work environment based on the multiple environmental images; wherein the at least two trajectory modes have different main travel directions, and the main travel direction refers to the travel direction that meets set conditions.

[0008] This application embodiment also provides a cleaning method, including: controlling a sweeping robot to perform a cleaning task in a work area using a first trajectory mode; after completing the cleaning task using the first trajectory mode, controlling the sweeping robot to continue to perform the cleaning task again in the work area using a second trajectory mode; wherein, the main travel direction of the first trajectory mode and the second trajectory mode are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0009] This application embodiment also provides an autonomous mobile device, including: a device body, the device body having a memory and a processor; the memory for storing a computer program; the processor, coupled to the memory, for executing the computer program to: control the autonomous mobile device to traverse the working environment using at least two trajectory modes, and acquire multiple environmental images during the traversal using at least two trajectory modes; mark the positions of the partition reference objects in the environmental map corresponding to the working environment according to the frequency and position of the partition reference objects in the multiple environmental images; partition the environmental map according to the positions of the partition reference objects in the environmental map; wherein the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0010] This application embodiment also provides an autonomous mobile device, including: a device body, wherein the device body is provided with a memory and a processor; the memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program for: controlling the autonomous mobile device to traverse the working environment using at least two trajectory modes; acquiring multiple environmental images during the traversal using the at least two trajectory modes; constructing a semantic map of the working environment based on the multiple environmental images; wherein the primary travel directions of the at least two trajectory modes are different, and the primary travel direction refers to the travel direction that meets the set conditions.

[0011] This application embodiment also provides an autonomous mobile device, including: a device body, wherein the device body is provided with a memory and a processor; the memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program for: controlling the autonomous mobile device to traverse the working environment using at least two trajectory modes; acquiring multiple environmental images during the traversal using the at least two trajectory modes; identifying target objects existing in the working environment based on the multiple environmental images; wherein the at least two trajectory modes have different main travel directions, and the main travel direction refers to the travel direction that meets the set conditions.

[0012] This application embodiment also provides an autonomous mobile device, including: a device body, the device body having a memory and a processor; the memory for storing computer programs; the processor, coupled to the memory, for executing the computer programs to: control a sweeping robot to perform a cleaning task in a work area using a first trajectory mode; and after completing the cleaning task using the first trajectory mode, control the sweeping robot to continue performing the cleaning task again in the work area using a second trajectory mode; wherein the main travel directions of the first trajectory mode and the second trajectory mode are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0013] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to implement the steps in the method of this application.

[0014] The map partitioning and construction, object recognition and cleaning methods, devices and storage media provided in this application embodiment, because the autonomous mobile device uses multiple trajectory modes with different main directions of travel to traverse the working environment, can cover the entire working environment as much as possible, thereby capturing as many partition reference objects in the working environment as possible, and increasing the probability of capturing frontal angle images of partition reference objects, which is conducive to more accurately identifying partition reference objects and their positions in the working environment, and improving the accuracy of map partitioning based on partition reference objects. Attached Figure Description

[0015] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0016] Figure 1 A flowchart illustrating a map partitioning method provided for an exemplary embodiment of this application;

[0017] Figure 2 This is a schematic diagram illustrating a W-shaped travel trajectory as exemplified in an embodiment of this application.

[0018] Figure 3a This is a schematic diagram illustrating a bow-shaped travel trajectory, which is an example of an embodiment of this application.

[0019] Figure 3b This is a schematic diagram illustrating an irregularly shaped travel trajectory, as exemplified in an embodiment of this application.

[0020] Figure 3c This is a schematic diagram illustrating another irregularly shaped travel trajectory, which is an example of an embodiment of this application.

[0021] Figure 4a This application provides an example of an interlaced "bow"-shaped cleaning path planning method.

[0022] Figure 4b This is another example of an interlaced "bow"-shaped cleaning path planning in an embodiment of this application;

[0023] Figure 5 This is an example of a partitioned environment map in an embodiment of this application;

[0024] Figure 6 A schematic flowchart of a cleaning method provided in an exemplary embodiment of this application;

[0025] Figure 7 A flowchart illustrating a semantic map construction method provided in an exemplary embodiment of this application;

[0026] Figure 8 A flowchart illustrating an object recognition method provided in an exemplary embodiment of this application;

[0027] Figure 9 A schematic diagram of the structure of a map partitioning device provided for an exemplary embodiment of this application;

[0028] Figure 10 A schematic diagram of a cleaning device provided for an exemplary embodiment of this application;

[0029] Figure 11 This is a schematic diagram of the structure of an autonomous mobile device provided as an exemplary embodiment of this application. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0031] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.

[0032] Figure 1 This is a flowchart illustrating a map partitioning method provided for an exemplary embodiment of this application. Figure 1 As shown, the method includes the following steps:

[0033] Step 101: Control the autonomous mobile device to traverse the working environment using at least two trajectory modes, and collect multiple environmental images during the traversal using at least two trajectory modes.

[0034] Step 102: Based on the frequency and location of the partition reference objects in the work environment in multiple environmental images, mark the location of the partition reference objects in the environmental map corresponding to the work environment.

[0035] Step 103: Divide the environment map into zones based on the position of the zone reference objects in the environment map; wherein, at least two trajectory modes have different main travel directions, and the main travel direction refers to the travel direction that meets the set conditions.

[0036] In this embodiment, the autonomous mobile device can be any device capable of autonomous movement, such as a robotic vacuum cleaner, a disinfection robot, or a shopping guide robot. Of course, the autonomous mobile device is not limited to the examples described. Specifically, the robotic vacuum cleaner can autonomously move within its working environment to clean it. The disinfection robot can autonomously move within its working environment to disinfect it. The guide robot can autonomously move within an area and act as a guide, explaining detailed information about the area.

[0037] In this embodiment, the autonomous mobile device supports at least two trajectory modes. Each trajectory mode supports slightly different travel trajectories, each with its own attribute data, and these trajectories can be distinguished by differences in their attribute data. In this embodiment, the attribute data of the travel trajectory includes at least the shape of the trajectory and the main direction of travel. Different trajectory modes support different main directions of travel, which refer to the direction of travel within the trajectory that meets set conditions. In practical applications, the set conditions for the main direction of travel are related to the shape of the trajectory and can be flexibly set as needed. Simultaneously, the selection of the main direction of travel is related to the shape of the trajectory; the selection of the main direction of travel differs for trajectories with different shapes. Different trajectory modes may have the same shape but different main directions of travel.

[0038] To facilitate understanding of the main direction of travel, the following explanation uses several different broken line shapes as examples.

[0039] Example 1:

[0040] Figure 2 This is a schematic diagram illustrating a W-shaped travel trajectory as exemplified in an embodiment of this application. The travel trajectory is W-shaped, and a W-shaped travel trajectory includes four segments (e.g., Figure 2 The line segments 1, 2, 3, and 4 shown in the diagram form a continuous straight-line trajectory, creating a broken-line pattern. The entire "W"-shaped trajectory has a primary direction of travel. When setting the primary direction of travel for a "W"-shaped trajectory, one of the straight-line trajectories (such as line segment 1) can be selected as the reference straight-line trajectory, and its direction can be set as the primary direction of travel for the "W"-shaped trajectory. Changing the direction of the reference straight-line trajectory results in "W"-shaped trajectories with different primary directions of travel, forming different trajectory patterns. For example, in trajectory pattern 1, the direction of the reference straight-line trajectory (i.e., the primary direction of travel in trajectory pattern 1) faces due south; in trajectory pattern 2, the direction of the reference straight-line trajectory (i.e., the primary direction of travel in trajectory pattern 2) faces due north. It should be noted that if the straight-line trajectories shown by line segments 1 and 3 are parallel, then their directions of travel are the same.

[0041] Example 2:

[0042] Figure 3a This is a schematic diagram illustrating a bow-shaped travel trajectory, as exemplified in an embodiment of this application. The trajectory is bow-shaped, and one bow-shaped travel trajectory includes six segments (e.g., Figure 3a The six line segments (1, 2, 3, 4, 5, and 6) shown in the diagram form a continuous straight-line trajectory, creating a broken-line pattern. The entire "bow"-shaped trajectory has a primary direction of travel. When setting the primary direction of travel for the "bow"-shaped trajectory, one of the line segments (such as segment 1) can be selected as the reference line segment. The autonomous mobile device's travel direction along the reference line segment is then set as the primary direction of travel for the "bow"-shaped trajectory. Changing the direction of the reference line segment results in "bow"-shaped trajectories with different primary directions of travel, forming different trajectory patterns. For example, in trajectory pattern 1, the direction of the reference line segment (i.e., the primary direction of travel in trajectory pattern 1) faces due south; in trajectory pattern 2, the direction of the reference line segment (i.e., the primary direction of travel in trajectory pattern 2) faces due north.

[0043] It should be noted that among the six straight segments in the "bow"-shaped trajectory, three segments are relatively long, and three segments are relatively short. For example, Figure 3a Line segments 1, 3, and 5 in the diagram represent three relatively long straight-line trajectories. Figure 3aLine segments 2, 4, and 6 in the diagram represent three shorter straight line trajectories. When selecting a reference straight line trajectory, you can choose any one of the three longer trajectories as the reference trajectory. For example, as shown in the diagram, you can choose the straight line trajectory shown in line segment 1 as the reference straight line trajectory. It should be noted that if the straight line trajectories shown in line segments 1, 3, and 5 are parallel, then the straight line trajectories shown in line segments 1 and 5 travel in the same direction, while the straight line trajectories shown in line segments 1 and 3 travel in opposite directions.

[0044] Example 3:

[0045] Figure 3b and Figure 3c This is a schematic diagram illustrating the trajectory of an irregularly shaped object, as exemplified in an embodiment of this application. Figure 3b The trajectory shown includes a long straight line and a curved line, which are connected. The curved line can be a regular curve, such as... Figure 3b As shown, the curved trajectory consists of multiple square waves that are connected end-to-end and equally spaced. In addition, the curved trajectory can also be irregular, such as... Figure 3c As shown, the curve trajectory includes multiple square-like waves that are connected end-to-end and have not all the same shape. Irregular curve trajectories are not limited to... Figure 3c As shown; correspondingly, regular curved trajectories are not limited to Figure 3b As shown. When setting the main direction of travel for irregular travel trajectories, a straight trajectory can be selected as the reference straight trajectory, and the travel direction of the autonomous mobile device along the reference straight trajectory is set as the main direction of travel. Changing the direction of the reference straight trajectory results in irregular travel trajectories with different main directions of travel, forming different trajectory patterns. For example, in trajectory pattern 1, the direction of the reference straight trajectory (i.e., the main direction of travel in trajectory pattern 1) is due south, while in trajectory pattern 2, the direction of the reference straight trajectory (i.e., the main direction of travel in trajectory pattern 2) is due north. As another example, in trajectory pattern 1, the direction of the reference straight trajectory (i.e., the main direction of travel in trajectory pattern 1) is 20° west of south, while in trajectory pattern 2, the direction of the reference straight trajectory (i.e., the main direction of travel in trajectory pattern 2) is 15° east of north.

[0046] In this embodiment, the reference coordinate system corresponding to the travel direction of the autonomous mobile device can be a Cartesian coordinate system XOY established on the horizontal plane of the operating environment, with the horizontal axis being the X-axis and the vertical axis being the Y-axis. The positive direction of the X-axis points north, and the negative direction of the X-axis points south. The positive direction of the Y-axis points east, and the negative direction of the Y-axis points west. In this case, the azimuth angle of the main travel direction ranges from 0 to 360°.

[0047] Autonomous mobile devices can perform tasks by repeatedly moving along the corresponding primary and secondary directions of travel in the working environment using supported trajectory modes. They can also use trajectory modes to traverse the working environment, thereby collecting surrounding environmental information using their onboard visual sensors. These visual sensors can be monocular, binocular, or depth sensors, but are not limited to these. Depending on the installation position of the visual sensor on the autonomous mobile device, its field of view has a certain relative relationship with the direction of travel of the autonomous mobile device. For example, the center line of the field of view may be aligned with the direction of travel, or the center line of the field of view may maintain a certain angle with the direction of travel. In this embodiment, it is preferred to illustrate that the center line of the visual sensor's field of view is aligned with the direction of travel of the autonomous mobile device. Considering the field of view of the visual sensor on the autonomous mobile device, the visual sensor can typically collect environmental information more comprehensively in front of the direction of travel, while it cannot collect or can only collect some environmental information on the left and right sides of the direction of travel.

[0048] In this embodiment, since the main travel direction of different trajectory modes is different, the environmental area covered by the visual sensor's field of view will be different. If an object in the working environment is not within the field of view of the visual sensor in the previous trajectory mode, it is very likely to appear within the field of view of the visual sensor in the next trajectory mode. Therefore, when the autonomous mobile device uses multiple different trajectory modes to traverse the working environment, it can collect multiple environmental images from different perspectives through its own onboard visual sensor, which can enable the autonomous mobile device to collect richer environmental information.

[0049] In this embodiment, controlling the autonomous mobile device to traverse the operating environment using at least two trajectory modes can be achieved in any of the following ways:

[0050] The first method: Regardless of the trajectory mode used, the autonomous mobile device will traverse the working environment from the same starting position. That is, after the autonomous mobile device has traversed the working environment using the previous trajectory mode, it will be controlled to return to the specified starting position and continue to traverse the working environment again using the next trajectory mode.

[0051] The second method is to control the autonomous mobile device to start from the current position and continue to traverse the working environment again using the next trajectory mode after the autonomous mobile device has completed the previous trajectory mode.

[0052] In this embodiment, when traversing the work environment using at least two trajectory patterns, the execution order between the at least two trajectory patterns is not limited. For example, the execution order between the at least two trajectory patterns can also be calculated according to a certain algorithm. In one optional embodiment, the execution order between the at least two trajectory patterns can be randomly determined, or the execution order between the at least two trajectory patterns can be preset. In another optional embodiment, the gaps between the travel trajectories of different trajectory patterns are different, such as... Figure 4b The first trajectory pattern shown (cleaning path A) ～ The corresponding horizontal bow-shaped trajectory pattern) and the second trajectory pattern (cleaning path B) ～ The corresponding vertical bow-shaped trajectory pattern), the first trajectory pattern has a smaller trajectory gap, and the second trajectory pattern has a larger trajectory gap. When using at least two trajectory patterns, the trajectory pattern with the smaller trajectory gap is preferred. Based on this, the execution order of at least two trajectory patterns can be determined according to the order of the trajectory gap from smallest to largest. In other words, for two adjacent trajectory patterns, the trajectory gap of the latter trajectory pattern is larger than that of the former. The first and second trajectory patterns are examples of two trajectory patterns supported when an autonomous mobile device is implemented as a robotic vacuum cleaner. Both the first and second trajectory patterns use a bow-shaped trajectory, the difference being the main direction of travel and the different trajectory gaps of the bow-shaped trajectory. Specifically, the main direction of travel corresponding to due south in the first trajectory pattern is due east (i.e.,...). Figure 4b Cleaning path A shown ～ (As shown by the arrow); the main direction of travel for the second trajectory pattern is due south (i.e., Figure 4b Cleaning path B shown ～ (as indicated by the arrow), and the second trajectory pattern corresponds to trajectory B. ～ The trajectory interval is greater than the travel trajectory A corresponding to the first trajectory pattern. ～ The trajectory interval.

[0053] When an autonomous mobile device traverses the work environment using each trajectory mode, it moves systematically according to the shape of the trajectory defined by that mode. Taking the first trajectory mode as an example, when the autonomous mobile device traverses the work environment along the trajectory defined by that mode, a target object in the work environment may be located directly in front of the autonomous mobile device's current direction of travel (which could be the main or secondary direction), or it may be located to the left or right of the autonomous mobile device's current direction of travel. If the frontal view of the vision sensor on the autonomous mobile device is consistent with its direction of travel, and the target object is directly in front of the autonomous mobile device's current direction of travel, it means that the target object is within the frontal view of the vision sensor. If the target object is to the left or right of the autonomous mobile device's current direction of travel, the target object may be within the side view of the vision sensor, or it may not be within any view of the vision sensor. When the target object is within the frontal view of the vision sensor, the vision sensor can capture an image of the target object from a frontal angle, resulting in an image of the target object from the frontal angle, referred to as a frontal angle image. When the target object is located within the side view of the vision sensor, the vision sensor can capture an image of the target object from the side angle, which is simply called a side-angle image. The features of the target object contained in the frontal angle image and the side-angle image will differ.

[0054] Furthermore, when the target object is not within any field of view of the visual sensor, the autonomous mobile device will not capture an image of the target object when traversing the operating environment using the first trajectory mode. However, when the autonomous mobile device traverses the operating environment using the next trajectory mode, the main direction of travel specified by the next trajectory mode is different from that specified by the previous trajectory mode, resulting in different field of view coverage of the visual sensor in the two trajectory modes. Therefore, when the autonomous mobile device changes its trajectory mode, it is very likely that the target object will be captured when the autonomous mobile device traverses according to the next trajectory mode.

[0055] It is understandable that when traversing the work environment based on different trajectory patterns, it is possible to capture images of more objects in the work environment and multiple images of the same object from different angles, without being limited by the field of view of the visual sensor.

[0056] Figure 4a This is an example of staggered, overlapping "bow"-shaped cleaning path planning. Figure 4aIn this example, the autonomous mobile device is a robotic vacuum cleaner, supporting a first trajectory mode and a second trajectory mode. The first trajectory mode requires movement along cleaning path A, while the second trajectory mode requires movement along cleaning path B. The main cleaning direction of cleaning path A is perpendicular to the main cleaning direction of cleaning path B. When the robotic vacuum cleaner uses the first trajectory mode and performs a regular "bow"-shaped cleaning pattern along cleaning path A in the work area, it can capture images of targets F and D from the front, but may capture images of targets C and E from the side or not. When the robotic vacuum cleaner uses the second trajectory mode and performs a regular "bow"-shaped cleaning pattern along cleaning path B in the work area, it can capture images of targets C and E from the front, but may capture images of targets F and D from the side or not. Because the robotic vacuum cleaner moves along different main directions, the field of view of its onboard visual sensors covers different areas. Therefore, compared to moving along only one main direction, moving along different main directions sequentially allows for the acquisition of images of targets at various locations in the work environment, resulting in richer environmental information.

[0057] It is understandable that during image recognition, the key features of the target object captured from a frontal angle image are more numerous than those captured from a non-frontal angle image. Therefore, the probability of recognizing the target object from a frontal angle image is greater than that from a non-frontal angle image. This embodiment controls the autonomous mobile device to traverse the working environment using various trajectory patterns, enabling it to capture images of the target object from multiple shooting angles (e.g., frontal and side angles), thereby increasing the probability of capturing a frontal angle image of the target object. Furthermore, using frontal angle images also improves the accuracy of target object recognition.

[0058] In real-world scenarios, the work environment is likely to have different functional areas. These areas can be divided by partitioning reference objects within the work environment. These partitioning reference objects are objects that divide the work environment and can be located on the boundaries between two functional areas; examples include various types of doors, partitions, or screens. There can be one or more partitioning reference objects. To ensure that the environmental map accurately reflects the functional areas of the work environment, it is necessary to partition the environmental map. Partitioning the environmental map aims to establish a one-to-one correspondence between the partitions on the environmental map and the functional areas in the work environment. This allows the environmental map to more accurately describe the actual work environment, enabling the autonomous mobile device to perform tasks in designated areas based on the partitioning results of the environmental map. For example, when the autonomous mobile device is a robotic vacuum cleaner, the environmental map might partition the area into bedroom, kitchen, balcony, bathroom, and living room. The user can control the robotic vacuum cleaner to clean any one of these partitions.

[0059] In this embodiment, the autonomous mobile device traverses the operating environment using at least two trajectory modes, potentially capturing images of all or nearly all partition reference objects in the environment, with a relatively high probability of capturing frontal images of the partition reference objects. Of course, among the multiple environmental images captured by the autonomous mobile device, some images include partition reference objects, while others do not. Therefore, environmental images containing partition reference objects can be identified from multiple images; furthermore, for environmental images containing partition reference objects, the positions of the partition reference objects within the environmental image can be calculated.

[0060] Because a functional zone reference point in the working environment may be captured once or multiple times during multiple traversals by the autonomous mobile device, the number of times a zone reference point appears in multiple environmental images may be one or more. The more times a zone reference point appears in multiple environmental images, the greater the probability that the zone reference point actually exists in the working environment. To mark the location of zone reference points on the environmental map more objectively and accurately, both the number of times and the location of the zone reference point appear in multiple environmental images can be considered. Specifically, the number of times the zone reference point appears in multiple environmental images can be counted based on the number of environmental images containing the zone reference point; finally, by combining the number of times and the location of the zone reference point appear in multiple environmental images, the location of the zone reference point is marked on the environmental map.

[0061] In this embodiment, the environmental map can be an environmental map constructed by autonomous mobile devices using SLAM (Simultaneous Localization and Mapping) computation, such as an LDS (Laser Direct Structuring) map, a raster map, or a semantic map.

[0062] In step 103, after identifying the location of the partition reference in the environment map, the areas on both sides of the partition reference in the environment map are separated by using the partition reference as a dividing line.

[0063] Figure 5 As an example, the partitioned environment map is divided into 6 areas. The gaps in areas 0, 1-4, and 6 represent the locations of doors. The area behind each door forms one area, and the area in front of each door forms area 5. Taking an indoor environment as an example, the environment map can be logically partitioned according to the positions of the doors. When the autonomous mobile device is a robotic vacuum cleaner, it can clean designated areas based on the partitioning results, and can classify rooms (e.g., bedroom, kitchen, balcony).

[0064] The map partitioning method provided in this application can cover the entire working environment as much as possible by using multiple trajectory modes with different main directions of travel on the autonomous mobile device. This allows for the capture of as many partitioning reference objects in the working environment as possible, and increases the probability of capturing frontal angle images of the partitioning reference objects. This is beneficial for more accurately identifying the partitioning reference objects and their locations in the working environment, and improves the accuracy of map partitioning based on partitioning reference objects.

[0065] In the above embodiments, an optional implementation of step 102 may be: selecting a target environment image containing a partition reference from multiple environment images; marking candidate locations containing partition references in the environment map according to the location of the partition references in the target environment image; removing candidate locations that have been marked less than a set threshold number of times to obtain the location of the partition references in the environment map.

[0066] In practical applications, some of the multiple environmental images collected by autonomous mobile devices contain zonal reference objects while others do not. Therefore, it is necessary to select the target environmental image that contains zonal reference objects from the multiple environmental images.

[0067] After locating the target environment image, the geographical location information of the zoning reference objects in the world coordinate system can be deduced based on their positions within the target environment image. The origin of the device coordinate system is determined based on the location information of the autonomous mobile device recorded when the target environment image was acquired. Using the origin of the device coordinate system and the transformation relationship between the world coordinate system and the device coordinate system, the geographical location information of the zoning reference objects in the world coordinate system is transformed to obtain their position information in the device coordinate system. The device coordinate system is the system established on the autonomous mobile device when creating the environment map. For example, the device coordinate system can be a rectangular coordinate system XOY, with the horizontal axis being the X-axis and the vertical axis being the Y-axis. The positive direction of the X-axis points north, and the negative direction points south. The positive direction of the Y-axis points east, and the negative direction points west.

[0068] Since the environment map is also built using the device coordinate system, the location information of the partition reference objects in the device coordinate system can be marked on the environment map.

[0069] In practical applications, the higher the frame rate of the visual sensor on the autonomous mobile device, the more times the reference objects within the visual sensor's field of view are captured. Furthermore, when the autonomous mobile device traverses the operating environment according to multiple different trajectory patterns, it also increases the number of times the reference objects within the visual sensor's field of view are captured.

[0070] In this embodiment, the more times the same partition reference appears in multiple environmental maps, the greater the probability that the partition reference actually exists in the working environment. The number of times the same partition reference appears in multiple environmental maps is reflected on the environmental map as the number of times candidate locations are marked. Therefore, in order to more accurately identify the actual location of the partition reference in the environmental map, the location of the partition reference in the environmental map is determined by comparing the number of times candidate locations are marked on the environmental map with a set threshold.

[0071] Continue using the above Figure 5 The environmental map shown illustrates this. The thick solid lines corresponding to regions 0, 1-4, and 6 represent potential candidate locations for doors. The number of thick solid lines indicates the frequency of each candidate location being marked. Statistical analysis of the marked candidate locations on the environmental map reveals that regions 0, 1-4, and 6 have a relatively high number of marked candidate locations. Therefore, the average value of candidate locations within the same location range in each region can be calculated. This average value is taken as the final location of the door within that location range on the environmental map and marked accordingly. In other words, the gaps corresponding to regions 0, 1-4, and 6 represent the final location of the marked door on the environmental map.

[0072] Based on the above embodiments, optionally, an image recognition model can be used to automatically and more accurately select the target environment image containing the partition reference from multiple environment images.

[0073] Specifically, an image recognition model is pre-trained using sample images containing partition references. Image recognition models include, but are not limited to: CNN (Convolutional Neural Networks) models, RNN (Recurrent Neural Network) models, and LSTM (Long Short-Term Memory) models.

[0074] Furthermore, to improve the accuracy of identifying partitioned reference objects and reduce their false recognition rate, for any given environment image from multiple environment images, the image recognition model first extracts multiple candidate detection boxes. Then, it extracts the image features of each candidate detection box and inputs these features into a multi-class SVM (Support Vector Machine) classifier to determine the probability value of the target object within the candidate detection box belonging to each class. One SVM classifier is trained for each class. If the probability value of the target object within the candidate detection box belonging to a partitioned reference object meets the criteria for a partitioned reference object, then the target object within the candidate detection box is determined to be a partitioned reference object, and the environment image of that object is identified as the target environment image. Additionally, the image recognition model also has a bounding box correction function, which can correct candidate detection boxes containing partitioned reference objects, improving the accuracy of partitioned reference object localization.

[0075] or

[0076] Furthermore, to improve the accuracy of identifying partitioned reference objects and reduce the false recognition rate, the image recognition model can also be a keypoint-based object detection model. Preferably, the object detection model is a center point regression-based model. This model models the target object as the center point of the bounding box, and then directly regresses the other attributes of the target object. Specifically, the model finds the center point of the target object through keypoint estimation, and then regresses all other attributes of the target object, such as size, 3D position, orientation, and pose. Simultaneously, the model can also predict the target's category.

[0077] When training a centroid regression-based target detection model, a large number of images of the work environment can be collected in advance, and all target objects can be manually labeled with category tags (such as bed, sofa, dining table, clothes rack, etc.). Then, various parameter adjustments can be performed to finally determine a centroid regression-based target detection model with strong generalization ability and stable performance.

[0078] When performing image recognition using a center-point regression-based target detection model, for any given environment image from multiple environment images, the image features of that environment image are first extracted to obtain a feature map. Next, a single feature point on the feature map is used as the center point of the target object. The size of the target object is regressed based on its center point, and the category of the target object is predicted based on the center point. If the probability value of the target object belonging to a partition reference satisfies the partition reference judgment condition, then the target object is determined to be a partition reference, and the environment image containing the target object is designated as the target environment image.

[0079] When using an image recognition model, the probability value of the target object belonging to the partition reference object, as determined by the above-mentioned criteria, is greater than the probability threshold in the image recognition model. For example, the probability threshold used by the image recognition model is 0.9. In an optional embodiment, the probability threshold used by the image recognition model can be calibrated during training based on a large number of experimental samples. Because the environmental images acquired in this embodiment are of high quality, the probability threshold used by the image recognition model is relatively high, resulting in higher recognition accuracy. A process for calibrating the probability threshold during model training is as follows:

[0080] It can control autonomous mobile devices to traverse different working environments, collect environmental images, and train an image recognition model based on a large number of environmental images. During training, the probability threshold in the image recognition model is continuously adjusted until the probability threshold enables the image recognition model to achieve high accuracy in recognizing zoning reference objects in different working environments.

[0081] For example, before the autonomous mobile device leaves the factory, an image recognition model is trained using a massive amount of environmental images. After the autonomous mobile device leaves the factory, the probability threshold in the image recognition model remains fixed, enabling it to accurately identify zoning reference objects in different working environments.

[0082] In this embodiment, the recognition accuracy of the image recognition model is directly related to the image quality and the value of the probability threshold used in the image recognition model to determine whether the image includes a partition reference object. It is understood that if the environmental image quality is poor, the probability threshold used by the image recognition model to determine whether a target object in the detection box belongs to a partition reference object cannot be very high. If the probability threshold is set too high, misjudgments may occur, misidentifying objects that belong to partition reference objects as not belonging to them, increasing the likelihood of missed detections of partition reference objects. Conversely, if the environmental image quality is good, the probability threshold used by the image recognition model to determine whether a target object in the detection box belongs to a partition reference object can be higher. If the probability threshold is set too low, misjudgments may also occur, misidentifying objects that do not belong to partition reference objects as belonging to them, increasing the likelihood of misidentification of partition reference objects.

[0083] Therefore, in some embodiments of this application, the probability threshold in the image recognition model can be dynamically adjusted to enable the image recognition model to accurately identify partition reference objects in different working environments. Specifically, to improve the recognition accuracy of the image recognition model, before inputting multiple environmental images into the image recognition model for recognition, the method may further include: acquiring image quality information of multiple environmental images, wherein the image quality information includes at least: image sharpness and / or the number of images containing frontal features of partition reference objects; and adjusting the probability threshold in the image recognition model used to determine whether an image contains partition reference objects based on the image quality information of the multiple environmental images.

[0084] In this embodiment, an adjustment strategy for the probability threshold can be formulated according to the actual situation. As an example, the higher the image quality level, the higher the probability threshold; conversely, the lower the image quality level, the lower the probability threshold. When the probability threshold is set to a certain value, the probability threshold corresponding to a low image quality level is less than the probability threshold corresponding to a high image quality level. For example, the probability threshold for a low image quality level is 0.3, and the probability threshold for a high image quality level is 0.5.

[0085] Specifically, the image quality level of multiple environmental images can be determined manually based on two dimensions: image sharpness and the number of images including frontal features of the partitioned reference objects. If most of the environmental images have good image sharpness and / or most of the environmental images include frontal features of the partitioned reference objects, then the image quality level of the multiple environmental images is high; if most of the environmental images have poor image sharpness and / or most of the environmental images do not include frontal features of the partitioned reference objects, then the image quality level of the multiple environmental images is low.

[0086] Based on the above embodiments, optionally, the threshold number for determining whether to remove candidate locations in the environmental map can be adjusted according to the image quality information of multiple environmental images. The higher the image quality level of the multiple environmental images, the higher the threshold number; the lower the image quality level, the lower the threshold number. The correspondence between image quality level and threshold number is set according to the actual situation. For example, a threshold number of 5 corresponds to a low image quality level, and a threshold number of 15 corresponds to a high image quality level.

[0087] Based on the above embodiments, optionally, step 103 can be implemented as follows: according to the position of the partition reference in the environment map, the area behind each partition reference is divided into a partition, and the remaining area is divided into a partition.

[0088] Continue Figure 5The environment map is shown after partitioning. The environment map is divided into 6 regions. The gaps in regions 0, 1-4, and 6 are the locations of doors. The area behind each door is a separate region, and the remaining area is region 5.

[0089] In some optional embodiments of this application, the threshold number of times a partition reference is set to determine whether a candidate location in the environmental map actually contains a partition reference may include an upper limit and a lower limit. Based on this, a partitioning method includes: first, controlling an autonomous mobile device to traverse the operating environment using a trajectory mode W1, and acquiring environmental images during the traversal using trajectory mode W1; selecting a target environmental image containing a partition reference from the acquired environmental images; and marking candidate locations where the partition reference exists in the environmental map according to the position of the partition reference in the target environmental image.

[0090] For each candidate location in the environment map, if the number of times the candidate location is marked is less than or equal to the lower limit, it is determined that the candidate location does not have a partition reference. If the number of times the candidate location is marked is greater than or equal to the upper limit, it is determined that the candidate location has a partition reference. If the number of times the candidate location is marked is greater than the lower limit but less than the upper limit, it can be further verified whether the candidate location has a partition reference.

[0091] Specifically, the autonomous mobile device can be controlled to acquire multiple environmental images around the candidate location using other trajectory modes W2. The number of times the candidate location is marked can be determined again using the multiple environmental images around the candidate location acquired by trajectory mode W2. If the number of times the candidate location is marked again is less than or equal to the lower limit, it is determined that there is no partition reference at the candidate location; if the number of times the candidate location is marked again is greater than or equal to the upper limit, it is determined that there is a partition reference at the candidate location.

[0092] The trajectory pattern W2 can be one or more. When selecting trajectory pattern W2, the orientation information of the candidate location can be determined based on the environment map, and a trajectory pattern whose main travel direction is opposite to the orientation information can be selected based on the orientation information of the candidate location; or, a trajectory pattern whose main travel direction is approximately opposite to the orientation information can be selected.

[0093] In one optional embodiment, a trajectory pattern can be randomly selected as trajectory pattern W1, or a default trajectory pattern can be selected as trajectory pattern W1, or a trajectory pattern with a smaller (e.g., the smallest) trajectory interval can be selected as trajectory pattern W1.

[0094] Then, based on the determined partition references and their positions on the environment map, the areas on both sides of the partition references in the environment map are separated using the partition references as dividing lines.

[0095] Based on the above embodiments, optionally, at least two trajectory patterns include: a horizontal bow-shaped trajectory pattern and a vertical bow-shaped trajectory pattern, wherein the main travel directions of the horizontal bow-shaped trajectory pattern and the vertical bow-shaped trajectory pattern are perpendicular to each other, and the main travel direction is the direction of the long side trajectory in the bow-shaped trajectory.

[0096] In this embodiment, using the horizontal bow-shaped trajectory pattern and the vertical bow-shaped trajectory pattern can achieve more regular traversal in the working environment, which is beneficial for obtaining richer environmental information.

[0097] Continue Figure 4a The staggered "bow" shaped cleaning path planning shown has a horizontal bow-shaped trajectory pattern for cleaning path A and a vertical bow-shaped trajectory pattern for cleaning path B.

[0098] In different application scenarios, autonomous mobile devices may use at least two trajectory modes to traverse the operating environment in different ways.

[0099] In the first application scenario, when the autonomous mobile device traverses the work environment using at least two trajectory modes, it sequentially employs both modes while capturing environmental images during the traversal. Taking a horizontal bow-shaped trajectory mode and a vertical bow-shaped trajectory mode as an example, the autonomous mobile device first uses the horizontal bow-shaped trajectory mode to traverse the entire work environment, capturing environmental images during the process; then it uses the vertical bow-shaped trajectory mode to traverse the entire work environment, capturing environmental images during the same process. In this application scenario, the autonomous mobile device traverses the work environment comprehensively, but the amount of information collected is large, resulting in a longer information processing time and higher resource consumption due to the extended time spent traversing the work environment.

[0100] In the second application scenario, when the autonomous mobile device traverses the work environment using at least two trajectory modes, it first uses one trajectory mode to traverse the entire work environment, acquiring environmental images and performing image recognition. If the recognition results indicate that the probability of a local reference object is not high, the location range of the local area is identified, and another trajectory mode is used to traverse the local area separately, acquiring environmental images within the location range of the local area during the traversal. It is understandable that if a local reference object truly exists, switching the trajectory mode increases the probability of acquiring a frontal image of the reference object in that local area, thereby increasing the probability of identifying the reference object in the local area in the recognition results. This ensures improved accuracy in identifying reference objects in the work environment while reducing the acquisition of unnecessary information, shortening information processing time, and reducing the time spent traversing the work environment with minimal resource consumption.

[0101] Autonomous mobile devices can employ at least two different trajectory modes to perform tasks, traversing the work environment and collecting environmental images during the process. They can also partition the environment map based on the positions of zoning reference points identified from the collected environmental images. Furthermore, the tasks performed will vary depending on the device's implementation. For example, a robotic vacuum cleaner performs cleaning tasks, a disinfection robot performs disinfection tasks, and a guidance robot acts as a guide, explaining detailed information about the area.

[0102] The following example illustrates how an autonomous mobile robot operates in a work environment using different trajectory modes. Specifically, the application scenario involves a robotic vacuum cleaner performing cleaning tasks in a work area using at least two trajectory modes.

[0103] Figure 6 This is a schematic flowchart illustrating a cleaning method provided in an exemplary embodiment of this application. Figure 6 As shown, the method includes the following steps:

[0104] Step 601: Control the sweeping robot to perform the cleaning task in the work area using the first trajectory mode.

[0105] Step 602: After completing the cleaning task using the first trajectory mode, control the robot vacuum cleaner to continue to perform the cleaning task in the work area again using the second trajectory mode; wherein, the main travel direction of the first trajectory mode and the second trajectory mode are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0106] Currently, the most common method for robotic vacuum cleaners to perform cleaning tasks is through random collision-based path planning. In this trajectory mode, the robot explores its path by constantly colliding with its surroundings, increasing the cleaning coverage by accumulating cleaning cycles. However, this trajectory mode is prone to chaotic cleaning and missed areas. Furthermore, its ability to acquire environmental information is limited; when map partitioning is required using environmental information, insufficient information can lead to inaccurate partitioning.

[0107] In this embodiment, the robotic vacuum cleaner will clean the work area multiple times according to a regular cleaning path based on the first trajectory mode and the second trajectory mode. The first trajectory mode and the second trajectory mode can be found in the description of trajectory modes in the above embodiments. The first trajectory mode and the second trajectory mode can correspond to the same operation mode. For example, the first trajectory mode and the second trajectory mode can be two different trajectory modes in an edge cleaning mode; or, for example, the first trajectory mode and the second trajectory mode can be two different trajectory modes in a spot cleaning mode. Alternatively, different operation modes can be defined for different trajectory modes; for example, the operation mode executed using the first trajectory mode can be called the first operation mode, and the operation mode executed using the second trajectory mode can be called the second operation mode, and so on.

[0108] For ease of understanding, we will use the first trajectory pattern as a horizontal bow-shaped trajectory pattern and the second trajectory pattern as a vertical bow-shaped trajectory pattern as examples. The main directions of travel in the horizontal and vertical bow-shaped trajectory patterns are perpendicular to each other, and the main direction of travel is the direction of the longer side of the bow-shaped trajectory.

[0109] Continue Figure 4a The illustrated staggered "bow"-shaped cleaning path planning, in the first trajectory mode, when the robot vacuum cleaner follows cleaning path A to perform a regular "bow"-shaped cleaning in the work area, can capture images of targets F and D from the front, but may capture images of targets C and E from the side or not. In the second trajectory mode, when the robot vacuum cleaner follows cleaning path B to perform a regular "bow"-shaped cleaning in the work area, it can capture images of targets C and E from the front, but may capture images of targets F and D from the side or not.

[0110] Clearly, when cleaning the work area based on different trajectory patterns, more objects in the work area can be captured without being limited by the field of view of the visual sensor.

[0111] The cleaning method provided in this embodiment enables the robotic vacuum cleaner to clean the work area according to a regular cleaning path, effectively improving cleaning coverage and reducing missed areas. Furthermore, cleaning the work area based on different trajectory patterns minimizes the limitations of the visual sensor's field of view, thereby acquiring richer environmental information about the work area and improving the accuracy of map zoning.

[0112] In this embodiment, there are no restrictions on how the robot vacuum cleaner switches its trajectory mode.

[0113] As an example, regardless of the trajectory mode used, the robot vacuum cleaner traverses the working environment from the same starting position. That is, it first controls the robot vacuum cleaner to start from the designated starting position and performs the cleaning task in the working area using the first trajectory mode; then, it controls the robot vacuum cleaner to return to the designated starting position and starts from the designated starting position again to perform the cleaning task in the working area using the second trajectory mode.

[0114] In this embodiment, unifying the starting point position of each trajectory pattern can effectively improve cleaning coverage and reduce missed areas.

[0115] Continue Figure 4a The illustrated staggered "bow"-shaped cleaning path planning works as follows: When the robot vacuum moves along cleaning path A from the upper left corner (the designated starting position) to the lower left corner of the work area, it completes the cleaning task using the first trajectory mode. At this point, the robot vacuum moves in a straight line from the lower left corner back to the upper left corner (returning to the designated starting position). Next, the robot vacuum moves along the cleaning path from the upper left corner (the designated starting position) to the lower right corner of the work area, completing the cleaning task using the second trajectory mode. Then, the robot vacuum moves in a straight line from the lower right corner back to the upper left corner (returning to the designated starting position).

[0116] As another example, you can first control the robot vacuum to perform a cleaning task in the work area using the first trajectory mode; after completing the cleaning task using the first trajectory mode, you can control the robot vacuum to start from the current position and continue to perform the cleaning task in the work area again using the second trajectory mode.

[0117] It should be noted that before controlling the robot vacuum to resume cleaning in the work area using the second trajectory mode from its current position, the robot vacuum's current direction of travel needs to be adjusted to be the same as the main direction of travel supported by the second trajectory mode.

[0118] In practice, user needs directly influence the cleaning trajectory patterns of a robotic vacuum cleaner. For example, when a robotic vacuum cleaner performs its first cleaning task in a home, it needs to create a floor plan. To accurately divide the floor plan into zones, the robot needs to use at least two trajectory patterns to perform the cleaning task. However, when the robotic vacuum cleaner performs its second cleaning task, since there's no need to create a floor plan or divide the floor plan into zones, it can use only one trajectory pattern. Similarly, if the user requires a deep clean, the robotic vacuum cleaner can use two trajectory patterns. It's understandable that, based on user needs, a robotic vacuum cleaner can use at least two trajectory patterns or only one trajectory pattern to perform a cleaning task.

[0119] Therefore, based on the above embodiments, optionally, before step 601, at least one of the following judgment operations may be included: determining whether a cleaning task is being performed in the work area for the first time; determining whether a work instruction indicating to perform the cleaning task using two trajectory modes has been received; if the result of at least one judgment operation is yes, then the robot vacuum cleaner is controlled to perform the cleaning task in the work area in sequence using the first trajectory mode and the second trajectory mode.

[0120] To better understand the cleaning method of this application, two scenario embodiments are described below:

[0121] Scenario Example 1

[0122] Taking a home environment as an example, when a user purchases a robot vacuum cleaner, the robot vacuum cleaner supports... Figure 4a The two trajectory modes are shown. When the robot vacuum cleaner is used for the first time to perform a cleaning task in the home, in the first trajectory mode, the robot vacuum cleaner performs a regular "bow"-shaped cleaning pattern in the home environment according to cleaning path A, and collects environmental images of the home environment. Then, in the second trajectory mode, the robot vacuum cleaner performs a regular "bow"-shaped cleaning pattern in the home environment according to cleaning path B, and collects environmental images of the home environment.

[0123] During or after a cleaning task, environmental images of the home environment can be collected to create a home environmental map. Optionally, the created environmental map can be a semantic map. For information on creating a semantic map, please refer to [link to documentation / reference]. Figure 7 The illustrated embodiment will not be described in detail here. Furthermore, the semantic map can be partitioned; the partitioning process can be implemented using… Figure 1 The method described in the illustrated embodiments can also be adopted using the following methods. Figure 7 The method described in the illustrated embodiment.

[0124] After constructing a semantic map with partitions, users can use a robot vacuum to clean specific areas. For example, they can display the semantic map on their mobile phone, set the target area to be cleaned on the semantic map, such as the living room or balcony, and issue cleaning commands to the robot vacuum via their mobile phone. During subsequent cleaning sessions, the robot vacuum uses a trajectory mode to perform cleaning tasks in the target area. The trajectory mode used by the robot vacuum can be a default trajectory mode or a trajectory mode specified by the user. For example, during subsequent cleaning sessions, the robot vacuum can use the first trajectory mode to perform regular "bow"-shaped cleaning in the target area according to cleaning path A; or it can use the second trajectory mode to perform regular "bow"-shaped cleaning in the target area according to cleaning path B.

[0125] Scenario Example 2:

[0126] Following on from Scenario 1 above, in cases involving non-initial cleaning tasks, if deep cleaning is required, the user can issue a command via their mobile phone to execute the cleaning task using two trajectory modes. Upon receiving the command, the robot vacuum cleaner, in the first trajectory mode, performs a regular "bow"-shaped cleaning motion within the home environment following cleaning path A. Then, in the second trajectory mode, the robot vacuum cleaner performs a regular "bow"-shaped cleaning motion within the home environment following cleaning path B. Using two trajectory modes along different main directions of travel reduces the probability of missed areas, increases cleaning coverage, and meets the needs of deep cleaning.

[0127] Figure 7 This is a flowchart illustrating a semantic map construction method provided as an exemplary embodiment of this application. Figure 7 As shown, the method includes the following steps:

[0128] Step 701: Control the autonomous mobile device to traverse the operating environment using at least two trajectory modes.

[0129] Step 702: Acquire multiple environmental images during the traversal process using the at least two trajectory modes.

[0130] Step 703: Based on the multiple environmental images, construct a semantic map of the working environment; at least two trajectory modes have different main directions of travel, and the main direction of travel refers to the direction of travel that meets the set conditions.

[0131] The implementation methods and beneficial effects of steps 701 and 702 in this embodiment can be referred to step 101 in the above embodiment, and will not be repeated here.

[0132] In this embodiment, because at least two traversals are performed and the main travel directions of the two traversals are different, the field of view coverage of the visual sensor will be different. Therefore, richer environmental images of the working environment can be collected. For example, environmental images from multiple perspectives may be collected for the same object. Based on this, semantic maps are constructed from the objects contained in the more information-rich environmental images, which can more accurately describe the working environment. The semantic map has higher precision and accuracy.

[0133] In an optional embodiment of this application, step 703 can be implemented as follows: based on the multiple environmental images, identify objects and their semantic information existing in the work environment; based on the objects and their semantic information existing in the work environment, determine at least two partitions and their semantic information contained in the work environment; based on the at least two partitions and their semantic information, construct a semantic map of the work environment.

[0134] The objects in the work environment can be any object within the work environment. For example, in a home environment, objects include, but are not limited to, doors, televisions, refrigerators, range hoods, toilets, clothes racks, beds, sofas, and coffee tables. Optionally, a trained image recognition model can be used to identify the objects and their semantic information in the work environment. Image recognition models include, but are not limited to, CNN models, RNN models, and LSTM models. Furthermore, to improve the accuracy of target object recognition, for any given environmental image, the image recognition model first extracts multiple candidate detection boxes. Then, the model extracts the image features of each candidate detection box, and then inputs these features into a multi-class SVM classifier to determine the objects contained within the candidate detection boxes and their semantic information. One SVM classifier is trained for each class. Additionally, the image recognition model also has a bounding box correction function, which can correct candidate detection boxes containing target objects, improving the accuracy of target object localization.

[0135] After identifying the objects and their semantic information in the work environment, at least two partitions and their semantic information within the work environment can be determined based on these objects and semantic information. One exemplary implementation includes: dividing the work environment into at least two partitions based on partition reference objects and their semantic information; and determining the at least two partition semantic information based on non-partition reference objects and their semantic information contained within the at least two partitions.

[0136] Taking a home environment as an example, objects within a home environment include, but are not limited to: doors to each room, televisions, refrigerators, range hoods, toilets, clothes racks, beds, sofas, coffee tables, etc. Doors in a home environment serve as partitioning references, dividing the home environment into different zones. Conversely, refrigerators, range hoods, toilets, clothes racks, beds, sofas, coffee tables, etc., do not serve a partitioning function and can be considered non-partitioning references. For each zone, the objects included will differ depending on the function it serves. For example, bedrooms are generally for rest, so they typically include beds; kitchens are for cooking, so they usually include refrigerators, range hoods, etc.; and living rooms are generally for entertaining guests or leisure, so they usually include sofas, coffee tables, and televisions. Based on this, for each zone, the semantic information of the zone can be determined based on the objects it contains and their semantic information, i.e., whether the zone is a living room, bedroom, kitchen, or bathroom, etc. If a zone includes a television, refrigerator, and sofa, then that zone is a living room. If a zone includes a bed, then that zone is a bedroom. If a zone includes a refrigerator and a range hood, then that zone is a kitchen. If a zone includes a clothes rack, then that zone is a balcony. If a zone includes a toilet, then that zone is a bathroom. Thus, a family floor plan is divided into zones such as bedroom, living room, kitchen, balcony, and bathroom.

[0137] After obtaining the partitions and their semantic information contained in the working environment, as well as the objects contained in each partition and their semantic information, a semantic map can be created. This semantic map includes the partitions existing in the working environment and their semantic information, and may further include the objects contained in each partition and their semantic information.

[0138] Figure 8 This is a schematic flowchart illustrating an object recognition method provided as an exemplary embodiment of this application. Figure 8 As shown, the method includes the following steps:

[0139] Step 801: Control the autonomous mobile device to traverse the operating environment using at least two trajectory modes.

[0140] Step 802: Acquire multiple environmental images during the traversal process using the at least two trajectory modes.

[0141] Step 803: Based on the multiple environmental images, identify the target objects present in the working environment; at least two trajectory modes have different main travel directions.

[0142] The implementation methods and beneficial effects of steps 801 and 802 in this embodiment can refer to step 101 in the above embodiment, and will not be repeated here.

[0143] In this embodiment, because at least two traversals are performed and the main travel directions of the two traversals are different, the field of view coverage of the visual sensor will be different. Therefore, richer environmental images in the working environment can be collected. For example, environmental images from multiple perspectives may be collected for the same object. Based on this, objects in the working environment can be identified more accurately and more objects in the working environment can be identified. This improves the ability of the autonomous mobile device to acquire environmental information, which is beneficial for obstacle avoidance and positioning of the autonomous mobile device, and also helps to improve the accuracy of the environmental map and its partitions.

[0144] The target object in the work environment can be any object in the work environment. For example, in a home environment, objects in the environment include, but are not limited to: doors, televisions, refrigerators, range hoods, toilets, clothes racks, beds, sofas, coffee tables, etc.

[0145] After acquiring multiple environmental images, a trained image recognition model can be used to identify target objects in the work environment. The image recognition model can identify the object's category and semantic information, or only the object's category. Image recognition models include, but are not limited to, CNN models, RNN models, and LSTM models. Furthermore, to improve the accuracy of target object recognition, the image recognition model first extracts multiple candidate detection boxes from the environmental image. Then, the model extracts the image features of each candidate detection box, and then inputs these features into a multi-class SVM classifier to determine the probability value of the target object contained in the candidate detection box belonging to each category. The category with the highest probability value is identified as the final category of the target object. Each category has its own SVM classifier. In addition, the image recognition model also has a bounding box correction function, which can correct the candidate detection boxes containing the target object, improving the accuracy of target object localization.

[0146] It should be noted that the execution subject of each step of the method provided in the above embodiments can be the same device, or the method can be executed by different devices. For example, the execution subject of steps 101 to 103 can be device A; or the execution subject of steps 101 and 102 can be device A, and the execution subject of step 103 can be device B; and so on.

[0147] Furthermore, some processes described in the above embodiments and accompanying drawings include multiple operations appearing in a specific order. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or they may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not represent any execution order. Additionally, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first" and "second" in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to different types.

[0148] Figure 9 This is a schematic diagram of a map partitioning device provided for an exemplary embodiment of this application. Figure 9 As shown, the device includes a control module 91, a data acquisition module 92, a tagging module 93, and a partitioning module 94.

[0149] Control module 91 is used to control the autonomous mobile device to traverse the operating environment using at least two trajectory modes;

[0150] Acquisition module 92 is used to acquire multiple environmental images during the traversal process;

[0151] The marking module 93 is used to mark the position of the partition reference in the environmental map corresponding to the work environment based on the number of times and position of the partition reference in the multiple environmental images;

[0152] The partitioning module 94 is used to partition the environment map according to the position of the partitioning reference in the environment map;

[0153] Among them, the main travel direction of the at least two trajectory modes is different, and the main travel direction refers to the travel direction that meets the set conditions.

[0154] Furthermore, the control module 91 is specifically used for:

[0155] After the autonomous mobile device has traversed the working environment using the previous trajectory mode, control the autonomous mobile device to return to the designated starting position and continue to traverse the working environment again using the next trajectory mode;

[0156] or

[0157] After the autonomous mobile device has traversed the working environment using the previous trajectory mode, it is controlled to start from the current position and continue to traverse the working environment again using the next trajectory mode.

[0158] Furthermore, the marking module 93 includes a selection unit, a marking unit, and a filtering unit:

[0159] The selection unit is used to select a target environment image containing the partition reference from the plurality of environment images;

[0160] A marking unit is used to mark candidate locations where the partition reference is present in the environment map based on the location of the partition reference in the target environment image;

[0161] The filtering unit is used to remove candidate locations whose number of markings is less than a set threshold, so as to obtain the location of the partition reference in the environment map.

[0162] Furthermore, the selection unit is specifically used for:

[0163] The multiple environmental images are input into an image recognition model for recognition to obtain a target environmental image including the partition reference; wherein, the image recognition model is pre-trained using sample images containing the partition reference.

[0164] Furthermore, before inputting the multiple environmental images into the image recognition model for recognition, the selection unit is also used to:

[0165] Obtain image quality information of the multiple environmental images, wherein the image quality information includes at least: image sharpness and / or the number of images including the frontal features of the partition reference;

[0166] Based on the image quality information of the multiple environmental images, the probability threshold used in the image recognition model to determine whether an image contains a partition reference object is adjusted.

[0167] Furthermore, partition module 94 is specifically used for:

[0168] Based on the location of the partition reference points in the environment map, the area behind each partition reference point is divided into a partition, and the remaining area is divided into a partition.

[0169] Furthermore, the at least two trajectory patterns include: a horizontal bow-shaped trajectory pattern and a vertical bow-shaped trajectory pattern, wherein the main travel directions of the horizontal bow-shaped trajectory pattern and the vertical bow-shaped trajectory pattern are perpendicular to each other, and the main travel direction is the direction of the longer side of the bow-shaped trajectory.

[0170] The map partitioning device provided in this application embodiment can cover the entire working environment as much as possible by using multiple trajectory modes with different main directions of travel on the autonomous mobile device. This allows for capturing as many partitioning reference objects in the working environment as possible and increases the probability of capturing frontal angle images of the partitioning reference objects. This is beneficial for more accurately identifying the partitioning reference objects and their locations in the working environment and improving the accuracy of map partitioning based on partitioning reference objects.

[0171] Figure 10 This is a schematic diagram of a cleaning device provided for an exemplary embodiment of this application. (See attached diagram.) Figure 10 As shown, the device includes a first control module 101 and a second control module 102.

[0172] The first control module 101 is used to control the sweeping robot to perform cleaning tasks in the work area using the first trajectory mode;

[0173] The second control module 102 is used to control the sweeping robot to continue to perform the sweeping task in the work area again using the second trajectory mode after completing the sweeping task using the first trajectory mode; wherein, the main travel direction of the first trajectory mode and the second trajectory mode are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0174] Furthermore, the first control module 101 is used to control the sweeping robot to perform cleaning tasks in the work area starting from a designated starting position and using a first trajectory mode;

[0175] Accordingly, the second control module 102 is used to control the sweeping robot to return to the designated starting position and to control the sweeping robot to resume the cleaning task in the work area from the designated starting position using the second trajectory mode.

[0176] Furthermore, the second control module 102 is used to control the sweeping robot to continue to perform the cleaning task in the work area from its current position using the second trajectory mode.

[0177] The cleaning device provided in this embodiment enables the robotic vacuum cleaner to clean the work area according to a regular cleaning path, effectively improving cleaning coverage and reducing missed areas. Furthermore, cleaning the work area based on different trajectory patterns minimizes the limitations of the visual sensor's field of view, thereby acquiring richer environmental information about the work area and improving the accuracy of map zoning.

[0178] Figure 11 This is a schematic diagram of the structure of an autonomous mobile device provided as an exemplary embodiment of this application. Figure 11As shown, the autonomous mobile device 100 includes a device body 110, on which a memory 11 and a processor 12 are provided.

[0179] Memory 11 is used to store computer programs and can be configured to store various other data to support operation on the processor. Examples of this data include instructions, messages, images, videos, etc., for any application or method that operates on the processor.

[0180] The memory 11 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0181] Processor 12, coupled to memory 11, is used to execute computer programs in memory 11 for:

[0182] Control the autonomous mobile device to traverse the operating environment using at least two trajectory modes, and acquire multiple environmental images during the traversal using at least two trajectory modes;

[0183] Based on the frequency and location of the zoning reference objects existing in the work environment in the multiple environmental images, mark the location of the zoning reference objects in the environmental map corresponding to the work environment;

[0184] The environment map is divided into zones based on the location of the zone reference points in the environment map; wherein, the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0185] Furthermore, when the processor 12 controls the autonomous mobile device to traverse the operating environment, it is specifically used for:

[0186] After the autonomous mobile device has traversed the working environment using the previous trajectory mode, control the autonomous mobile device to return to the designated starting position and continue to traverse the working environment again using the next trajectory mode;

[0187] or

[0188] After the autonomous mobile device has traversed the working environment using the previous trajectory mode, it is controlled to start from the current position and continue to traverse the working environment again using the next trajectory mode.

[0189] Furthermore, when marking the location of the partition reference, the processor 12 specifically performs the following functions:

[0190] Select the target environment image containing the partition reference from the plurality of environment images;

[0191] Based on the position of the partition reference in the target environment image, candidate locations where the partition reference exists are marked in the environment map;

[0192] Candidate locations that have been marked less than a set threshold number are removed to obtain the location of the partition reference in the environment map.

[0193] Furthermore, when selecting a target environment image containing the partition reference, the processor 12 specifically performs the following:

[0194] The multiple environmental images are input into an image recognition model for recognition to obtain a target environmental image including the partition reference; wherein, the image recognition model is pre-trained using sample images containing the partition reference.

[0195] Furthermore, before inputting the multiple environmental images into the image recognition model for recognition, the processor 12 is also used to:

[0196] Obtain image quality information of the multiple environmental images, wherein the image quality information includes at least: image sharpness and / or the number of images including the frontal features of the partition reference;

[0197] Based on the image quality information of the multiple environmental images, the probability threshold used in the image recognition model to determine whether an image contains a partition reference object is adjusted.

[0198] Furthermore, when partitioning the environment map, the processor 12 specifically performs the following functions:

[0199] Based on the location of the partition reference points in the environment map, the area behind each partition reference point is divided into a partition, and the remaining area is divided into a partition.

[0200] Furthermore, the at least two trajectory patterns include: a horizontal bow-shaped trajectory pattern and a vertical bow-shaped trajectory pattern, wherein the main travel directions of the horizontal bow-shaped trajectory pattern and the vertical bow-shaped trajectory pattern are perpendicular to each other, and the main travel direction is the direction of the longer side of the bow-shaped trajectory.

[0201] Figure 11 The autonomous mobile device shown can execute the methods of the above embodiments. For parts not described in detail in this embodiment, please refer to the relevant descriptions of the above embodiments. The execution process and technical effects of this technical solution are described in the above embodiments and will not be repeated here.

[0202] Furthermore, such as Figure 11As shown, the autonomous mobile device also includes other components such as a communication component 13, a display 14, a power supply component 15, and an audio component 16. Figure 11 The diagram only shows some components and does not mean that the processor only includes... Figure 11 The components shown. Additionally... Figure 11 The components shown in the dashed box are optional, not mandatory, and their specific requirements depend on the implementation of the partitioning device.

[0203] In an alternative embodiment, Figure 11 The autonomous mobile device shown can be implemented as a robotic vacuum cleaner, which can then perform... Figure 6 For detailed descriptions of the steps in the cleaning method provided in the illustrated embodiment, please refer to the relevant descriptions of the above embodiments, which will not be repeated here. Furthermore, when the autonomous mobile device is implemented as a robotic vacuum cleaner, the device body is also equipped with cleaning components (such as floor brushes, dust collection boxes, etc.), anti-collision components (such as anti-collision plates located at the front of the robot), laser sensors, and infrared sensors for infrared recharging, among other sensors.

[0204] The above Figure 11 The communication component is configured to facilitate wired or wireless communication between the device containing the communication component and other devices. The device containing the communication component can access wireless networks based on communication standards, such as WiFi, 2G, 3G, 4G / LTE, 5G, or combinations thereof. In one exemplary embodiment, the communication component receives broadcast signals or broadcast-related information from an external broadcast management system via a broadcast channel. In one exemplary embodiment, the communication component further includes a Near Field Communication (NFC) module to facilitate short-range communication. For example, the NFC module may be implemented based on Radio Frequency Identification (RFID), Infrared Data Association (IrDA) technology, Ultra-Wideband (UWB) technology, Bluetooth (BT) technology, and other technologies.

[0205] The above Figure 11 The display includes a screen, which may include a liquid crystal display (LCD) and a touch panel (TP). If the screen includes a touch panel, the screen can be implemented as a touchscreen to receive input signals from the user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of the touch or swipe action, but also the duration and pressure associated with the touch or swipe operation.

[0206] The above Figure 11 The power supply component provides power to the various components of the device in which it resides. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device in which it resides.

[0207] The above Figure 11 The audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC) configured to receive external audio signals when the device containing the audio component is in an operating mode, such as call mode, recording mode, or voice recognition mode. The received audio signals can be further stored in memory or transmitted via a communication component. In some embodiments, the audio component also includes a speaker for outputting audio signals.

[0208] Accordingly, embodiments of this application also provide a computer-readable storage medium storing a computer program, which, when executed, can perform the steps that can be executed by the processor 11 in the above method embodiments.

[0209] An exemplary embodiment of this application also provides an autonomous mobile device, the structure of which is... Figure 11 The structure of the autonomous mobile device shown is the same, and the computer programs stored in its memory and executed by its processor are identical. Figure 11 The autonomous mobile device shown is different. Specifically, this autonomous mobile device includes: a device body, on which a memory and a processor are provided; the memory is used to store computer programs; the processor, coupled to the memory, is used to execute the computer programs for: controlling the autonomous mobile device to traverse the working environment using at least two trajectory modes; acquiring multiple environmental images during the traversal using the at least two trajectory modes; constructing a semantic map of the working environment based on the multiple environmental images; the at least two trajectory modes have different main travel directions, where the main travel direction refers to the travel direction that meets set conditions.

[0210] An exemplary embodiment of this application also provides an autonomous mobile device, the structure of which is... Figure 11 The structure of the autonomous mobile device shown is the same, and the computer programs stored in its memory and executed by its processor are identical. Figure 11 The autonomous mobile device shown is different. Specifically, this autonomous mobile device includes: a device body, on which a memory and a processor are provided; the memory is used to store computer programs; the processor, coupled to the memory, is used to execute the computer programs for: controlling the autonomous mobile device to traverse the working environment using at least two trajectory modes; acquiring multiple environmental images during the traversal using the at least two trajectory modes; and identifying target objects present in the working environment based on the multiple environmental images; wherein the primary travel directions of the at least two trajectory modes are different, and the primary travel direction refers to the travel direction that meets set conditions.

[0211] An exemplary embodiment of this application also provides an autonomous mobile device, the structure of which is... Figure 11 The structure of the autonomous mobile device shown is the same, and the computer programs stored in its memory and executed by its processor are identical. Figure 11 The autonomous mobile device shown is different. Specifically, the autonomous mobile device includes: controlling the robot vacuum to perform a cleaning task in the work area using a first trajectory mode; after completing the cleaning task using the first trajectory mode, controlling the robot vacuum to continue to perform the cleaning task again in the work area using a second trajectory mode; wherein, the main travel direction of the first trajectory mode and the second trajectory mode are different, and the main travel direction refers to the travel direction that meets the set conditions.

[0212] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0213] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0214] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0215] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0216] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0217] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0218] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0219] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0220] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A map partitioning method, characterized in that, include: The autonomous mobile device is controlled to traverse the operating environment using at least two trajectory modes, and multiple environmental images are acquired during the traversal using the at least two trajectory modes. Based on the frequency and location of the zoning reference objects existing in the work environment in the multiple environmental images, mark the location of the zoning reference objects in the environmental map corresponding to the work environment; The environment map is divided into zones based on the location of the zone reference in the environment map; wherein, the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions, and the cleaning paths of the at least two trajectory modes are interleaved and superimposed.

2. The method according to claim 1, characterized in that, Control the autonomous mobile device to traverse the operating environment using at least two trajectory modes, including: After the autonomous mobile device has traversed the working environment using the previous trajectory mode, control the autonomous mobile device to return to the designated starting position and continue to traverse the working environment again using the next trajectory mode; or After the autonomous mobile device has traversed the working environment using the previous trajectory mode, it is controlled to start from the current position and continue to traverse the working environment again using the next trajectory mode.

3. The method according to claim 1, characterized in that, Based on the frequency and location of zoning reference points in the multiple environmental images, mark the locations of the zoning reference points in the environmental map corresponding to the work environment, including: Select the target environment image containing the partition reference from the plurality of environment images; Based on the position of the partition reference in the target environment image, candidate locations where the partition reference exists are marked in the environment map; Candidate locations that have been marked less than a set threshold number are removed to obtain the location of the partition reference in the environment map.

4. The method according to claim 3, characterized in that, Selecting the target environment image containing the partition reference from the plurality of environment images includes: The multiple environmental images are input into an image recognition model for recognition to obtain a target environmental image including the partition reference object; wherein, the image recognition model is pre-trained using sample images containing the partition reference object.

5. The method according to any one of claims 1-4, characterized in that, The at least two trajectory patterns include a horizontal bow-shaped trajectory pattern and a vertical bow-shaped trajectory pattern. The main travel directions of the horizontal bow-shaped trajectory pattern and the vertical bow-shaped trajectory pattern are perpendicular to each other, and the main travel direction is the direction in which the autonomous mobile device travels along the longer side of the bow-shaped trajectory.

6. The method according to any one of claims 1-4, characterized in that, The at least two trajectory modes include: a horizontal irregular trajectory mode and a vertical irregular trajectory mode. The travel trajectories corresponding to the horizontal irregular trajectory mode and the vertical irregular trajectory mode include connected straight line trajectories and curved trajectories. The travel direction of the autonomous mobile device along the straight line trajectory is the main travel direction. The main travel directions of the horizontal irregular trajectory mode and the vertical irregular trajectory mode are perpendicular to each other.

7. The method according to any one of claims 1-4, characterized in that, The trajectory gaps between different trajectory patterns are different, and for two adjacent trajectory patterns, the trajectory gap of the latter trajectory pattern is greater than that of the former trajectory pattern.

8. A semantic map construction method, characterized in that, include: The autonomous mobile device is controlled to traverse the operating environment using at least two trajectory modes; Multiple environmental images are acquired during the traversal process employing at least two of the aforementioned trajectory modes; Based on the multiple environmental images, a semantic map of the working environment is constructed; at least two trajectory modes have different main travel directions, which refer to the travel direction that meets the set conditions, and the cleaning paths of the at least two trajectory modes are interleaved and superimposed.

9. The method according to claim 8, characterized in that, Based on the multiple environmental images, a semantic map of the working environment is constructed, including: Based on the multiple environmental images, identify the objects and their semantic information in the work environment; Based on the objects and their semantic information present in the work environment, determine at least two partitions and their semantic information contained in the work environment; Based on the at least two partitions and their semantic information, construct a semantic map of the operating environment.

10. The method according to claim 9, characterized in that, Based on the objects and their semantic information existing in the work environment, determine at least two partitions and their semantic information contained in the work environment, including: Based on the partition references and their semantic information existing in the work environment, the work environment is divided into at least two partitions; The semantic information of the at least two partitions is determined based on the non-partition references and their semantic information contained in the at least two partitions.

11. An object recognition method, characterized in that, include: The autonomous mobile device is controlled to traverse the operating environment using at least two trajectory modes; Multiple environmental images are acquired during the traversal process employing at least two of the aforementioned trajectory modes; Based on the multiple environmental images, target objects in the working environment are identified; the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions, and the cleaning paths of the at least two trajectory modes are interleaved and superimposed.

12. An autonomous mobile device, characterized in that, include: The device body, which is equipped with a memory and a processor; The memory is used to store computer programs; The processor, coupled to the memory, is configured to execute the computer program for: The autonomous mobile device is controlled to traverse the operating environment using at least two trajectory modes, and multiple environmental images are collected during the traversal process; Based on the frequency and location of the zoning reference objects existing in the work environment in the multiple environmental images, mark the location of the zoning reference objects in the environmental map corresponding to the work environment; The environment map is divided into zones based on the location of the zone reference in the environment map; wherein, the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions, and the cleaning paths of the at least two trajectory modes are interleaved and superimposed.

13. An autonomous mobile device, characterized in that, include: The device body, which is equipped with a memory and a processor; The memory is used to store computer programs; The processor, coupled to the memory, is configured to execute the computer program for: The autonomous mobile device is controlled to traverse the operating environment using at least two trajectory modes; Multiple environmental images are acquired during the traversal process employing at least two of the aforementioned trajectory modes; Based on the multiple environmental images, a semantic map of the working environment is constructed; at least two trajectory modes have different main travel directions, which refer to the travel direction that meets the set conditions, and the cleaning paths of the at least two trajectory modes are interleaved and superimposed.

14. An autonomous mobile device, characterized in that, include: The device body, which is equipped with a memory and a processor; The memory is used to store computer programs; The processor, coupled to the memory, is configured to execute the computer program for: The autonomous mobile device is controlled to traverse the operating environment using at least two trajectory modes; Multiple environmental images are acquired during the traversal process employing at least two of the aforementioned trajectory modes; Based on the multiple environmental images, target objects in the working environment are identified; the main travel directions of the at least two trajectory modes are different, and the main travel direction refers to the travel direction that meets the set conditions, and the cleaning paths of the at least two trajectory modes are interleaved and superimposed.

15. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method according to any one of claims 1-11.

Images