Target detection method, self-moving mowing equipment and computer readable medium

Abstract

The invention relates to a target detection method, a self-moving mowing device and a computer readable medium, the self-moving mowing device comprises a non-independently arranged target detection module, a rain detection function is realized by using a non-redundant sensor, and in the detection process, target feature information is determined based on at least two sets of target information, so that the self-moving mowing device can detect the target. And whether rain is detected is identified according to the determined target feature data, so that the cost problem of multi-sensor collaboration can be solved, the problem of insufficient collaboration of multi-source data can be solved, and the problems of instability and loud noise of single-frame data can be overcome. Furthermore, by limiting data detection of a specific space, external interference is effectively reduced, and the accuracy of data is improved.

Description

Technical Field

[0001] The present invention relates to the field of environmental perception, and in particular to a target detection method implemented by using a non-independent sensor, a self-propelled mowing device, and a computer-readable medium. Background Art

[0002] Smart lawn mowers can mow the lawn autonomously, without the need for direct human control or operation. They are low-power, quiet, and compact, significantly reducing manual labor. While the number of smart lawn mower users continues to increase, the functional requirements for smart lawn mowers are also posing new challenges.

[0003] To ensure safe operation in complex outdoor environments, extend equipment life, improve efficiency, and adapt to changing weather conditions, more and more smart lawn mowers are adding rain detection features. Current approaches for implementing rain detection include: 1) Using independent environmental sensors for rain detection. Environmental sensors, such as raindrop sensors, humidity sensors, and air pressure sensors, utilize humidity and temperature changes during rainy days to detect rain. However, for a robot already equipped with radar for positioning, adding environmental sensors requires additional procurement costs and additional sensor interfaces on the processor. Furthermore, raindrop sensors are affected by residual rain on the sensor surface, and will continue to detect rain for a period of time after the rain stops. Humidity sensors cannot distinguish humidity changes caused by evaporation, for example. Air pressure sensors do not detect directly, making it difficult to distinguish between cloudy days, storms, and rainfall. 2) Vision-based rain detection: Using computer vision to analyze camera image data can determine whether there are water stains caused by raindrops on the ground. However, this approach typically requires extensive data collection for model training and cannot detect rain at night. 3) Sound-based rain detection: Rain can be detected by identifying the sound of raindrops falling on the housing. However, this increases the hardware cost of the audio equipment and the accuracy of detecting light rain is low. 4) Multi-sensor collaboration: For example, rain detection requires integration with humidity sensors and LiDAR (for identifying areas of accumulated water) to avoid misjudgments by a single sensor. 5) Rain detection is achieved through LiDAR point cloud analysis. In the existing LiDAR point cloud analysis process, single-frame data sets are used for rain detection to improve data processing efficiency.

[0004] Through the analysis of the existing technology, the main defects are as follows:

[0005] 1) Inadequate environmental adaptability: Traditional rain sensors are susceptible to interference from non-rainfall factors (such as artificial spray and condensation), leading to false triggering of return calls or missed rainfall detections. LiDAR point cloud quality degrades in rainy and foggy weather, making it difficult to distinguish real obstacles from raindrop noise in a single frame. 2) Inadequate multi-source data synergy: Most solutions use a single sensor (such as a rain sensor or a lidar alone) without integrating meteorological data (such as humidity and wind speed) with dynamic point cloud information, resulting in biased decision-making. 3) System complexity and cost conflict: Redundant sensor deployment (such as a rain sensor + lidar + camera) increases hardware cost and energy consumption, and multi-sensor data fusion algorithms are difficult to run in real time on embedded platforms. 4) LiDAR generates spatial point cloud information by transmitting and receiving signals. For individual points, rain detection can be determined by analyzing the characteristics of the received signal. If a frame of the point cloud contains a large number of rain points, it can be considered rain. This solution relies on the sensor performance of the radar manufacturer; weak detection performance can result in inability to identify rain points.

[0006] In view of this, it is indeed necessary to provide an improved self-propelled mowing device to overcome the defects of the prior art. Summary of the Invention

[0007] To address the shortcomings of existing technologies, this invention addresses the challenges of rain affecting the positioning and obstacle avoidance of robots equipped with 3D LiDAR. The goal of this invention is to directly use the robot's own 3D LiDAR for rain detection. By collecting data multiple times and taking into account features such as the number of specific frames, this invention effectively addresses issues such as single-frame inaccuracy, insufficient environmental adaptability, high system complexity, and poor multi-source data fusion.

[0008] The present invention provides a target detection method, which is applied to a self-propelled lawn mowing device, wherein the self-propelled lawn mowing device includes a non-independent target detection module, and the target detection method includes:

[0009] Obtain at least two sets of target information in the target space;

[0010] Running a target detection program to process at least two sets of target information to determine whether the target information includes target feature information;

[0011] When at least one of the at least two sets of target information does not include target feature information, corresponding indication information is fed back or target information is continued to be detected.

[0012] Further improvement plans are:

[0013] The at least two sets of target data each include multiple frames of point cloud data, and the point cloud data of each frame includes the position coordinates and intensity value of each point.

[0014] Further improvement plans are:

[0015] The target space includes a first subspace and a second subspace; and the method further includes:

[0016] Get the number n1 of points with low intensity values ​​in the first subspace;

[0017] Get the number n2 of points in the second subspace;

[0018] When n1 and n2 meet the target detection requirements, the target information includes target feature information.

[0019] Further improvement plans are:

[0020] The target space includes a plurality of sectors, each sector includes a short-range area and a long-range area, and the method further includes:

[0021] Get the number n1 of points with low intensity values ​​in the close-range area of ​​each sector;

[0022] Get the number n2 of the points in the long-distance area of ​​each sector;

[0023] When n1 and n2 in the same sector meet the target detection requirements, the sector is the target sector;

[0024] When the number of the target sectors exceeds a specified threshold, the target information includes target feature information.

[0025] Further improvement plans are:

[0026] The target space includes a zero-point space; the method further includes: when target information is detected in the zero-point space, the self-propelled mowing device performs abnormal feedback.

[0027] Further improvement plans are:

[0028] Before acquiring at least two sets of target information in the target space, the method further includes: performing a cleaning operation on the target detection module of the self-propelled mowing device.

[0029] Further improvement plans are:

[0030] The method further includes: when the at least two sets of target information both include target feature information, outputting a control instruction to control the self-propelled mowing device to perform a regression operation.

[0031] The present invention also provides a self-propelled lawn mowing device, which includes a target detection module, a processing module, and an execution module that are not independently arranged;

[0032] The target detection module is used to obtain at least two sets of target information in the target space;

[0033] The processing module is used to run the target detection program and process at least two sets of target information to determine whether the target information includes target feature information; when at least one of the at least two sets of target information does not include target feature information, it feeds back corresponding indication information or continues to detect target information.

[0034] The present invention also provides a self-moving lawn mowing device, which includes a memory and a processor. The memory stores a computer program that can be run on the processor. The method is characterized in that the processor implements the method when executing the computer program.

[0035] The present invention also provides a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code causes the processor to execute the method.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] The present invention provides a target detection method for a self-moving lawn mowing device, wherein the self-moving lawn mowing device includes a non-independent target detection module. The present invention obtains at least two sets of target information in the target space; runs a target detection program, processes the at least two sets of target information to determine whether the target information includes target feature information; when at least one of the at least two sets of target information does not include target feature information, feedback is given of corresponding indication information or the target information is continued to be detected. In the above manner, a non-redundantly set sensor is used to implement the rain detection function, and during the detection process, the target feature information is determined based on at least two sets of target information, and whether rain is detected is identified based on the determined target feature data. This not only solves the cost problem of multi-sensor collaboration, but also solves the lack of collaboration of multi-source data, and overcomes the problems of instability and high noise in single-frame data. Furthermore, the present invention effectively reduces external interference and improves data accuracy by limiting data detection to a specific space. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 A schematic structural diagram of a preferred embodiment of the present invention;

[0039] Figure 2a A top view of a radar scanning area according to a preferred embodiment of the present invention;

[0040] Figure 2b A three-dimensional schematic diagram of a radar scanning area according to a preferred embodiment of the present invention;

[0041] Figure 3Flowchart of a method according to a preferred embodiment of the present invention. DETAILED DESCRIPTION

[0042] The terms used in this invention are for the purpose of describing specific embodiments only and are not intended to limit the invention. This invention addresses the problem of single-frame inaccuracy, insufficient environmental adaptation, high system complexity, and poor multi-source data fusion by using the robot's onboard 3D LiDAR to detect rain and other weather conditions. By collecting data multiple times and taking into account features such as the number of specific frames, this invention effectively addresses the issues of single-frame inaccuracy, insufficient environmental adaptation, high system complexity, and poor multi-source data fusion.

[0043] The present invention provides a target detection method, which is applied to a self-propelled lawn mowing device, wherein the self-propelled lawn mowing device includes a non-independent target detection module, wherein the non-independent target detection module refers to a detection module capable of performing at least two functions. Preferably, the detection module may be a laser radar, which can perform at least two of the functions of positioning, obstacle detection, rain detection, and slope detection. The target detection method includes:

[0044] At least two sets of target information in the target space are acquired; the at least two sets of target data each include multiple frames of point cloud data, and the point cloud data of each frame includes the position coordinates and intensity value of each point.

[0045] Running a target detection program to process at least two sets of target information to determine whether the target information includes target feature information;

[0046] When at least one of the at least two sets of target information does not include target feature information, corresponding indication information is fed back or target information is continued to be detected.

[0047] The method further includes: when the at least two sets of target information both include target feature information, outputting a control instruction to control the self-propelled mowing device to perform a regression operation.

[0048] In a specific embodiment of the present invention, the self-propelled mowing device is equipped with a 3D laser radar for sensing the environment, such as achieving positioning, obstacle detection, rain perception, etc. The self-propelled device is also equipped with an actuator, such as wheels, a blade, a lifting mechanism, etc., so as to perform related actions according to the control signal sent by the MCU. Figure 1As shown, the laser radar collects point cloud data in space and sends it to a processing module, such as an MCU, at a preset frequency (e.g., 10 Hz). The MCU runs a rain detection program responsible for processing the point cloud data, analyzing whether it is raining, and sending the results to the main control program, which controls and plans the machine's further actions (e.g., rain regression). In alternative embodiments, the present invention can also detect various weather conditions such as rain, fog, hail, and haze by presetting corresponding detection programs and then matching the point cloud distribution model to achieve corresponding weather detection. The self-propelled mowing device involved in the present invention communicates with multiple interactive modules. The preset frequency is related to the number of interactive modules and the time required for data transmission of each interactive module. The setting of the preset frequency can effectively prevent data conflicts and prevent multiple interactive modules from sending data to the MCU at the same time.

[0049] In order to prevent the radar single frame data from being unstable, in a specific embodiment, as shown in FIG. Figure 3 As shown, when the rain detection function is activated, several data sets can be continuously collected as a set of data, and then another set of data is collected at a preset interval. When more data sets are needed, data collection is carried out sequentially at the same preset interval. To reduce the system's computing power requirements, the number of data sets can preferably be any value between 3 and 10. The number of point cloud frames included in at least two data sets can be the same or different, and the specific value can be determined based on the actual accuracy requirements. In a specific example, taking three frames of point cloud data as an example, the lidar continuously collects three frames of point cloud data as a set of data. At the same time, to prevent false detection caused by temporary interference, another three frames of point cloud data are collected at an interval of ΔT as another set of data. After obtaining sufficient data totaling 3*2 frames, point cloud statistical discrimination is performed, that is, each frame of data is traversed to perform statistical analysis on each frame. A single frame of point cloud data contains data for W points (x, y, z, i), where x, y, and z are the coordinate values ​​of the point, and i is the intensity value of the corresponding point (range 0-255). The number of single-frame point cloud data W is determined by the type / model of the radar.

[0050] In a further embodiment, the target space includes a first subspace and a second subspace; and the method further includes:

[0051] Get the number n1 of points with low intensity values ​​in the first subspace;

[0052] Get the number n2 of points in the second subspace;

[0053] When n1 and n2 meet the target detection requirements, the target information includes target feature information.

[0054] In an alternative embodiment,

[0055] The target space includes multiple sectors, and each sector includes a short - distance area and a long - distance area. The method further includes:

[0056] Obtaining the number n1 of points with low intensity values in the short - distance area of each sector;

[0057] Obtaining the number n2 of points in the long - distance area of each sector;

[0058] When the n1 and n2 in the same sector meet the target detection requirements, then the sector is a target sector;

[0059] When the number of the target sectors exceeds a specified threshold, then the target information includes target feature information.

[0060] In a specific embodiment, taking a 3D lidar as an example, as Figure 2a shown, the 3D lidar can detect data within a maximum detection distance Lf with the radar center position as the origin. Among them, Lf is greater than or equal to 40m, and its specific value is determined by the own parameters of the lidar. In the specific embodiment of the present invention, in order to improve the detection accuracy, only the target space is detected. Among them, the target space refers to the area with a ranging range of (Ln~a*Lf), where a is a value between 0.5 and 1, and preferably can be 0.5, 0.6, 0.7, 0.8, 0.9, 1; as Figure 2b shown, the target space includes a short - distance area (Ln~Lm) and a long - distance area (Lm~Lf); define the distance from the radome surface to the radar center as Lt, 0 < Ln < Lt, in a preferred embodiment, 0.8*Lt < Ln ≤ Lt, Lt ≤ Lm ≤ 1.5*Lt. In a specific embodiment, Ln can take values between 0 and 0.15, specifically can be 0.05, 0.08, 0.1, 0.12, etc.; Lm is a value between 0.15 and 0.3, specifically can be 0.15, 0.18, 0.20, 0.23, 0.25, 0.28, 0.3, etc. Among them, in the extremely short - distance area (0 - Ln) and the short - distance area (Ln~Lm), especially in the extremely short - distance area (0 - Ln), there is generally no point cloud during normal operation of the device. When there is a short - distance point cloud, there are two situations. One is that there is an object in the short - distance area (usually objects are not allowed to invade the extremely short - distance area to prevent damage to the radome), and the other is that there is an attachment on the radar surface (radome). According to the point cloud performance after rain attachment to the radar, when rain is attached, low - intensity noise points will be generated in the short - distance area (Ln~Lm) of the radar. Therefore, counting this part of the point cloud can achieve rain detection. At the same time, in order to prevent false rain detection caused by short - distance objects, zonal statistics and long - distance point cloud statistics are carried out.

[0061] In a further embodiment, when N1>0.0005W*4 and N2>0.005W*4, n1 and n2 meet the target detection requirements, and the target information includes target feature information, that is, the point cloud data includes rain frames.

[0062] In a further alternative embodiment, the present invention traverses each point in the point cloud and calculates:

[0063] Theta=arctan(y / x)

[0064] Divide the point cloud into N sectors along the circumference according to the Theta value, and the preferred N is 12. Calculate the distance between the point cloud and the origin of the radar coordinate system:

[0065] L=(x*x+y*y+z*z)^1 / 2

[0066] L between (Ln ~ Lm) is marked as a close-range point, and L outside Lm is marked as a long-range point. Intensity values ​​of 1 to 20 are marked as low-intensity points. For each sector, the number of low-intensity point clouds in the close-range area (Ln ~ Lm) and the number of point clouds in the long-range area (Lm ~ Lf) are counted separately. For a single sector, if n1 and n2 meet a certain threshold range, it is marked as a rain sector, that is, when N1>0.0005W and N2>0.005W; if there are 4 or more rain sectors in 12 sectors, it is marked as a rain frame.

[0067] If there is rain data in both the first three frames and the last three frames, it is determined to be raining.

[0068] In a preferred embodiment of the present invention, the target space includes a zero-point space; for an extremely close distance area (0-Ln), or an area with a distance of (0-Lt) from the center of the radar, the zero-point space can be defined, and the method further includes: when target information is detected in the zero-point space, the self-propelled mowing device performs abnormal feedback. Generally, the zero-point space is mainly the space within the radar cover. Under normal circumstances, there is no point cloud data in the zero-point space. If point cloud data appears, it means that the laser radar may have an abnormality, which affects the accuracy of the detection; the user is prompted to check the device through the abnormal error report. In an alternative embodiment, when an abnormal error report is detected, detection is performed after a period of time. If an abnormality is detected two or three times in a row, the user is prompted to check the device through the abnormal error report; in a further embodiment, when an abnormality is issued, the device is controlled to shut down or return to the charging pile.

[0069] In a further embodiment, in order to further improve the accuracy of detection, the interference of obstruction on the surface of the radar cover is reduced. Before obtaining at least two sets of target information in the target space, the method also includes: performing a cleaning operation on the target detection module of the self-moving mowing device. Specifically including: the self-moving mowing device performs a cleaning operation after returning to the charging pile, and / or, performs a cleaning operation before the self-moving mower leaves the charging pile. In a further preferred embodiment, the charging pile includes a cleaning head for cleaning the laser radar. After the cleaning is completed and the self-moving mowing device has not left the charging pile, the cleaning head maintains contact with the radar cover to prevent dust / particulate matter, etc. from attaching to the radar cover again, thereby ensuring the clean state of the radar cover.

[0070] Through the above approach, rain detection is achieved using non-redundant sensors. During the detection process, target feature information is determined based on at least two sets of target information. The presence of rain is then determined based on the determined target feature data. This not only addresses the cost of multi-sensor collaboration, but also addresses the lack of collaboration between multiple sources of data, as well as the issues of unstable and noisy single-frame data. Furthermore, by limiting data detection to a specific space, the present invention effectively reduces external interference and improves data accuracy.

[0071] The present invention also provides a self-propelled lawn mowing device, which includes a target detection module, a processing module, and an execution module that are not independently arranged;

[0072] The target detection module is used to obtain at least two sets of target information in the target space;

[0073] The processing module is used to run the target detection program and process at least two sets of target information to determine whether the target information includes target feature information; when at least one of the at least two sets of target information does not include target feature information, it feeds back corresponding indication information or continues to detect target information.

[0074] The present invention also provides a self-moving lawn mowing device, which includes a memory and a processor. The memory stores a computer program that can be run on the processor. The method is characterized in that the processor implements the method when executing the computer program.

[0075] The present invention also provides a computer-readable medium having a non-volatile program code executable by a processor, wherein the program code causes the processor to execute the method.

[0076] The present invention provides a target detection method for a self-moving lawn mowing device, wherein the self-moving lawn mowing device includes a non-independent target detection module. The present invention obtains at least two sets of target information in the target space; runs a target detection program, processes the at least two sets of target information to determine whether the target information includes target feature information; when at least one of the at least two sets of target information does not include target feature information, feedback is given of corresponding indication information or the target information is continued to be detected. In the above manner, a non-redundantly set sensor is used to implement the rain detection function, and during the detection process, the target feature information is determined based on at least two sets of target information, and whether rain is detected is identified based on the determined target feature data. This not only solves the cost problem of multi-sensor collaboration, but also solves the lack of collaboration of multi-source data, and overcomes the problems of instability and high noise in single-frame data. Furthermore, the present invention effectively reduces external interference and improves data accuracy by limiting data detection to a specific space.

[0077] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems and devices can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0078] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0079] Finally, it should be noted that the above-described embodiments are only specific implementations of the present invention, which are used to illustrate the technical solutions of the present invention, rather than to limit them. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the above-described embodiments, it should be understood by those skilled in the art that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the above-described embodiments within the technical scope disclosed by the present invention, or perform equivalent replacements on some of the technical features thereof. Such modifications, changes, or replacements do not deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A target detection method for a self-propelled lawn mowing device, characterized in that: The self-propelled mowing device includes a non-independent target detection module, and the target detection method includes: Obtain at least two sets of target information in the target space; Running a target detection program to process at least two sets of target information to determine whether the target information includes target feature information; When at least one of the at least two sets of target information does not include target feature information, corresponding indication information is fed back or target information is continued to be detected.

2. The target detection method according to claim 1, wherein: The at least two sets of target data each include multiple frames of point cloud data, and the point cloud data of each frame includes the position coordinates and intensity value of each point.

3. The target detection method according to claim 2, wherein: The target space includes a first subspace and a second subspace; and the method further includes: Get the number n1 of points with low intensity values ​​in the first subspace; Get the number n2 of points in the second subspace; When n1 and n2 meet the target detection requirements, the target information includes target feature information.

4. The target detection method according to claim 2, wherein: The target space includes a plurality of sectors, each sector includes a short-range area and a long-range area, and the method further includes: Get the number n1 of points with low intensity values ​​in the close-range area of ​​each sector; Get the number n2 of the points in the long-distance area of ​​each sector; When n1 and n2 in the same sector meet the target detection requirements, the sector is the target sector; When the number of the target sectors exceeds a specified threshold, the target information includes target feature information.

5. The target detection method according to claim 1, wherein: The target space includes a zero-point space; the method further includes: when target information is detected in the zero-point space, the self-propelled mowing device performs abnormal feedback.

6. The target detection method according to claim 1, wherein: Before acquiring at least two sets of target information in the target space, the method further includes: performing a cleaning operation on the target detection module of the self-propelled mowing device.

7. The target detection method according to claim 1, wherein: The method further comprises: When the at least two sets of target information both include target feature information, a control instruction is output to control the self-propelled mowing device to perform a regression operation.

8. A self-propelled mowing device, characterized in that: The self-propelled mowing device includes a target detection module, a processing module and an execution module that are not independently arranged; The target detection module is used to obtain at least two sets of target information in the target space; The processing module is used to run a target detection program to process at least two sets of target information to determine whether the target information includes target feature information; When at least one of the at least two sets of target information does not include target feature information, corresponding indication information is fed back or target information is continued to be detected.

9. A self-propelled mowing device, characterized in that: The self-propelled mowing device includes a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and is characterized in that the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable medium having a non-volatile program code executable by a processor, characterized in that: The program code enables the processor to execute the method according to any one of claims 1 to 7.

Images