Mobile device control system and method

Abstract

The invention relates to the technical field of control systems, and provides a mobile device control system and method, and the method comprises the steps: obtaining a target image through a multispectral sensor, analyzing the sub-target density in combination with a convolutional neural network, calculating a normalized target index, and fusing the two to generate a health degree matrix; the comprehensive perception of the target area from the growth state of the target to the environmental factors is realized; working parameters and working priorities are distributed based on the health degree matrix, so that the operation of the mobile equipment is more targeted, adaptive working parameters can be obtained in target areas in different health states, and the working efficiency and the target management quality are improved; meanwhile, the actual motor torque is monitored in real time, the motor torque is determined in combination with the gradient and the density, the mobile equipment can keep stable operation in complex terrains and different target density environments, when the torque deviates, local re-scanning is triggered, the health degree matrix is updated, the accuracy and reliability of long-term operation of the system are ensured, and the service life of the system is prolonged. And the autonomous operation capability of the mobile equipment is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of control systems, in particular to a mobile device control system and method. BACKGROUND

[0002] With the development of intelligent technology, mobile devices are increasingly widely used in the fields of agriculture and garden maintenance, and the performance of their control systems directly affects the work efficiency and management quality. At present, the control system of a mobile device has gradually introduced multispectral sensing technology, and some systems also monitor environmental parameters such as soil humidity to provide basic data for subsequent work decision-making. At the decision-making and control level, the existing system can assign work parameters and work priorities to the mobile device based on the obtained feature data, and generate corresponding control signals to drive the device to work. However, the existing technology still has many deficiencies.

[0003] Firstly, in terms of data fusion and analysis, most systems do not fully utilize the features of multispectral images, often relying on only a single feature or a simple combination of features for evaluation, and failing to effectively fuse physical and physiological features, resulting in difficulty in comprehensively and accurately reflecting the actual state of the target. Secondly, in terms of decision-making mechanism, the assignment of work parameters and work priorities is mostly based on fixed rules or simple threshold judgments, lacking the ability to adaptively adjust to the dynamic changes of the target health degree. When the state of the target area changes complexly, parameter mismatch and low work efficiency may occur.

[0004] Furthermore, in terms of feedback control and dynamic adjustment, the existing system's monitoring of operating parameters such as motor torque mostly stays at the level of simple threshold comparison, failing to dynamically determine reasonable torque in combination with actual environmental factors, and when parameter abnormalities are found, the adjustment mechanism triggered is not precise enough, often requiring re-detection of a larger area, which not only affects the continuity of work, but also causes decision-making deviation due to untimely data updating. Most of them do not solve the problem of how to realize the perception of the target state by the mobile device and the work adaptation through multi-source data fusion processing in complex environments. SUMMARY

[0005] In view of the deficiencies of the existing technology, the present application provides a mobile device control system and method.

[0006] In a first aspect, the present application provides a mobile device control system, which comprises: acquiring a target image of a target area through a multispectral sensor, analyzing the density of a sub-target based on the target image of the target area through a convolutional neural network, and calculating a normalized target index in real time, generating a health degree matrix of the target according to the density of the sub-target and the normalized target index, and monitoring the soil humidity of the target area.

[0007] The health degree matrix of the target is used to allocate working parameters and working priorities of the mobile device to generate a control signal, and to trigger external devices according to a normalized target index and soil moisture of the target area;

[0008] The control signal is received to drive the mobile device, the actual motor torque of the mobile device is monitored in real time, the target motor torque of the mobile device is determined according to the sub-target slope and the sub-target density, the actual motor torque and the target motor torque are compared to trigger local rescan of the target image, and the health degree matrix of the target is updated.

[0009] As an optional implementation, the generation logic of the health degree matrix of the target comprises:

[0010] The corrected target image is subjected to band operation to calculate the normalized target index in real time, the normalized target index is subjected to space-time calibration, and the normalized target index is subjected to outlier filtering;

[0011] The target area is divided into a plurality of grids based on the sub-target density, the normalized target index and the sub-target density in each grid are counted, and the level of the normalized target index is determined to obtain an index level;

[0012] The health degree level of each grid is determined according to the density level and the index level, and a risk flag is allocated to each grid based on the health degree level;

[0013] The health degree matrix of the target is generated based on the density level, the index level and the risk flag.

[0014] As an optional implementation, the analysis sub-logic of the sub-target density comprises:

[0015] The target image of the target area is obtained by a multispectral sensor, the target image comprises a visible light image and a near-infrared image, and the inclination data of the mobile device is monitored in real time to correct the geometric distortion of the target image;

[0016] The corrected target image is subjected to environmental compensation, a double channel is constructed, and the visible light image and the near-infrared image are subjected to pixel separation by threshold segmentation to extract the target in the visible light image and the near-infrared image;

[0017] The visible light image is processed by an edge detection algorithm in the visible light channel to identify the first feature contour of the target, the first feature distance is calculated to determine the physical density, the near-infrared reflectivity is calculated in the near-infrared channel, and the second feature reflection intensity of the target in the near-infrared image is determined to determine the physiological density;

[0018] The features of the two channels are extracted by a convolutional neural network, the spatial structure of the sub-target is extracted according to the physical density in the visible light channel, the physiological information of the sub-target is extracted based on the physiological density in the near-infrared channel, and the feature weights of the two channels are allocated based on the attention mechanism to fuse the features of the two channels and form a fusion feature;

[0019] According to the fusion feature analysis sub-target density, the fusion feature is mapped to a density score, the density level is divided by a clustering algorithm according to the density score, and the difference between the physical density and the physiological density is verified to trigger an abnormal warning.

[0020] As an optional implementation, the generation logic of the control signal includes:

[0021] Based on the working parameters, the motor energy consumption demand of the mobile device is determined, and the motor energy consumption demand of each state grid block is adjusted according to the sub-target density and the working priority of each state grid block;

[0022] According to the current battery power of the mobile device and the historical energy consumption data of each motor, the total power threshold that can be allocated is calculated, the motor power is allocated in combination with the motor energy consumption demand of each state grid block, and the motor power is adjusted according to the sub-target density;

[0023] Taking the center of the state grid block as the target position, combining the current position of the mobile device, the sub-target slope, the sub-target density and the motor energy consumption, the travel path of the mobile device is planned, and the travel path of the mobile device is optimized based on the obstacles of each state grid block to generate the control signal.

[0024] As an optional implementation, the allocation sub-logic of the working parameters and the working priority includes:

[0025] The health matrix of the target is processed by a density clustering algorithm to identify the health state of each grid, the grids with the same health state are merged to form a state grid block, and a working strategy is assigned to each state grid block;

[0026] Based on the working strategy, the working parameters of the mobile device are assigned to each state grid block, including the working height, the device component rotating speed and the travel speed;

[0027] According to the properties of the state grid block, the working priority of each state grid block is assigned, and the obstacles of each state grid block are monitored.

[0028] As an optional implementation, the external device is triggered according to the normalized target index and the soil moisture of the target area;

[0029] The priority of the external device is determined, and when multiple external devices are triggered at the same time, the triggering of the external device is cooperatively scheduled based on the priority of the external device.

[0030] As an optional implementation, the trigger condition of the local rescan of the target image comprises:

[0031] The torque difference between the actual motor torque and the target motor torque is calculated, and a difference threshold is configured, which includes a first difference and a second difference;

[0032] When the torque difference is greater than the first difference and less than or equal to the second difference, a single-point rescan of the target image is triggered;

[0033] When the torque difference is greater than the second difference, a region rescan of the target image is triggered;

[0034] After triggering the single-point rescan and the region rescan of the target image, the sub-target slope is verified to determine the target motor torque of the mobile device, and it is judged whether the target image needs to be scanned again.

[0035] As an optional implementation, the determination sub-logic of the target motor torque comprises:

[0036] A control signal is received to drive the mobile device, the sub-target slope is monitored, and the driving mode of the mobile device is determined according to the sub-target slope, the driving mode including a standard mode and a compensation mode;

[0037] The initial motor torque of the mobile device is determined based on the driving mode, and the initial motor torque is corrected according to the sub-target density;

[0038] The degree of wear of the device component of the mobile device is identified in real time through voiceprint analysis, and the corrected initial motor torque is compensated based on the degree of wear of the device component to determine the target motor torque.

[0039] As an optional implementation, the update logic of the health matrix comprises:

[0040] After performing the local rescan of the target image, the target image of the target region is reacquired to determine the torque change type, the torque change type including a physical change and a physiological change;

[0041] If it is a physical change, the change feature of the first feature contour is identified to reextract the spatial structure of the target;

[0042] If it is a physiological change, the second feature reflection intensity of the target is updated to reevaluate the risk mark of each grid in the target region;

[0043] The change rates of the spatial structure and the risk mark after three local rescans are compared to dynamically update the health matrix.

[0044] In a second aspect, the application provides a mobile device control method, which comprises: acquiring a target image of a target area by a multispectral sensor, analyzing a sub-target density based on the target image of the target area by a convolutional neural network;

[0045] calculating a normalized target index in real time, generating a health degree matrix of the target according to the sub-target density and the normalized target index, and monitoring soil humidity of the target area;

[0046] allocating a working parameter and a working priority of the mobile device based on the health degree matrix of the target to generate a control signal;

[0047] comprehensively judging according to the normalized target index and the soil humidity of the target area to trigger an external device;

[0048] receiving the control signal to drive the mobile device, monitoring an actual motor torque of the mobile device in real time, and comprehensively determining a target motor torque of the mobile device according to the sub-target slope and the sub-target density;

[0049] comparing the actual motor torque and the target motor torque to trigger a local rescan of the target image, and updating the health degree matrix of the target.

[0050] Compared with the prior art, the application has the beneficial effects that: the target image is acquired by the multispectral sensor, the sub-target density is analyzed by the convolutional neural network, the normalized target index is calculated, the health degree matrix is generated by fusing the two, the soil humidity is monitored, the overall perception of the target area from the growth state of the target itself to the environmental factors is realized, the limitation of the single parameter judgment of the traditional system is broken through, the evaluation of the health condition of the target is more accurate and comprehensive, and reliable basic data is provided for subsequent decision-making.

[0051] The working parameter and the working priority are allocated based on the health degree matrix, the operation of the mobile device is more targeted, the target area with different health states can obtain adaptive working parameters, the problems of over-operation or insufficient operation under the unified parameters are avoided, the operation efficiency and the target management quality are improved, the actual motor torque is monitored in real time, the motor torque is determined in combination with the slope and the density, the mobile device can keep stable operation in the complex terrain and the environment with different target densities, when the torque deviates, the local rescan is triggered and the health degree matrix is updated, a closed dynamic adjustment mechanism is formed, the operation deviation caused by the environmental change or the device state fluctuation is avoided, the accuracy and reliability of long-term operation of the system are ensured, the demand for manual intervention is reduced, and the autonomous operation ability of the mobile device is improved. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings. Among them:

[0053] Figure 1 The system flowchart of the mobile device control system provided by the embodiments of the present application;

[0054] Figure 2 The lawn density analysis sub-logic diagram of the mobile device control system provided by the embodiments of the present application;

[0055] Figure 3 The triggering condition diagram of the local rescan of the vegetation image of the mobile device control system provided by the embodiments of the present application;

[0056] Figure 4 The method flowchart of the mobile device control method provided by the embodiments of the present application. DETAILED DESCRIPTION

[0057] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be described clearly and completely in the following description of the drawings. Obviously, the described embodiments are only some of the embodiments of the present application, not all.

[0058] Embodiment 1

[0059] As shown in Figure 1 The system flowchart of the mobile device control system provided by the embodiments of the present application, the system includes a spectrum perception module, an intelligent decision module and a control feedback module.

[0060] Here, the scene of the unmanned mower is taken as an example, that is, the unmanned mower represents the mobile device, the vegetation image represents the target image, the lawn density represents the sub-target density, the normalized vegetation index represents the normalized target index, the cutting parameter represents the working parameter, the cutting priority represents the working priority, the lawn slope represents the sub-target slope, the leaf profile represents the first feature profile, the leaf spacing represents the first feature spacing, the leaf canopy reflection intensity represents the second feature reflection intensity, the cutting strategy represents the working strategy, the cutting height represents the working height, the blade rotation speed represents the device component rotation speed, and the blade wear degree represents the device component wear degree.

[0061] The spectral perception module is configured to acquire a vegetation image of the target area through the multispectral sensor, analyze the lawn density based on the vegetation image of the target area through a convolutional neural network, and calculate a normalized vegetation index in real time, generate a health degree matrix of the vegetation according to the lawn density and the normalized vegetation index, and monitor the soil humidity of the target area.

[0062] Further, as shown in Figure 2 The analysis sub-logic of the lawn density includes:

[0063] The vegetation image of the target area is acquired through the multispectral sensor, the vegetation image includes a visible light image and a near-infrared image, and the inclination data of the unmanned mower is monitored in real time to correct the geometric distortion of the vegetation image;

[0064] The corrected vegetation image is subjected to environmental compensation, a double channel is constructed, and the visible light image and the near-infrared image are subjected to pixel separation through threshold segmentation to extract the vegetation in the visible light image and the near-infrared image;

[0065] The visible light image is processed in the visible light channel through an edge detection algorithm to identify the leaf blade profile of the vegetation, and the leaf blade spacing is calculated to determine the physical density, while the near-infrared reflectivity is calculated in the near-infrared channel, and the leaf canopy layer reflectivity of the vegetation in the near-infrared image is determined to determine the physiological density;

[0066] The features of the double channel are extracted through a convolutional neural network, the spatial structure of the lawn is extracted in the visible light channel according to the physical density, the physiological information of the lawn is extracted in the near-infrared channel based on the physiological density, and the feature weights of the double channel are allocated based on an attention mechanism to fuse the features of the double channel to form a fused feature;

[0067] The lawn density is analyzed according to the fused feature, the fused feature is mapped to a density score, the density level is divided according to the density score through a clustering algorithm, and the difference between the physical density and the physiological density is verified to trigger an abnormality warning.

[0068] When the unmanned mower moves on the lawn, the vegetation images obtained by the multispectral sensor will produce geometric distortion due to the influence of the terrain and the change of the posture of the unmanned mower. If not corrected, the subsequent analysis of the lawn density will be deviated. At the same time, the vegetation images include visible light images and near-infrared images, which are the basic data sources for analyzing the physical structure and physiological state of the lawn. Therefore, the vegetation images need to be obtained and corrected first. A multispectral sensor is integrated on the unmanned mower, which can synchronously obtain the visible light images and near-infrared images of the target area. At the same time, an inertial measurement unit is installed to monitor the inclination data of the unmanned mower in real time, including the pitch angle and roll angle and other posture information. According to the inclination data, the geometric distortion correction of the vegetation images is carried out through the perspective transformation algorithm. Through the construction of the transformation matrix, the deformed images caused by the change of the posture of the unmanned mower are restored to the vegetation images under the normal viewing angle. Thus, the image distortion caused by the change of the posture of the unmanned mower is effectively eliminated, and the accuracy of the shape and position of the lawn vegetation in the vegetation images is ensured, providing a reliable data basis for the subsequent analysis of the lawn density.

[0069] The light intensity and other factors in the natural environment will affect the quality of the vegetation images, leading to the reduction of the distinguishability of the vegetation and the background such as soil in the images, thereby affecting the accuracy of the analysis of the lawn density. Therefore, environmental compensation is needed to enhance the image quality. At the same time, in order to accurately analyze the lawn density, the vegetation pixels and the background pixels such as soil in the vegetation images need to be separated out for subsequent analysis of the vegetation. First, the corrected vegetation images are analyzed for light, and the brightness and contrast of the vegetation images are adjusted through histogram equalization and other techniques, so that the vegetation images under different light conditions have consistent visual effects. In terms of pixel separation, a double channel is constructed, including a visible light channel and a near-infrared channel. The visible light images and the near-infrared images are processed respectively through the threshold segmentation algorithm.

[0070] For the visible light images, the green vegetation pixels are separated from the background pixels such as soil by setting a suitable color threshold based on the super green index ExG. For the near-infrared images, the vegetation and non-vegetation areas are distinguished by setting an index threshold based on the characteristics of the normalized difference vegetation index. The environmental compensation improves the quality of the vegetation images and enhances the contrast between the vegetation and the background, making the pixel separation more accurate. The vegetation pixels in the vegetation images are effectively extracted through threshold segmentation, and the interference of the background factors such as soil on the analysis of the lawn density is excluded. The accurately separated vegetation pixels provide pure data for the subsequent analysis of the physical density and physiological density of the lawn in the visible light channel and the near-infrared channel, ensuring the reliability of the density analysis results.

[0071] The turf density includes information of two levels of physical structure and physiological state. The physical density reflects the spatial distribution tightness of the turf vegetation, and can be determined by analyzing structural characteristics such as leaf spacing. The physiological density embodies the growth and health status of the turf vegetation, and is related to physiological indicators such as leaf canopy reflectance. The turf growth status can be comprehensively evaluated, and multi-dimensional data is provided for subsequent comprehensive analysis. In the visible light channel, the separated vegetation image is processed by an edge detection algorithm to identify the leaf outline of the vegetation, and the average distance between adjacent leaves, i.e., the leaf spacing, is calculated based on the information of the leaf outline, so as to determine the physical density of the turf. The smaller the leaf spacing, the higher the physical density of the turf. In the near-infrared channel, the near-infrared reflectance is calculated, and the physiological density is determined by analyzing the leaf canopy reflectance of the vegetation in the near-infrared image. The higher the leaf canopy reflectance, the better the physiological state of the turf vegetation, and the higher the physiological density. The turf density is analyzed from two angles of physical structure and physiological state to obtain more comprehensive turf growth information, so that the actual condition of the turf can be more accurately described, and the limitation of single-dimensional analysis can be avoided.

[0072] The turf density information in the visible light channel and the near-infrared channel has different characteristics, and the information of a single channel cannot fully reflect the overall condition of the turf. The features of the two channels are extracted by a convolutional neural network, and the advantages of the two channels are integrated by an attention mechanism, so that the key features can be highlighted, and the turf density can be more accurately analyzed. A convolutional neural network is constructed, and the spatial structure of the turf based on the physical density in the visible light channel and the physiological information of the turf based on the physiological density in the near-infrared channel are input into different branches of the convolutional neural network for feature extraction. In the feature extraction process, the convolutional neural network automatically learns and extracts features in the vegetation image through multiple convolution and pooling operations.

[0073] Then, based on the attention mechanism, different feature weights are assigned to the features of the two channels according to the importance of the features to the analysis of the turf density. When analyzing the area with more weeds, more attention is paid to the spatial structure features of the turf in the visible light channel, and higher feature weights are given to the features of this channel. When evaluating the health status of the turf, the physiological information of the near-infrared channel is more important, and the feature weight of the near-infrared channel is increased accordingly. Finally, the weighted feature weights are fused to form a fusion feature containing rich turf information. Thus, the advantages of the visible light and near-infrared channels are effectively integrated, the key information having important influence on the analysis of the turf density is highlighted, the expression ability of the features is enhanced, and the accuracy and reliability of the analysis of the turf density are improved.

[0074] The obtained fusion features need to be further analyzed to determine the density level of the lawn, provide a basis for the operation decision of the unmanned mower, and verify that the difference between the physical density and the physiological density can timely discover abnormal situations in the growth process of the lawn, including diseases and pests, and nutrient deficiency, so as to take corresponding measures for processing; the fusion features are normalized and input into a multilayer perception machine to convert the fusion features into a density score, and based on the density score, the lawn density is classified into levels by a clustering algorithm, and the lawn density is divided into three levels of sparse, medium and dense. After classifying the density levels, the physical density and the physiological density of each region are compared, and the difference between the two is calculated. When the difference is greater than a preset difference threshold, it is determined that there is an abnormal situation in the region, an abnormal early warning mechanism is triggered, and alarm information is sent to the operator to prompt further inspection and processing of the region.

[0075] Accurate classification of the lawn density level provides a clear reference for the cutting parameters and cutting priority allocation of the unmanned mower, so that the unmanned mower can reasonably operate according to the actual density of the lawn. By verifying the difference between the physical density and the physiological density, abnormal growth of the lawn can be timely discovered, which helps to quickly take measures to solve the problem and ensure the healthy growth of the lawn. The classification result of the lawn density directly affects the allocation of the cutting parameters and the cutting priority in the intelligent decision module, providing a decision basis for the efficient operation of the unmanned mower. The abnormal early warning information will trigger other related systems or manual intervention to optimize the management and maintenance process of the lawn.

[0076] Specifically, the generation logic of the health degree matrix of the vegetation includes:

[0077] The corrected vegetation image is subjected to band operation to calculate the normalized vegetation index in real time, and the normalized vegetation index is subjected to time and space calibration, and the normalized vegetation index is subjected to outlier filtering;

[0078] Based on the lawn density, the target region is divided into a plurality of grids, the normalized vegetation index and the lawn density in each grid are counted, and the level of the normalized vegetation index is determined to obtain an index level;

[0079] According to the density level and the index level, the health degree level of each grid is determined, and a risk flag is assigned to each grid based on the health degree level;

[0080] Based on the density level, the index level and the risk flag, a health degree matrix of the vegetation is generated.

[0081] The normalized difference vegetation index is an important indicator for measuring the growth status of vegetation. By calculating the normalized difference vegetation index, the health information of the lawn vegetation can be obtained. However, due to the influence of time variation and spatial difference, the normalized difference vegetation index may deviate, and there may be abnormal values in the normalized difference vegetation index, which may interfere with the accuracy of the health assessment. Therefore, it is necessary to calibrate the normalized difference vegetation index in time and space and filter the abnormal values to ensure the reliability of the data. After the corrected vegetation image is subjected to band operation, the normalized difference vegetation index is calculated. In the time calibration, the historical normalized difference vegetation index of the same period is referred to, and the time trend of the current normalized difference vegetation index is analyzed to correct the deviation caused by factors such as seasonal variation and diurnal difference. For example, in the early spring, the historical normalized difference vegetation index of the spring is combined to adjust the current data.

[0082] In the spatial calibration, the normalized difference vegetation index is subjected to spatial smoothing processing by the inverse distance weighted interpolation method to fill the data missing area caused by factors such as sensor coverage blind area or image occlusion, so that the normalized difference vegetation index is more continuous and accurate in space. For abnormal value filtering, the mean and standard deviation of the normalized difference vegetation index are calculated by statistical analysis method, and the data deviating from the mean by more than a certain multiple of the standard deviation is regarded as abnormal value and is removed or corrected. The time and space calibration makes the normalized difference vegetation index more accurately reflect the actual growth status of the lawn vegetation, eliminates the errors caused by time and space factors, and the abnormal value filtering removes the interference factors in the data, improves the quality of the normalized difference vegetation index, and provides reliable data support for the subsequent health assessment based on the normalized difference vegetation index.

[0083] In order to more finely evaluate the health status of the lawn, the target area needs to be divided into multiple grids for data statistics and analysis. The normalized difference vegetation index and lawn density in each grid are calculated, and the normalized difference vegetation index grade is determined to obtain the growth information of the lawn in different local areas, which provides data basis for determining the health grade of each grid. Based on the analysis result of the lawn density, the target area is divided into multiple grids of appropriate size. The normalized difference vegetation index of all vegetation pixels contained in each grid is calculated, the average normalized difference vegetation index of the grid is calculated, and the normalized difference vegetation index grade of the grid is determined according to the range of the normalized difference vegetation index, which is divided into three intervals of low, medium and high. At the same time, the density grade in each grid is calculated. Through the grid division, the lawn is finely managed and analyzed, the growth difference of the lawn in different local areas can be more accurately understood, the index grade and density grade of each grid are calculated, which provides detailed data for the subsequent comprehensive evaluation of the health grade of each grid, and helps to find the local problems in the growth process of the lawn.

[0084] According to the turf density and the normalized vegetation index of each grid, the health degree grade of each grid is determined, the growth and health condition of the turf in each grid can be directly reflected, a risk mark is assigned to each grid, the potential risks faced by the turf in different regions are determined, and targeted management measures are taken; an evaluation system of the health degree grade is established, the density grade and the index grade are comprehensively considered to determine the health degree grade of each grid, when the density grade is high and the index grade is also high, the health degree grade of the grid is determined as excellent, when the density grade is low and the index grade is also low, the health degree grade is determined as poor; according to the health degree grade, a corresponding risk mark is assigned to each grid, when the health degree grade of the grid is excellent, a green risk mark is assigned, which indicates that the turf in the region grows well, when the health degree grade of the grid is poor, a red risk mark is assigned, which indicates that there are serious problems in the turf in the region and the region needs to be focused on and treated, and the rest are assigned with a yellow risk mark; thus, the health degree grade of each grid is clearly divided, the growth condition of the turf in different regions can be quickly understood by the operator, the reasonable assignment of the risk mark helps to highlight the region to be focused on, and targeted turf management strategies are facilitated, and the efficiency and effect of turf management are improved.

[0085] The density grade, the index grade and the risk mark are integrated to generate the health degree matrix, the health condition of the turf in the target region can be comprehensively displayed in a structured form, and intuitive and detailed data support is provided for intelligent decision-making of the control system; taking the grid as a basic unit, the density grade, the index grade and the risk mark information of each grid are encoded, a multi-dimensional matrix is constructed according to the position order of the grid in the target region, each element of the matrix corresponds to a grid, and the related information code of the grid is stored, in this way, a complete vegetation health degree matrix is formed, and the spatial distribution characteristics of the turf health condition in the target region can be clearly presented.

[0086] The health degree matrix integrates the multi-dimensional information related to the turf health in an intuitive form, facilitates the quick query and analysis of the control system, and provides a comprehensive and accurate data basis for the intelligent decision-making module, the matrix information can be used to make reasonable allocation of the cutting parameters and the cutting priority and the trigger judgment of the external equipment, realize the intelligent and precise operation of the unmanned mower, the generated health degree matrix is an important basis for the decision-making of the intelligent decision-making module, directly affects the subsequent operation processes such as the operation mode of the unmanned mower and the working state of the external equipment, and realizes the efficient operation of the entire control system.

[0087] The intelligent decision-making module is used for allocating the cutting parameters and the cutting priority of the unmanned mower based on the health degree matrix of the vegetation, to generate a control signal, and comprehensively judging the normalized vegetation index and the soil humidity of the target region to trigger the external equipment.

[0088] Further, the allocation of the cutting parameters and the cutting priority includes:

[0089] The health matrix of the vegetation is processed by a density clustering algorithm to identify the health status of each grid, and the grids with the same health status are combined to form a state grid block, and a cutting strategy is assigned to each state grid block;

[0090] The cutting parameters of the unmanned mower are assigned to each state grid block based on the cutting strategy, including the cutting height, blade speed and travel speed;

[0091] A cutting priority is assigned to each state grid block according to the attributes of the state grid block, and the obstacles in each state grid block are monitored.

[0092] The health status of different areas of the lawn is different, and if a uniform cutting method is used, it will cause over-pruning or insufficient pruning, affecting the health and appearance of the lawn, so the health matrix of the vegetation needs to be processed to identify the health status of each grid, and the grids with the same health status are combined to form a state grid block, so that different state lawn areas can be assigned a targeted cutting strategy; the health matrix of the vegetation is analyzed by a density clustering algorithm, which can automatically aggregate grids with similar health status according to the similarity of the density level, index level and risk flag between grids, and divide grids with high density level, high index level and no risk flag into healthy grid blocks, and divide grids with low density level, low index level and red risk flag into problem grid blocks, and the state grid block also includes a regular grid block, by setting appropriate distance threshold and density threshold during clustering, to ensure that the state grid block divided is consistent with the actual growth conditions of the lawn and has a certain scale for subsequent processing; thereby realizing the fine classification of lawn areas, so that the control system can clearly distinguish between lawn areas with different health status to accurately develop a cutting strategy.

[0093] Different health status of the lawn area has different needs for cutting, in order to ensure the healthy growth and good landscape effect of the lawn, appropriate cutting strategy needs to be assigned to the state grid block according to its health status to guide the unmanned mower to adopt the correct operation mode; different cutting strategies are assigned to different types of state grid blocks, for healthy grid blocks, a light pruning strategy is selected to maintain the appearance and health of the lawn and avoid excessive pruning damage to the lawn, for problem grid blocks, a targeted treatment strategy is implemented to focus on cleaning weeds and repairing degraded lawns, and for regular grid blocks, a standard pruning strategy is implemented to prune according to the regular operation requirements, when assigning the cutting strategy, the control system will assign the most suitable cutting strategy according to the detailed attribute information of each state grid block; so that the unmanned mower can operate according to the actual needs of the lawn, improving the scientificity and effectiveness of lawn maintenance, which helps to promote the healthy growth of the lawn and improve the overall quality of the lawn.

[0094] The cutting strategy needs to be realized by specific cutting parameters, different cutting strategies correspond to different combinations of cutting height, blade rotation speed and travel speed, and reasonable allocation of these parameters can ensure that the unmanned mower completes the cutting task efficiently and accurately according to the cutting strategy; according to the allocated cutting strategy, a parameter mapping relationship is established, when the cutting strategy is light pruning, the cutting height is set to a higher value to protect the normal growth of the lawn, the blade rotation speed is set to a lower value to reduce the damage to the lawn, and the travel speed is appropriately accelerated to improve the work efficiency; while when the cutting strategy is target processing, in the area with more weeds, the cutting height is lowered to completely remove the weeds, the blade rotation speed is increased to enhance the cutting ability, and the travel speed is slowed down to ensure the cutting effect; the control system will automatically select the corresponding parameter value from the parameter mapping relationship according to the cutting strategy of each state grid block for allocation; thereby realizing the accurate matching of cutting parameters and cutting strategies, enabling the unmanned mower to carry out fine work according to the actual situation of the lawn, improving the mowing quality and efficiency, and reducing energy consumption and equipment wear and tear.

[0095] The importance and emergency treatment demand of different areas on the lawn are different, and the areas with obstacles need special treatment, reasonable allocation of cutting priority and monitoring of obstacles can ensure that the unmanned mower processes the key areas first and avoids colliding with obstacles, ensuring safe and efficient operation; according to the properties of the state grid block, the cutting priority is determined, among which the problem grid block is set to high priority and is processed first to improve the poor condition of the lawn as soon as possible, and the regular grid block is set to low priority and is processed after the high priority area is processed.

[0096] For obstacle monitoring, sensors such as laser radar and camera are installed on the unmanned mower to scan the environmental information in each state grid block in real time, identify the position, shape and size of the obstacle, and when the obstacle is detected, the control system will automatically mark the area and avoid these areas when allocating cutting priority and planning travel path; so that the unmanned mower can orderly cut the lawn and solve the problem areas of the lawn first, improving the pertinence and efficiency of lawn maintenance, while obstacle monitoring effectively ensures the safety of the mower, reducing the risk of equipment damage, and cutting priority and obstacle information will directly affect the travel path planning in the control signal generation process to plan the optimal mowing path and ensure efficient and safe completion of the work task.

[0097] Specifically, the generation logic of the control signal includes:

[0098] Based on the cutting parameters, the motor energy consumption demand of the unmanned mower is determined, and according to the lawn density and cutting priority of each state grid block, the motor energy consumption demand of each state grid block is adjusted;

[0099] According to the current battery power of the unmanned mower and the historical energy consumption data of each motor, the total power threshold that can be allocated is calculated, the motor power is allocated in combination with the motor energy consumption demand of each state grid block, and the motor power is adjusted according to the lawn density;

[0100] Taking the center of the state grid block as the target position, the travel path of the unmanned mower is planned in combination with the current position of the unmanned mower, the lawn slope, the lawn density and the motor energy consumption, and the travel path of the unmanned mower is optimized based on the obstacles of each state grid block to generate a control signal.

[0101] Different cutting parameters and lawn density and cutting priority will result in different motor energy consumption demands of the unmanned mower when working in different state grid blocks. Accurate judgment and adjustment of motor energy consumption demand can ensure that the mower meets the working requirements and reasonably allocates energy to avoid energy waste or insufficient supply. The control system first preliminarily judges the motor energy consumption demand of the unmanned mower when working in each state grid block based on the allocated cutting parameters and historical data. Preferably, higher blade speed and lower cutting height will increase the motor load and correspondingly increase the energy consumption demand.

[0102] Then, the energy consumption demand preliminarily judged is adjusted according to the lawn density and cutting priority of each state grid block. In high-density lawn areas, the energy consumption demand is increased to ensure the cutting effect. For high-priority areas, the energy supply is prioritized, and if necessary, the energy consumption demand of other areas can be appropriately reduced. In this way, the motor energy consumption demand of each state grid block is dynamically adjusted to better meet the actual working conditions. Thus, precise estimation and flexible adjustment of motor energy consumption demand are realized, energy utilization efficiency is improved, the endurance time of the unmanned mower is prolonged, and the unmanned mower can stably operate under different working conditions.

[0103] The battery power of the unmanned mower is limited, and the power demands of different motors are different during operation. According to the current battery power, historical energy consumption data and motor energy consumption demand of each state grid block, the motor power is reasonably allocated to ensure normal operation of the mower while avoiding battery overload or premature power depletion. The control system first calculates the total power threshold that can be allocated according to the current battery power of the unmanned mower and the historical energy consumption data of each motor, and then allocates the motor power according to the priority and demand ratio in combination with the adjusted motor energy consumption demand of each state grid block.

[0104] For high-priority and high-energy-demand areas, sufficient power is allocated first, and for low-priority areas, power allocation is appropriately reduced under the premise of meeting basic operation requirements. Meanwhile, considering the power characteristics of different motors, they are allocated differently. When climbing or processing high-density lawns, the power ratio of the driving motor is appropriately increased to ensure the driving power of the mower. When performing fine cutting, the power of the cutting motor is increased. Thus, the scientific and reasonable allocation of motor power is realized, the battery capacity is fully utilized, the working efficiency and endurance of the unmanned mower are improved, and the problem of performance degradation or battery life shortening caused by unreasonable power allocation is avoided.

[0105] To enable the unmanned mower to efficiently and safely complete the cutting task, the center of the state grid block is taken as the target position, the current position of the mower, the slope of the lawn, the density of the lawn, the motor energy consumption, and obstacles are comprehensively considered, the optimal travel path is planned, and the corresponding control signal is generated to drive the mower to run according to the planned path. Through the path planning algorithm, the center of the state grid block is taken as the target node, and the current position of the unmanned mower is taken as the starting node. The path search graph is constructed. In the search process, the factors such as the slope of the lawn, the density of the lawn, and the motor energy consumption are converted into the weight parameters of the path cost function. That is, the area with a larger slope of the lawn will increase the path cost, so that the path planning algorithm will try to avoid selecting this path. The area with high energy consumption demand will also correspondingly increase the path cost, guiding the path planning algorithm to select a more energy-saving path.

[0106] Meanwhile, based on the obstacle information of each state grid block, the travel path is optimized to ensure that the path avoids obstacles. After the optimal travel path is planned, the control system converts the path information and cutting parameters into control signals, including signals for controlling motor speed and steering angle, and sends them to the unmanned mower through wireless communication or wired connection to drive the mower to perform the cutting task according to the planned path. The planned travel path can balance the operation efficiency, energy consumption, and safety, so that the unmanned mower can efficiently operate in complex lawn environments, reduce invalid travel and energy waste, avoid colliding with obstacles, and ensure the safety of equipment and personnel. The generated control signal accurately conveys the operation instruction, ensuring that the unmanned mower cuts according to the predetermined path and parameters. The generated control signal will be transmitted to the control feedback module to drive the unmanned mower to operate. Meanwhile, the control feedback module will monitor the operating state of the mower in real time and feedback and adjust the control signal according to the actual situation, forming a closed-loop control system to ensure the stability and accuracy of the operation of the unmanned mower.

[0107] According to the normalized vegetation index and the soil moisture of the target area, the external equipment is triggered, including irrigation devices and fertilization devices;

[0108] The priority of the external device is judged, and when multiple external devices trigger at the same time, the triggering of the external devices is cooperatively scheduled based on the priority of the external devices.

[0109] The growth of the lawn requires appropriate moisture and nutrients, and whether the lawn needs irrigation or fertilization and other interventions of external devices can be determined according to the normalized difference vegetation index and soil moisture, and the external devices are triggered in time to provide good growth conditions for the lawn and promote the healthy growth of the lawn; the control system obtains the normalized difference vegetation index and soil moisture monitored by the spectral sensing module in real time, and when the normalized difference vegetation index is less than a preset index threshold and the soil moisture is also less than a set moisture value, it is determined that the lawn is in a water and fertilizer deficient state, and the irrigation device and the fertilization device are triggered; when the normalized difference vegetation index is within a normal range but the soil moisture is continuously low, only the irrigation device is triggered, and the control system will continuously compare the real-time data with the conditions in the rule base, and once the triggering condition is met, an external device triggering signal is immediately generated; thereby realizing automatic triggering of external devices according to the actual growth needs of the lawn, avoiding the subjectivity and hysteresis of manual judgment, improving the timeliness and accuracy of lawn maintenance, and helping to maintain the good growth state of the lawn.

[0110] When multiple external devices meet the triggering condition at the same time, in order to reasonably arrange the working order of the devices and avoid resource conflicts and waste, the priority of the external devices needs to be judged to ensure that the key devices are prioritized to run and improve the device use efficiency and the lawn maintenance effect; the priority of the external devices is formulated, and the priority of the irrigation device is set to be higher than that of the fertilization device, because water is the basic requirement for the growth of the lawn; the control system obtains the corresponding priority information from the priority rules according to the type of the triggered external device, and prioritizes the devices triggered at the same time; thereby the working order of the external devices is clear, the problems of resource shortage and mutual interference caused by simultaneous operation of the external devices are avoided, the external devices can work in order, and the overall use efficiency of the devices is improved, better meeting the needs of lawn maintenance.

[0111] In order to fully play the role of the external devices and improve the comprehensive effect of lawn maintenance, the multiple external devices need to be cooperatively scheduled based on the priority when they are triggered, the working time and resources of the devices are reasonably allocated, and the lawn is ensured to be comprehensively and effectively maintained; a cooperative scheduling strategy is formulated according to the priority of the external devices, when the irrigation device and the fertilization device are triggered at the same time, the irrigation device is started first, and after the soil moisture of the lawn reaches a certain level, the fertilization device is started to ensure that the fertilizer can be better absorbed by the lawn; in the scheduling process, the control system will monitor the working state of the external devices and the environmental changes of the lawn in real time, and dynamically adjust the working parameters and working time of the devices; wherein during irrigation, if it is detected that it is raining, the irrigation device is suspended, and the start time of the subsequent fertilization device is re-evaluated, and the working order and connection between different external devices are coordinated to avoid conflicts and interference between the devices.

[0112] Thus, efficient cooperation of external devices is realized, the utilization rate of device resources is improved, the lawn can be properly maintained at the appropriate time, the overall effect of lawn maintenance is improved, the healthy growth of the lawn is ensured, and the result of the cooperative scheduling of the external devices directly affects the growth condition of the lawn, and the change of the growth condition of the lawn is fed back to the spectrum sensing module, affecting the subsequent health matrix updating and intelligent decision-making process, forming a complete closed-loop management system for lawn maintenance.

[0113] The control feedback module is configured to receive a control signal to drive the unmanned mower, monitor the actual motor torque of the unmanned mower in real time, and determine the target motor torque of the unmanned mower according to the lawn slope and the lawn density, compare the actual motor torque with the target motor torque to trigger local rescan of the vegetation image, and update the health matrix of the vegetation.

[0114] Further, the determination sub-logic of the target motor torque comprises:

[0115] The control signal is received to drive the unmanned mower, the lawn slope is monitored, and the driving mode of the unmanned mower is determined according to the lawn slope, the driving mode including a standard mode and a compensation mode;

[0116] The initial motor torque of the unmanned mower is determined based on the driving mode, and the initial motor torque is corrected according to the lawn density;

[0117] The degree of blade wear of the unmanned mower is identified in real time through voiceprint analysis, and the corrected initial motor torque is compensated based on the degree of blade wear to determine the target motor torque.

[0118] The lawn slope will significantly affect the driving resistance and working load of the unmanned mower. If the same motor torque control mode is used under different slopes, it will cause the mower to be insufficient in power or waste energy. Therefore, the driving mode of the mower needs to be determined according to the lawn slope to provide a basis for the reasonable setting of the motor torque. A slope sensor is installed on the unmanned mower to monitor the lawn slope of the working area of the unmanned mower in real time. A slope threshold is set. When the slope sensor detects that the slope is less than or equal to the slope threshold, it is determined as the standard mode. At this time, the unmanned mower is in a normal working environment, and the driving resistance is small. When the slope is greater than the slope threshold, it is determined as the compensation mode, indicating that the mower is in a special terrain such as climbing or descending, and additional power compensation or brake control is needed. The driving mode determination result is transmitted to the control unit by the control system as the basis for determining the initial motor torque. By accurately determining the driving mode, the unmanned mower can adapt to different slope lawn environments, avoid power redundancy in the standard mode, and ensure working power in the compensation mode, thereby improving the environmental adaptability and energy utilization efficiency of the mower.

[0119] The initial motor torque provides the basis power for the mower operation, and the lawn density further affects the load of the cutting operation, so the initial torque needs to be corrected according to the lawn density to ensure that the mower can effectively operate on lawns of different densities, and to avoid poor cutting effect due to insufficient torque or energy waste due to excessive torque; based on the determined driving mode, the control system retrieves the corresponding initial motor torque value from the preset torque parameter table, in the standard mode, the initial torque is set to the basic value that meets the regular lawn cutting and driving requirements; in the compensation mode, the initial torque will be appropriately increased according to the slope size to cope with greater driving resistance.

[0120] After retrieving the initial torque, the lawn density information provided by the spectrum sensing module is used for correction, for high-density lawn areas, the initial torque is increased by a certain percentage to enhance the cutting ability; for low-density lawn areas, the initial torque is appropriately reduced to reduce energy consumption, that is, when the lawn is high-density and in the compensation mode, a certain percentage of torque value is added to the initial torque in the compensation mode; thereby realizing the preliminary matching of the motor torque and the lawn operation environment, so that the mower can obtain appropriate power output under different slope and lawn density conditions, improve the stability and efficiency of the mowing operation, and reduce unnecessary energy consumption.

[0121] Blade wear can reduce its cutting efficiency and increase the load of the motor, if not compensated, it will affect the mowing quality and equipment life, so the degree of blade wear needs to be identified in real time through voiceprint analysis, and based on this, the corrected initial motor torque is compensated to determine the final target motor torque, ensuring that the mower operates continuously, stably and efficiently; a voiceprint sensor is installed near the blade driving part of the unmanned mower to obtain the sound signal generated during the blade cutting operation in real time, the obtained sound signal is analyzed, the difference between the sound characteristics of the normal blade operation and the current sound signal is compared, the degree of blade wear is judged, and the degree of wear is divided into different levels such as light, medium and heavy.

[0122] For different wear grades, corresponding torque compensation strategies are set. When the wear is slight, a small proportion of the modified initial torque is compensated for; when the wear is moderate, the compensation proportion is increased accordingly; when the wear is severe, in addition to a large proportion of torque, a blade replacement prompt is triggered. After blade wear compensation, the final target motor torque is determined, and the target motor torque is taken as the control target for the operation of the mower motor; thereby effectively addressing the impact of blade wear on the working performance of the mower, maintaining the cutting efficiency and working quality of the mower through real-time torque compensation, prolonging the service life of the blade and the motor, reducing equipment maintenance costs, and determining the target motor torque as a standard for judging whether the actual motor torque is normal, for comparison with the actual motor torque, and then triggering subsequent operations such as local rescan of the vegetation image, to realize closed-loop control of the working state of the mower.

[0123] Specifically, as shown in Figure 3 The triggering conditions of the local rescan of the vegetation image include:

[0124] The torque difference between the actual motor torque and the target motor torque is calculated, and a difference threshold is configured, including a first difference and a second difference;

[0125] When the torque difference is greater than the first difference and less than or equal to the second difference, a single-point rescan of the vegetation image is triggered;

[0126] When the torque difference is greater than the second difference, a regional rescan of the vegetation image is triggered;

[0127] After triggering the single-point rescan and the regional rescan of the vegetation image, the lawn slope is verified to determine the target motor torque of the unmanned mower, and it is judged whether the vegetation image needs to be scanned again.

[0128] The difference between the actual motor torque and the target motor torque can reflect whether the lawn mower operation state is normal. By calculating the torque difference and comparing it with the configured difference threshold, it can be determined whether the vegetation image needs to be locally rescanned to timely discover the lawn condition changes or equipment abnormalities, ensuring the accuracy and effectiveness of the mowing operation. During the operation of the unmanned lawn mower, the control feedback module obtains the actual motor torque in real time and calculates the difference with the target motor torque. Meanwhile, the control system pre-configures two difference thresholds, i.e., the first difference and the second difference, with the first difference being smaller than the second difference. The first difference is used to determine whether single-point rescanning is needed, mainly capturing smaller torque abnormalities caused by local lawn condition changes or slight equipment fluctuations. The second difference is used to determine whether regional rescanning is needed, aiming at larger torque abnormalities indicating serious lawn problems or equipment failures. These difference thresholds can be flexibly adjusted and set according to the model, performance, and actual operation environment of the lawn mower. The clear torque difference calculation and threshold configuration provide a quantitative basis for determining whether to trigger vegetation image rescanning, enabling the control system to accurately identify abnormal conditions in the lawn mower operation and avoid resource waste caused by blind rescanning, while ensuring timely discovery of potential problems.

[0129] When the difference between the actual motor torque and the target motor torque reaches different levels, it indicates that the lawn condition or equipment operation state has different degrees of abnormalities. By triggering single-point rescanning or regional rescanning, more accurate vegetation image information of the abnormal area can be obtained to analyze the problem in depth and timely adjust the operation parameters of the lawn mower or perform equipment maintenance.

[0130] If the torque difference is greater than the second difference, the control system determines that there is a more serious abnormal condition and triggers regional rescanning. The unmanned lawn mower controls the multi-spectral sensor to comprehensively acquire images of a larger area centered on the torque abnormal point. The image acquisition range is dynamically determined according to the abnormality degree and the actual lawn condition, i.e., covering multiple state grid blocks. During the triggering of rescanning, the control system will pause or adjust the operation of the lawn mower in the area to ensure the accuracy of image acquisition. According to the different levels of torque difference, rescanning is triggered in stages, achieving accurate positioning and targeted detection of lawn abnormal conditions, improving the efficiency and accuracy of problem troubleshooting, and helping to quickly discover lawn growth abnormalities or equipment failures, providing reliable basis for timely measures.

[0131] After triggering the vegetation image rescan, the vegetation image obtained by rescan needs to be analyzed to reevaluate the lawn condition and determine whether the current target motor torque is still appropriate and whether rescan needs to be performed again to ensure that the mower is always in the best working state and effectively responds to the dynamic changes in the lawn environment; after the rescan is completed, the control system processes and analyzes the newly obtained vegetation image, reanalyzes the lawn density and normalized vegetation index and other parameters, and verifies the lawn slope information, based on these reevaluated data, determines whether the current target motor torque can meet the actual working requirements, if the evaluation result shows that the target motor torque still needs to be adjusted, the control system will determine the target motor torque again according to the new lawn condition and device running state, at the same time, the control system will analyze the trend and stability of the torque change, if it is found that the torque fluctuation is still large or there is a continuous abnormality and cannot be solved by adjusting the target motor torque, the local rescan of the vegetation image will be triggered again to further investigate the problem.

[0132] Through the evaluation and rescan judgment mechanism after rescan, dynamic monitoring and continuous optimization of the working state of the unmanned mower are realized, which ensures that the unmanned mower can adapt to the changes in the lawn environment, discovers and solves potential problems in time, improves the quality of mowing and the stability of device operation, and the results of evaluation and rescan judgment after rescan will determine whether the health degree matrix of vegetation needs to be updated and whether information needs to be fed back to the intelligent decision module to prompt the intelligent decision module to adjust the cutting parameters, cutting priority and other decisions, forming a closed-loop control process of the working state of the unmanned mower.

[0133] Specifically, the update logic of the health degree matrix includes:

[0134] When the local rescan of the vegetation image is performed, the vegetation image of the target area is reacquired to determine the torque change type, and the torque change type includes physical change and physiological change;

[0135] If it is a physical change, the change characteristics of the leaf profile are identified to reextract the spatial structure of the vegetation;

[0136] If it is a physiological change, the leaf canopy layer reflectance intensity of the vegetation is updated to reevaluate the risk flag of each grid in the target area;

[0137] The change rates of the spatial structure and the risk flag after three local rescan are compared to dynamically update the health degree matrix.

[0138] After performing the local rescan of the vegetation image, it is necessary to determine the cause of the torque change and distinguish whether it is caused by changes in the physical structure or physiological state of the lawn. Different types of torque changes require different processing methods to accurately update the health matrix and provide reliable basis for subsequent mowing decisions of the unmanned mower. The control system performs multi-dimensional analysis on the vegetation image obtained by rescan, extracts the physical structure characteristics and physiological state indicators of the lawn, and determines the type of torque change by comparing the changes in these characteristics and indicators before and after rescan. If abnormal deformation of the leaf profile or sudden appearance of foreign matter in the lawn is found, and there is no significant change in the normalized vegetation index and other physiological indicators, it is determined to be a physical change. If the normalized vegetation index decreases significantly and the leaf canopy reflectance decreases, but there is no obvious abnormality in the physical structure, it is determined to be a physiological change. Thus, the type of torque change can be accurately determined, which helps the control system to take appropriate processing strategies for different reasons, avoids decision-making errors caused by incorrect analysis, improves the accuracy and reliability of the health matrix update, and better guides the mowing of the mower and the maintenance and management of the lawn.

[0139] If the torque change is a physical change, re-extracting the spatial structure of the vegetation can accurately reflect the change in the physical state of the lawn and provide a basis for adjusting the mowing strategy of the mower. If it is a physiological change, updating the leaf canopy reflectance of the vegetation and re-evaluating the risk flag can help to timely grasp the changes in the health status of the lawn so as to take targeted maintenance measures. When it is determined to be a physical change, the control system uses edge detection and image segmentation algorithms to accurately identify the leaf profile of the vegetation in the rescan image, re-extract the spatial structure of the vegetation, and update the relevant information of the corresponding grid in the health matrix.

[0140] If it is determined to be a physiological change, the control system re-calculates the leaf canopy reflectance of the vegetation according to the rescan image, combines with the normalized vegetation index and other physiological indicators, and re-evaluates the risk flag of each grid in the target area according to the pre-set evaluation rules. When the leaf canopy reflectance and the normalized vegetation index decrease significantly, the risk flag of the corresponding grid is updated from green to yellow or red. Different types of torque changes are processed accordingly, so that the health matrix can timely and accurately reflect the actual condition of the lawn, and whether it is a physical structure change or a physiological state change can be effectively recorded and updated, providing more accurate data support for intelligent decision-making of the mower.

[0141] By comparing the change rates of the spatial structure and the risk sign of the lawn after multiple local rescans, the dynamic evolution trend of the lawn condition can be comprehensively understood, and the misjudgment caused by single data fluctuation can be avoided. Based on the change rate, the health matrix is dynamically updated, so that the matrix information can always match the actual change of the lawn, and accurate and real-time decision basis is provided for the control system. The control system records the spatial structure information and the risk sign of the lawn updated after each local rescan. When three local rescans are completed, the change rates of the spatial structure and the risk sign between the adjacent two rescans are calculated, and a change rate threshold is set. When the change rate of the spatial structure or the risk sign is greater than the change rate threshold, the control system considers that the lawn condition has changed significantly, and the health matrix is dynamically updated comprehensively. The latest spatial structure information, risk sign, and related lawn density and normalized vegetation index information are integrated into the health matrix. If the change rate is not greater than the change rate threshold, it is considered that the lawn condition is relatively stable, and only part of the detailed information in the health matrix is adjusted or remains unchanged.

[0142] Thus, dynamic and accurate updating of the health matrix is realized, which can timely capture the subtle changes and long-term evolution trend of the lawn condition, provide real-time and reliable data basis for intelligent decision of the unmanned mower, help the unmanned mower to plan the operation strategy more scientifically, and improve the effect and efficiency of lawn maintenance. The updated health matrix is fed back to the intelligent decision module, the intelligent decision module reallocates the cutting parameters and cutting priority based on the new matrix information, generates a new control signal, and drives the unmanned mower to perform operation more in line with the current lawn condition, forming a complete closed-loop management process from data acquisition, analysis to decision execution.

[0143] Embodiment 2

[0144] As shown in Figure 4 a method flowchart of a mobile device control method is provided for the embodiments of the present application. The method comprises:

[0145] The vegetation image of the target area is acquired by the multispectral sensor, and the lawn density is analyzed based on the vegetation image of the target area by the convolutional neural network;

[0146] The normalized vegetation index is calculated in real time, and the health matrix of the vegetation is generated according to the lawn density and the normalized vegetation index, and the soil humidity of the target area is monitored;

[0147] The cutting parameters and cutting priority of the unmanned mower are allocated based on the health matrix of the vegetation to generate a control signal;

[0148] The external device is triggered according to the normalized vegetation index and the soil humidity of the target area;

[0149] The control signal is received to drive the unmanned mower, the actual motor torque of the unmanned mower is monitored in real time, and the target motor torque of the unmanned mower is determined comprehensively according to the lawn slope and the lawn density;

[0150] The actual motor torque and the target motor torque are compared to trigger local rescan of the vegetation image, and the health matrix of the vegetation is updated.

[0151] Since the principle of the method in the embodiment of the application solves the problem is similar to the system described above, the implementation of the method refers to the implementation of the system, and the repeated parts will not be repeated.

Claims

1. A mobile device control system, characterized in that: include: Spectral perception module, intelligent decision-making module and control feedback module; The spectral perception module is used to obtain a target image of the target area through a multispectral sensor. Based on the target image of the target area, the sub-target density is analyzed through a convolutional neural network, and the normalized target index is calculated in real time. The target health matrix is ​​generated based on the sub-target density and normalized target index, and the soil moisture in the target area is monitored at the same time. The intelligent decision-making module is used to allocate the working parameters and working priorities of the mobile devices based on the target health matrix to generate control signals, and trigger external devices based on the normalized target index and the soil moisture in the target area. The control feedback module is used to receive control signals to drive the mobile device, monitor the actual motor torque of the mobile device in real time, and comprehensively determine the target motor torque of the mobile device based on the sub-target slope and sub-target density. The actual motor torque is compared with the target motor torque to trigger a partial rescan of the target image and update the target health matrix.

2. The mobile device control system according to claim 1, wherein: The generation logic of the health matrix of the target includes: Perform band operations on the corrected target image to calculate the normalized target index in real time, perform spatiotemporal calibration on the normalized target index, and filter outliers on the normalized target index. The target area is divided into multiple grids based on the sub-target density, the normalized target index and sub-target density in each grid are counted, and the level of the normalized target index is determined to obtain the index level; Determine the health level of each grid based on the density level and index level, and assign a risk flag to each grid based on the health level; Generate a health matrix of targets based on density level, index level and risk indicator.

3. The mobile device control system according to claim 2, wherein: The sub-logic of analyzing the sub-target density includes: The target image of the target area is acquired through a multispectral sensor. The target image includes a visible light image and a near-infrared image, and the tilt data of the mobile device is monitored in real time to correct the geometric distortion of the target image. Perform environmental compensation on the corrected target image, construct a dual-channel, and perform pixel separation on the visible light image and near-infrared image through threshold segmentation to extract the target in the visible light image and near-infrared image; Processing the visible light image by an edge detection algorithm in the visible light channel to identify a first characteristic outline of the target and calculating a first characteristic spacing to determine the physical density, while calculating the near-infrared reflectivity in the near-infrared channel and determining a second characteristic reflection intensity of the target in the near-infrared image to determine the physiological density; A convolutional neural network is used to extract dual-channel features. The spatial structure of the sub-target is extracted based on the physical density in the visible light channel, and the physiological information of the sub-target is extracted based on the physiological density in the near-infrared channel. The feature weights of the two channels are assigned based on the attention mechanism to fuse the features of the two channels to form a fusion feature. The sub-target density is analyzed based on the fusion features, and the fusion features are mapped into density scores. The density levels are divided by the clustering algorithm based on the density scores, and the difference between physical density and physiological density is verified to trigger abnormal warnings.

4. The mobile device control system according to claim 3, wherein: The control signal generation logic includes: Determine the motor energy consumption requirements of the mobile device based on the working parameters, and adjust the motor energy consumption requirements of each state grid block according to the sub-goal density and work priority of each state grid block; Based on the current battery level of the mobile device and the historical energy consumption data of each motor, the total power threshold that can be allocated is calculated. The motor power is allocated based on the motor energy consumption requirements of each state grid block, and the motor power is adjusted according to the sub-goal density. Taking the center of the state grid block as the target position, the mobile device's travel path is planned based on the current position of the mobile device, sub-target slope, sub-target density, and motor energy consumption. The mobile device's travel path is optimized based on the obstacles in each state grid block and a control signal is generated.

5. The mobile device control system according to claim 4, wherein: The sub-logic of allocating the working parameters and working priorities includes: The target health matrix is ​​processed by a density clustering algorithm to identify the health status of each grid, grids with the same health status are merged to form a state grid block, and a working strategy is assigned to each state grid block; Based on the working strategy, the working parameters of the mobile device are assigned to each state grid block. The working parameters include working height, rotation speed of the equipment components and travel speed. The working priority is assigned to each state grid block according to the attributes of the state grid block, and obstacles of each state grid block are monitored.

6. The mobile device control system according to claim 5, wherein: Triggering external devices based on a comprehensive judgment of the normalized target index and the soil moisture in the target area; Determine the priority of the external device. When multiple external devices are triggered at the same time, coordinate and schedule the triggering of the external devices based on their priority.

7. The mobile device control system according to claim 6, wherein: The triggering conditions for the partial rescanning of the target image include: Calculating a torque difference between an actual motor torque and a target motor torque, and configuring a difference threshold, where the difference threshold includes a first difference and a second difference; When the torque difference is greater than the first difference and less than or equal to the second difference, triggering a single-point rescan of the target image; When the torque difference is greater than the second difference, triggering a rescan of the target image area; After triggering the single-point rescan and area rescan of the target image, the sub-target slope is verified to determine the target motor torque of the mobile device and to determine whether the target image needs to be scanned again.

8. The mobile device control system according to claim 7, wherein: The target motor torque determination sub-logic includes: receiving a control signal to drive the mobile device, monitoring a sub-target slope and determining a driving mode of the mobile device according to the sub-target slope, the driving mode including a standard mode and a compensation mode; Determine the initial motor torque of the mobile device based on the driving mode, and modify the initial motor torque according to the sub-target density; The degree of wear of the device components of the mobile device is identified in real time through voiceprint analysis, and the initial motor torque is compensated and corrected based on the degree of wear of the device components to determine the target motor torque.

9. The mobile device control system according to claim 8, wherein: The update logic of the health matrix includes: After performing a partial rescan of the target image, a target image of the target area is reacquired to determine a torque change type, where the torque change type includes a physical change and a physiological change; If it is a physical change, identifying the change characteristics of the first feature contour to re-extract the spatial structure of the target; If it is a physiological change, the second characteristic reflection intensity of the target is updated to re-evaluate the risk sign of each grid in the target area; The change rates of spatial structure and risk markers after three local rescans are compared to dynamically update the health matrix.

10. A mobile device control method, implemented based on the mobile device control system according to any one of claims 1 to 9, characterized in that: include: A target image of the target area is acquired through a multispectral sensor, and sub-target density is analyzed through a convolutional neural network based on the target image of the target area; Calculate the normalized target index in real time, generate the target health matrix based on the sub-target density and the normalized target index, and monitor the soil moisture in the target area at the same time; Allocate working parameters and working priorities of mobile devices based on the target health matrix to generate control signals; Triggering external devices based on a comprehensive judgment of the normalized target index and the soil moisture in the target area; receiving a control signal to drive the mobile device, monitoring the actual motor torque of the mobile device in real time, and comprehensively determining the target motor torque of the mobile device based on the sub-target slope and the sub-target density; The actual motor torque is compared with the target motor torque to trigger a partial rescan of the target image and update the target's health matrix.

Images