Greenhouse inspection method and system

Abstract

The present application provides a method and system for greenhouse inspection. The method includes: according to the plant growth demand information in the greenhouse, using an abnormal detection algorithm to process the parameter values ​​of environmental parameters to identify abnormal environmental parameters, determining the inspection target according to the abnormal environmental parameters, and determining the initial abnormality degree according to the parameter values ​​of the abnormal environmental parameters; determining the target inspection area according to the location information of the automation equipment and the initial abnormality degree, determining the corresponding inspection strategy according to the initial abnormality degree and the target inspection area, and generating an inspection plan according to the inspection target, the image acquisition and processing method in the inspection strategy, and the target inspection area; different inspection strategies contain the optimal inspection paths obtained by different algorithms and different image acquisition and processing methods; sending the inspection plan to the automation equipment to obtain the image processing result, and generating the inspection result according to the image processing result. The present application improves the accuracy of greenhouse inspection.

Description

Technical Field

[0001] The present application relates to the field of intelligent inspection technology, and in particular to an inspection method and system for a greenhouse. Background Art

[0002] Greenhouses are a common facility in modern agriculture that promotes plant growth by controlling environmental conditions (such as temperature, humidity, light, etc.). In order to ensure that the environmental conditions in the greenhouse always meet the growth needs of the plants, regular inspections are required.

[0003] Traditional inspection methods mainly rely on manual labor, which is inefficient and prone to omissions. In recent years, with the development of automation technology and artificial intelligence, the use of automated equipment for greenhouse inspection has gradually become a research hotspot. However, the existing methods of using automated equipment for greenhouse inspection have the following defects: the inspection path is fixed and cannot be flexibly adjusted according to actual conditions, and the image acquisition and processing methods are single, which makes it difficult to cope with complex and changing environments, resulting in low accuracy of greenhouse inspection. Summary of the invention

[0004] The embodiments of the present application provide a greenhouse inspection method and system to solve the problem of low accuracy of greenhouse inspection in the prior art due to the fixed inspection path and difficulty in coping with complex and changing environments.

[0005] In a first aspect, an embodiment of the present application provides a greenhouse inspection method, comprising:

[0006] Obtain parameter values ​​of environmental parameters collected by sensors in the greenhouse, information on plant growth requirements in the greenhouse, and location information of automation equipment;

[0007] According to the plant growth demand information in the greenhouse, an abnormality detection algorithm is used to process the parameter value of the environmental parameter to identify abnormal environmental parameters, an inspection target is determined according to the abnormal environmental parameters, and an initial abnormality degree is determined according to the parameter value of the abnormal environmental parameter;

[0008] Determine the target inspection area according to the location information and the initial abnormality degree of the automation equipment, determine the corresponding inspection strategy according to the initial abnormality degree and the target inspection area, and generate an inspection plan according to the inspection target, the image acquisition and processing method in the inspection strategy and the target inspection area; different inspection strategies include optimal inspection paths obtained by different algorithms and different image acquisition and processing methods;

[0009] The inspection plan is sent to the automation equipment through wireless communication technology, so that the automation equipment performs inspection along the optimal inspection path according to the inspection plan, processes the image according to the image acquisition and processing method during the inspection, obtains the image processing result, and generates the inspection result according to the image processing result.

[0010] Optionally, determining a corresponding inspection strategy according to the initial abnormality degree and the target inspection area includes:

[0011] Obtain performance information of automated equipment, including maximum travel speed, maximum operating radius, battery life, computing power, and sensor type;

[0012] When the initial abnormality is mild, combined with the maximum driving speed and maximum operating radius of the automation equipment, an improved genetic algorithm is used to perform path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the mild degree, and the image acquisition and processing method corresponding to the mild degree is set to combine the computing power and sensor type of the automation equipment, and use a lightweight image processing algorithm to perform image acquisition and processing on the target inspection area; the improved genetic algorithm is a genetic algorithm that introduces a local search algorithm, a heuristic crossover operator and an adaptive mutation operator;

[0013] When the initial abnormality is moderate, a multi-objective optimization algorithm is used to perform multi-objective comprehensive optimization processing on the target inspection area in combination with the maximum driving speed, maximum operating radius, battery life and computing power of the automation equipment to obtain the optimal inspection path corresponding to the moderate degree, and the image acquisition and processing method corresponding to the moderate degree is set to combine the computing power and sensor type of the automation equipment, and use a deep learning image recognition algorithm to collect and process images of the target inspection area;

[0014] When the initial abnormality level is severe, combined with the maximum driving speed, maximum operating radius, battery life and obstacle avoidance capability of the automation equipment, an emergency response path planning algorithm based on deep learning is used to perform rapid response path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the severity, and the image acquisition and processing method corresponding to the severity is set to combine the computing power and sensor type of the automation equipment, and a semantic segmentation algorithm is used to perform real-time image analysis and processing on the target inspection area.

[0015] Optionally, the maximum driving speed and maximum operating radius of the automated equipment are combined to perform path planning processing on the target inspection area using an improved genetic algorithm to obtain a lightly corresponding optimal inspection path, including:

[0016] The initial population is randomly generated according to the target inspection area, and each individual represents an initial inspection path;

[0017] Use the local search algorithm to optimize the initial population and obtain the optimized initial population;

[0018] The fitness values ​​of individuals in the population are calculated based on a multi-objective fitness evaluation function; the multi-objective fitness evaluation function introduces constraints including a maximum driving speed and a maximum operating radius of the automated equipment; the population is the initial population in the first iteration process and is the updated population in other iteration processes;

[0019] Use the roulette wheel selection algorithm to perform selection operations based on the fitness values ​​of the individuals to determine the selected individuals;

[0020] Use a heuristic crossover operator to select individuals from the screened individuals for crossover operation, and use an adaptive mutation operator to mutate the individuals in the population after the crossover operation to obtain an updated population;

[0021] Determine whether an iteration termination condition is met; the iteration termination condition is that the maximum number of iterations is reached or the fitness value of the optimal individual in the updated population is greater than or equal to a preset threshold;

[0022] If not, the calculation, selection, crossover and mutation operations are repeated until the iteration termination condition is met; the best individual in the last iteration is the optimal inspection path.

[0023] Optionally, the maximum driving speed, maximum operating radius, battery life and computing power of the automated equipment are combined to perform multi-objective comprehensive optimization processing on the target inspection area using a multi-objective optimization algorithm to obtain the optimal inspection path corresponding to the medium, including:

[0024] Quantitatively evaluate the maximum driving speed, maximum operating radius, battery life and computing power of automated equipment to form a multi-dimensional performance indicator matrix;

[0025] A multi-objective optimization model is constructed based on the geographic information and environmental characteristics of the target inspection area; the overall objective function of the multi-objective optimization model includes the following sub-objectives: minimizing the inspection path length, minimizing the inspection time, minimizing the energy consumption, and maximizing the task completion rate; each sub-objective corresponds to a weight, and the weight is obtained by adaptively adjusting the multi-dimensional performance indicator matrix using a preset dynamic weight adjustment strategy;

[0026] A differential evolution algorithm with memory mechanism, adaptive control parameter strategy, elite retention strategy and neighborhood update mechanism is used to obtain a non-dominated solution set.

[0027] Decision preference information is obtained, and an optimal inspection path is selected from the non-dominated solution set according to the decision preference information.

[0028] Optionally, the maximum driving speed, maximum operating radius, battery life and obstacle avoidance capability of the automated equipment are combined to use a deep learning-based emergency response path planning algorithm to perform rapid response path planning processing on the target inspection area to obtain a heavily corresponding optimal inspection path, including:

[0029] Determine the location of the abnormal point based on abnormal environmental parameters;

[0030] Obtaining map data and obstacle locations of a target inspection area, and integrating the map data, obstacle locations, and abnormal point locations into a multi-channel image; different channels represent different types of information;

[0031] Using a pre-trained convolutional neural network to extract features from the multi-channel image to obtain a feature map containing terrain features and obstacle features;

[0032] The position information of the automation equipment, the position of the abnormal point and the characteristic map are used as input sequences, and combined with the maximum driving speed and obstacle avoidance capability of the automation equipment, a recurrent neural network is used to process the input sequence to obtain a predicted inspection path;

[0033] Combining the maximum operating radius and battery life of the automated equipment, the predicted inspection path is optimized through a reinforcement learning algorithm to obtain the optimal inspection path.

[0034] Optionally, the method combines the computing power and sensor type of the automation equipment to acquire and process images of the target inspection area using a lightweight image processing algorithm, including:

[0035] Use the camera on the automation equipment to collect image data of the target inspection area;

[0036] Preprocessing the image data to obtain preprocessed image data;

[0037] Selecting an initial image processing algorithm according to the computing power of the automation equipment and the sensor type corresponding to the camera; the initial image processing algorithm includes: an edge detection algorithm, a feature extraction algorithm or a classifier;

[0038] The initial image processing algorithm is trained and optimized in combination with the computing power of the device to obtain a lightweight image processing algorithm;

[0039] A lightweight image processing algorithm is used to collect and process images of the target inspection area.

[0040] Optionally, the method combines the computing power and sensor type of the automation equipment to collect and process images of the target inspection area using a deep learning image recognition algorithm, including:

[0041] Use cameras on automated equipment to collect multi-scale images and use infrared sensors on automated equipment to collect sensor data;

[0042] Select multiple deep learning models for fusion, introduce knowledge distillation technology in the fusion process to obtain a fusion model, and perform distributed training on the fusion model to obtain a pre-trained fusion model;

[0043] The multi-stage detection strategy and the pre-trained fusion model are integrated into a deep learning image recognition algorithm, and the deep learning image recognition algorithm is used to collect and process images of the target inspection area.

[0044] In a second aspect, an embodiment of the present application provides a greenhouse inspection system, including:

[0045] An acquisition module is used to obtain parameter values ​​of environmental parameters collected by sensors in the greenhouse, plant growth requirement information in the greenhouse, and location information of automation equipment;

[0046] A processing and determination module, for processing the parameter values ​​of the environmental parameters using an abnormality detection algorithm according to the plant growth demand information in the greenhouse to identify abnormal environmental parameters, determine inspection targets according to the abnormal environmental parameters, and determine the initial abnormality degree according to the parameter values ​​of the abnormal environmental parameters;

[0047] A determination generation module is used to determine a target inspection area according to the location information and the initial abnormality degree of the automation equipment, determine a corresponding inspection strategy according to the initial abnormality degree and the target inspection area, and generate an inspection plan according to the inspection target, the image acquisition and processing method in the inspection strategy, and the target inspection area; different inspection strategies include optimal inspection paths obtained by different algorithms and different image acquisition and processing methods;

[0048] The sending module is used to send the inspection plan to the automation equipment through wireless communication technology, so that the automation equipment performs inspection along the optimal inspection path according to the inspection plan and processes the image according to the image acquisition and processing method during the inspection to obtain the image processing result, and generates the inspection result according to the image processing result.

[0049] In a third aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a greenhouse inspection method as described in any one of the first aspects.

[0050] In a fourth aspect, an embodiment of the present application provides a computer storage medium storing a computer program, which, when executed by a computer, implements a greenhouse inspection method as described in any one of the first aspects.

[0051] In the embodiment of the present application, the parameter values ​​of the environmental parameters collected by the sensors in the greenhouse, the plant growth demand information in the greenhouse and the location information of the automation equipment are obtained; according to the plant growth demand information in the greenhouse, the parameter values ​​of the environmental parameters are processed by an abnormal detection algorithm to identify abnormal environmental parameters, the inspection targets are determined according to the abnormal environmental parameters, and the initial abnormal degree is determined according to the parameter values ​​of the abnormal environmental parameters; the target inspection area is determined according to the location information and the initial abnormal degree of the automation equipment, the corresponding inspection strategy is determined according to the initial abnormal degree and the target inspection area, and the inspection plan is generated according to the inspection target, the image acquisition and processing method in the inspection strategy and the target inspection area; different inspection strategies include the optimal inspection paths obtained by different algorithms and different image acquisition and processing methods; the inspection plan is sent to the automation equipment through wireless communication technology, so that the automation equipment performs the inspection according to the image acquisition and processing method during the inspection along the optimal inspection path according to the inspection plan, and the image processing result is obtained, and the inspection result is generated according to the image processing result. In the embodiment of the present application, the abnormal environmental parameters identified by the abnormal detection algorithm can quantify the degree of abnormality, so as to determine which areas or parameters need to be focused on. Determination of the initial abnormality level makes inspections more targeted and avoids blind inspections. The most appropriate inspection strategy can be dynamically selected based on the initial abnormality level and the target inspection area. This embodiment can dynamically adjust the inspection strategy according to actual conditions to ensure the flexibility and adaptability of the inspection process. Different algorithms and methods can cope with different environments and needs and improve the comprehensiveness and accuracy of inspections. When the initial abnormality level is mild, the use of an improved genetic algorithm has the advantages of high computational efficiency, good path quality, and strong applicability; when the initial abnormality level is moderate, the use of a multi-objective optimization algorithm has the advantages of high flexibility and high reliability; when the initial abnormality level is severe, the use of an emergency response path planning algorithm based on deep learning has the advantages of rapid response, high precision, intelligent decision-making, and strong adaptability.

[0052] These and other aspects of the present application will become more clearly understood in the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0054] Figure 1 A flow chart of a greenhouse inspection method provided in an embodiment of the present application;

[0055] Figure 2 A schematic diagram of the structure of a greenhouse inspection system provided in an embodiment of the present application;

[0056] Figure 3 A schematic diagram of the structure of a computing device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0058] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this article or executed in parallel. The sequence numbers of the operations, such as 11, 12, etc., are only used to distinguish between different operations, and the sequence numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to different types.

[0059] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0060] Figure 1 A flow chart of a greenhouse inspection method provided in an embodiment of the present application, such as Figure 1 As shown, the method includes:

[0061] S11, obtaining parameter values ​​of environmental parameters collected by sensors in the greenhouse, plant growth requirement information in the greenhouse, and location information of automation equipment.

[0062] It should be understood that the sensor in the greenhouse can be a temperature sensor, a humidity sensor, etc., and the environmental parameters refer to various environmental conditions in the greenhouse, such as temperature, humidity, light, carbon dioxide concentration, etc. The parameter value refers to the specific value of the environmental condition. For example, the temperature is 25°C and the humidity is 70%.

[0063] It should also be understood that plant growth demand information refers to the requirements of plants for environmental conditions at different growth stages, including suitable temperature range, humidity range, light intensity, etc. For example, a certain plant requires a temperature range of 20-25°C and a humidity range of 60-70% during its growth period. Automation equipment may refer to inspection equipment, such as mobile robots, drones, etc. Correspondingly, the location information of the automation equipment refers to the current location of the inspection equipment in the greenhouse.

[0064] S12. Based on the plant growth demand information in the greenhouse, an anomaly detection algorithm is used to process the parameter values ​​of the environmental parameters to identify abnormal environmental parameters, determine inspection targets based on the abnormal environmental parameters, and determine the initial abnormality degree based on the parameter values ​​of the abnormal environmental parameters.

[0065] It should be understood that an anomaly detection algorithm is an algorithm for identifying outliers in data, and is often used to detect whether environmental parameters deviate from the normal range. Exemplarily, the anomaly detection algorithm is a statistical method (such as standard deviation) or a machine learning method (such as isolation forest, support vector machine, etc.). Abnormal environmental parameters refer to environmental parameters that deviate from the normal range identified by the anomaly detection algorithm. Exemplarily, if the temperature sensor detects that the temperature in a certain area is 30°C, which is 20-25°C beyond the normal range, then the temperature in the area is an abnormal environmental parameter. Inspection targets refer to parameters that need to be inspected in key areas or plants and equipment in greenhouses. Exemplarily, in the case where the temperature in the area is an abnormal environmental parameter, the inspection targets may include plants, and the inspection targets may also include air conditioners that provide cooling for the area, etc., to check the operation of the air conditioner. The initial abnormality level refers to the severity of the abnormal environmental parameter, which is usually determined by a quantitative method.

[0066] For example, the temperature abnormality is 80%, and the humidity abnormality is 60%. It can also be determined by level. For example, when the quantified temperature abnormality is any value between 1% and 40%, the initial abnormality can be mild; when the quantified temperature abnormality is any value between 41% and 60%, the initial abnormality can be moderate; when the quantified temperature abnormality is any value above 61%, the initial abnormality can be severe. The final abnormality can be included in the inspection results generated later, and the final abnormality can be determined by combining the initial abnormality and the image processing results.

[0067] Optionally, if the temperature in a certain area rises abnormally and the initial abnormality is high, this embodiment can prioritize the area as an inspection target to ensure timely discovery and handling of problems. The abnormal environmental parameters identified by the anomaly detection algorithm can quantify the degree of abnormality, thereby determining which areas or parameters need to be focused on. The determination of the initial abnormality makes the inspection more targeted and avoids blind inspections.

[0068] S13. Determine the target inspection area based on the location information of the automated equipment and the initial abnormality level, determine the corresponding inspection strategy based on the initial abnormality level and the target inspection area, and generate an inspection plan based on the inspection target, the image acquisition and processing method in the inspection strategy, and the target inspection area; different inspection strategies include optimal inspection paths obtained by different algorithms and different image acquisition and processing methods.

[0069] It should be understood that the number of automated devices can be one or more. A target inspection area is an inspection area of ​​an automated device. When the number of automated devices is multiple, the target inspection areas of all automated devices constitute the area to be inspected. Based on the principle of proximity, this embodiment determines the target inspection area based on the location information of the automated devices to achieve reasonable scheduling of the automated devices.

[0070] It should also be understood that this embodiment is centered on the location information of the sensor that provides abnormal environmental parameters. When the initial abnormality is mild, the area to be inspected is determined with the first value as the radius. When the initial abnormality is moderate, the area to be inspected is determined with the second value as the radius. When the initial abnormality is severe, the area to be inspected is determined with the third value as the radius. Among them, the first value is less than the second value, and the second value is less than the third value. After determining the area to be inspected, the target inspection area is determined based on the location information of the automation equipment according to the scheduling strategy of the automation equipment. The inspection strategy refers to the inspection method determined according to the inspection target and the target inspection area, including the optimal inspection path and the image acquisition and processing method. The optimal inspection path refers to the shortest or most efficient inspection path generated by the path planning algorithm corresponding to the initial abnormality degree based on the initial abnormality degree and the target inspection area. The image acquisition and processing method refers to the image acquisition technology and image processing algorithm used by the inspection equipment during the inspection process. For example, for areas with severe temperature abnormalities, select. For example, if the initial abnormality level in a certain area is high, this embodiment selects an inspection strategy that includes an emergency response path planning algorithm and a semantic segmentation algorithm to ensure that the automated equipment can quickly reach the area for detailed inspection. Therefore, according to the initial abnormality level and the target inspection area, this embodiment can dynamically select the most appropriate inspection strategy. Different inspection strategies contain optimal inspection paths obtained by different algorithms, which can ensure the accuracy of the inspection path design under different circumstances. In addition, according to the initial abnormality level and the target inspection area, the most suitable image acquisition and processing method is selected to ensure the quality of the acquired images and the accuracy of the processing results. Therefore, this embodiment can dynamically adjust the inspection strategy according to the actual situation to ensure the flexibility and adaptability of the inspection process. Different algorithms and methods can cope with different environments and needs, and improve the comprehensiveness and accuracy of inspections.

[0071] S14. Send an inspection plan to the automation equipment through wireless communication technology, so that the automation equipment can perform inspection along the optimal inspection path according to the inspection plan, process the image according to the image acquisition and processing method during the inspection, obtain the image processing result, and generate the inspection result according to the image processing result.

[0072] Among them, wireless communication technology refers to technology used for wireless data transmission, which is often used for remote control and data transmission. Inspection results refer to the final report generated based on the data collected during the inspection and the image processing results, including the details of the abnormal area and treatment suggestions. For example, the inspection report includes the specific location of the temperature abnormal area, the degree of abnormality and treatment suggestions. The inspection plan includes information such as the inspection speed of the automation equipment.

[0073] This embodiment uses wireless communication technology to send an inspection plan to the automation equipment, and the inspection equipment can perform inspections according to the optimal inspection path and image acquisition and processing method to generate inspection results. The entire process is highly intelligent, reducing human intervention and improving work efficiency.

[0074] The embodiment of the present application can quantify the degree of abnormality through abnormal environmental parameters identified by the anomaly detection algorithm, so as to determine which areas or parameters need to be focused on. The determination of the initial degree of abnormality makes the inspection more targeted and avoids blind inspections. The most appropriate inspection strategy can be dynamically selected according to the initial degree of abnormality and the target inspection area. This embodiment can dynamically adjust the inspection strategy according to actual conditions to ensure the flexibility and adaptability of the inspection process. Different algorithms and methods can cope with different environments and needs, and improve the comprehensiveness and accuracy of the inspection. In addition, this embodiment improves the level of intelligence of greenhouse inspections, and also provides strong support for the management and maintenance of greenhouses, which helps to improve the growth quality and yield of plants.

[0075] In some optional embodiments, in S12, a corresponding inspection strategy is determined according to the initial abnormality level and the target inspection area, including:

[0076] Step 121, obtaining performance information of the automation equipment, the performance information including maximum driving speed, maximum operating radius, battery life, computing power and sensor type. For example, the maximum driving speed is 1 m / s, the maximum operating radius is 50 m, the battery life is 2 hours, the computing power is 1 GHz, and the sensor types include temperature sensor, humidity sensor and camera.

[0077] It should be understood that in this embodiment, after executing step 121, any one of steps 122 to 124 may be executed.

[0078] Step 122, when the initial abnormality is mild, in combination with the maximum driving speed and maximum operating radius of the automation equipment, an improved genetic algorithm is used to perform path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the mild degree, and the image acquisition and processing method corresponding to the mild degree is set to be combined with the computing power and sensor type of the automation equipment, and a lightweight image processing algorithm is used to perform image acquisition and processing on the target inspection area. The improved genetic algorithm is a genetic algorithm that introduces a local search algorithm, a heuristic crossover operator, and an adaptive mutation operator. The shortest path generated by the improved genetic algorithm is used to ensure that the inspection equipment can efficiently complete the inspection task. A lightweight image processing algorithm refers to an image processing algorithm with low computing resource requirements, which is suitable for automation equipment with limited computing power.

[0079] Step 123, when the initial abnormality level is moderate, a multi-objective optimization algorithm is used to perform multi-objective comprehensive optimization processing on the target inspection area in combination with the maximum driving speed, maximum operating radius, battery life and computing power of the automation equipment to obtain the optimal inspection path corresponding to the moderate level, and the image acquisition and processing method corresponding to the moderate level is set to combine the computing power and sensor type of the automation equipment, and use a deep learning image recognition algorithm to perform image acquisition and processing on the target inspection area. A multi-objective optimization algorithm is an algorithm that can optimize multiple targets at the same time. The multi-objective optimization algorithm is used to simultaneously consider the path length, inspection time and resource consumption to generate a comprehensive optimal inspection path. A deep learning image recognition algorithm is an image recognition algorithm based on a deep neural network that can learn complex patterns and rules from large amounts of data. Exemplarily, a convolutional neural network is used for image recognition to identify plant growth conditions and environmental anomalies.

[0080] Step 124, when the initial abnormality level is severe, combined with the maximum driving speed, maximum operating radius, battery life and obstacle avoidance capability of the automation equipment, the emergency response path planning algorithm based on deep learning is used to perform rapid response path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the severity, and the image acquisition and processing method corresponding to the severity is set to combine the computing power and sensor type of the automation equipment, and the semantic segmentation algorithm is used to perform real-time image analysis and processing on the target inspection area. The emergency response path planning algorithm based on deep learning is an algorithm that can quickly generate an optimal inspection path and is suitable for rapid response in emergency situations. Exemplarily, a deep reinforcement learning algorithm is used to generate a rapid response path to ensure that the inspection equipment can quickly reach the abnormal area. The semantic segmentation algorithm is an algorithm that can classify each pixel in an image into different categories and is suitable for real-time image analysis. For example, this embodiment uses a semantic segmentation algorithm to perform real-time analysis of images to identify plant growth conditions and environmental abnormalities.

[0081] Through the above method, the embodiment of the present application can significantly improve the pertinence, efficiency and accuracy of greenhouse inspections. Among them, for areas with mild abnormalities, this embodiment selects an improved genetic algorithm to generate the optimal inspection path, and uses a lightweight image processing algorithm for image acquisition and processing to ensure that the inspection equipment can efficiently complete the task. For areas with moderate abnormalities, this embodiment selects a multi-objective optimization algorithm to generate the optimal inspection path, and uses a deep learning image recognition algorithm for image acquisition and processing to ensure that the growth conditions of plants and environmental abnormalities can be accurately identified. For areas with severe abnormalities, this embodiment selects an emergency response path planning algorithm based on deep learning to generate a rapid response path, and uses a semantic segmentation algorithm for real-time image analysis and processing to ensure that problems can be discovered and handled in a timely manner.

[0082] In the following embodiment, the algorithm of the optimal inspection path is designed as follows for each initial abnormality degree. In general, when the initial abnormality degree is mild, the use of the improved genetic algorithm has the advantages of high computational efficiency, good path quality, and strong applicability; when the initial abnormality degree is moderate, the use of the multi-objective optimization algorithm has the advantages of high flexibility and high reliability; when the initial abnormality degree is severe, the use of the emergency response path planning algorithm based on deep learning has the advantages of fast response, high precision, intelligent decision-making, and strong adaptability. Specifically:

[0083] In the above embodiment, in step 122, the improved genetic algorithm is used to perform path planning processing on the target inspection area in combination with the maximum driving speed and the maximum operating radius of the automation equipment to obtain the optimal inspection path corresponding to the light, including:

[0084] Step a1: randomly generate an initial population according to the target inspection area, and each individual represents an initial inspection path.

[0085] Step a2: Use a local search algorithm to optimize the initial population to obtain an optimized initial population.

[0086] Step a3: Calculate the fitness values ​​of individuals in the population based on the multi-objective fitness evaluation function. The multi-objective fitness evaluation function introduces constraints including the maximum driving speed and maximum operating radius of the automated equipment. The population is the initial population in the first iteration and the updated population in other iterations. The multi-objective fitness evaluation function uses the following formula:

[0087] ;

[0088] in, For individuals The fitness value of For individuals The corresponding inspection path length may refer to the total distance of the inspection path; For individuals The inspection time refers to the time required for the inspection equipment to complete the inspection path. For individuals Energy consumption refers to the energy required for the inspection equipment to complete the inspection path. , and They are , and The weight coefficient of , They are , The penalty coefficient. For individuals The maximum driving speed constraint ensures that the maximum driving speed in the inspection path does not exceed the maximum driving speed of the equipment.

[0089] ;

[0090] in, Indicates the node in the inspection path To Node The driving speed, is the maximum driving speed of the automated equipment, and n is the number of nodes. The maximum operating radius constraint ensures that each node in the inspection path is within the maximum operating radius of the device.

[0091] ;

[0092] in, Representation Node To base point The distance It is the maximum operating radius of the automated equipment.

[0093] Step a4: Use the roulette wheel selection algorithm to perform a selection operation based on the fitness value of the individual to determine the screened individuals.

[0094] Step a5, use the heuristic crossover operator to select individuals from the screened individuals for crossover operation, and use the adaptive mutation operator to mutate the individuals in the population after the crossover operation to obtain an updated population. The heuristic crossover operator is an improved crossover operator that combines heuristic information to make the crossover operation more efficient and targeted. The heuristic information can be the inspection path length, node distance, etc., which is used to guide the crossover operation and generate better quality offspring individuals. The heuristic crossover operator can implement any of partial mapping crossover, sequential crossover or cyclic crossover. The adaptive mutation operator is a mutation operator that dynamically adjusts the mutation probability according to the state of the current population.

[0095] Step a6: Determine whether the iteration termination condition is met. The iteration termination condition is that the maximum number of iterations is reached or the fitness value of the best individual in the updated population is greater than or equal to a preset threshold.

[0096] If not, the calculation in step a3, the selection operation in step a4, the crossover operation in step a5 and the mutation operation in step a6 are repeatedly performed until the iteration termination condition is met.

[0097] Step a7: determine the best individual in one iteration, and determine the best individual in the last iteration as the best inspection path.

[0098] This embodiment can balance the importance of each objective and constraint condition by dynamically adjusting the weight coefficient and penalty coefficient in the multi-objective fitness evaluation function, and improve the flexibility and adaptability of the improved genetic algorithm. The crossover operation is made more efficient and targeted by the heuristic crossover operator; by dynamically adjusting the mutation probability, exploration and development can be better balanced, and the convergence speed of the improved genetic algorithm and the quality of the optimal individual can be improved.

[0099] In the above embodiment, in step 123, a multi-objective optimization algorithm is used to perform multi-objective comprehensive optimization processing on the target inspection area in combination with the maximum driving speed, maximum operating radius, battery life and computing power of the automation equipment to obtain the optimal inspection path corresponding to the medium, including:

[0100] Step b1: Maximum driving speed of automated equipment , Maximum operating radius , Battery life and computing power Conduct quantitative evaluation and form a multi-dimensional performance indicator matrix.

[0101] Step b2: Construct a multi-objective optimization model based on the geographic information and environmental characteristics of the target inspection area. The overall objective function of the multi-objective optimization model includes the following sub-objectives: minimizing the inspection path length, minimizing the inspection time, minimizing energy consumption, and maximizing the task completion rate. Each sub-objective corresponds to a weight, and the weight is obtained by adaptively adjusting the preset dynamic weight adjustment strategy based on the multi-dimensional performance indicator matrix. The preset dynamic weight adjustment strategy provides a corresponding weight coefficient for each performance indicator. The overall objective function of the multi-objective optimization model adopts the following formula: ;

[0102] in, For individuals The total objective function value is , , and is the weight coefficient corresponding to the sub-goal, Represents an individual The minimum inspection path length is Represents an individual Minimize inspection time. Represents an individual Minimize energy consumption, Represents an individual Maximize task completion rate.

[0103] If the maximum operating radius of the automation equipment is larger, it means that the automation equipment can work in a larger area, and the weight of the inspection path length can be appropriately reduced. Similarly, if the maximum driving speed of the automated equipment is higher, it means that the automated equipment can complete the inspection task faster. Therefore, the embodiment of the present application can appropriately reduce the weight coefficient of the inspection time. .

[0104] Step b3, adopt a differential evolution algorithm that introduces a memory mechanism, an adaptive control parameter strategy, an elite retention strategy, and a neighborhood update mechanism to obtain a non-dominated solution set. The embodiment of the present application can perform non-dominated sorting on individuals in the population to generate multiple non-dominated layers. A non-dominated layer means that all individuals in a layer are mutually non-dominated, that is, no individual is better than another individual in all objectives. Calculate the crowding distance of individuals in each non-dominated layer; the crowding distance is used to measure the distribution density of individuals in the non-dominated layer; for each individual in the non-dominated layer, calculate the distance between its neighbors on each target, and add these distances to obtain the total crowding distance; subsequently perform a selection crossover operation, and record the best individual in the non-dominated solution set in each generation, and add the best individual in memory to the current population in each generation to prevent the algorithm from converging prematurely. In addition, in each generation, the best individuals in the previous generations are retained to ensure that these high-quality individuals are not discarded. Therefore, this embodiment can directly add elite individuals to the next generation population. In each generation, a local search is performed on the neighborhood of each individual to generate new individuals. Individuals in the neighborhood refer to individuals that are similar to the current individual in some dimensions. Through local search, the local optimization ability of the solution can be improved.

[0105] Step b4: Obtain decision preference information, and select the optimal inspection path from the non-dominated solution set according to the decision preference information. The decision preference information may refer to user preference. For example, the user may be more concerned about the length of the inspection path or the task completion rate. In this embodiment, the general non-dominated solution set may be re-sorted according to the user preference to select the optimal solution.

[0106] In this embodiment, the differential evolution algorithm is more suitable for handling moderate abnormal situations because it can handle complex multi-objective optimization problems and has strong global search and adaptive capabilities. The improved genetic algorithm is more suitable for handling mild abnormal situations because it has high computational efficiency, good path quality, strong applicability, and can run efficiently under limited resource conditions. This division of labor enables this embodiment to select the most appropriate algorithm according to different abnormality levels and improve the efficiency and accuracy of inspections.

[0107] In the above embodiment, in step 124, combined with the maximum driving speed, maximum operating radius, battery life and obstacle avoidance capability of the automation equipment, a deep learning-based emergency response path planning algorithm is used to perform rapid response path planning processing on the target inspection area to obtain the optimal inspection path with high degree of correspondence, including:

[0108] Step c1, determining the abnormal point location according to the abnormal environmental parameters. The abnormal point location refers to the geographical location where the abnormal environmental parameters are located.

[0109] Step c2, obtain the map data and obstacle positions of the target inspection area, and integrate the map data, obstacle positions and abnormal point positions into a multi-channel image. Different channels represent different types of information. The map data of the target inspection area refers to the geographical information describing the target inspection area, including boundaries, terrain, etc. Exemplarily, the map data includes the floor plan and terrain information of the greenhouse. The obstacle position refers to the location of fixed or mobile obstacles in the target inspection area. The obstacle position includes the location information of plants, equipment, walls, etc. A multi-channel image refers to integrating different types of information into a multi-channel image, and each channel represents a type of information. Exemplarily, the first channel represents terrain information, the second channel represents obstacle positions, and the third channel represents abnormal point positions.

[0110] Step c3: Use a pre-trained convolutional neural network to extract features from the multi-channel image to obtain a feature map containing terrain features and obstacle features. The pre-trained convolutional neural network extracts features from the multi-channel image through operations such as convolutional layers, pooling layers, and activation functions. The feature map refers to an intermediate representation of image features extracted by the convolutional neural network. The feature map can represent terrain features, obstacle features, etc.

[0111] Step c4: The location information, abnormal point location and feature map of the automation equipment are used as input sequences, and the input sequences are processed by a recurrent neural network in combination with the maximum driving speed and obstacle avoidance capability of the automation equipment to obtain a predicted inspection path. The recurrent neural network processes the input sequence through recursive connections to generate a predicted inspection path.

[0112] Step c5: Combine the maximum operating radius and battery life of the automated equipment and optimize the predicted inspection path through a reinforcement learning algorithm to obtain the optimal inspection path.

[0113] In this embodiment, severe abnormal situations usually require rapid response and timely processing. The emergency response path planning algorithm based on deep learning can quickly generate inspection paths to ensure that the inspection equipment can quickly reach the abnormal point location. The environment under severe abnormal conditions is complex and requires high-precision feature extraction. Therefore, this embodiment extracts rich terrain features and obstacle features from multi-channel images through convolutional neural networks to provide accurate environmental information. The inspection path under severe abnormal conditions needs to consider information of multiple time steps. Recurrent neural networks can process sequence data, capture dependencies in time series, and generate more reasonable inspection paths. The environment changes greatly under severe abnormal conditions, and the inspection path needs to be adjusted dynamically. Therefore, this embodiment adopts a reinforcement learning algorithm through trial and error and reward mechanisms, comprehensively considers multiple factors, and dynamically optimizes and predicts the inspection path to ensure the optimality and feasibility of the inspection path.

[0114] In the following embodiment, the image acquisition and processing method is designed as follows for each initial abnormality level:

[0115] In the above embodiment, in step 122, a lightweight image processing algorithm is used to collect and process images of the target inspection area in combination with the computing power and sensor type of the automation equipment, including:

[0116] Step d1: Use the camera on the automation equipment to collect image data of the target inspection area.

[0117] Step d2: preprocess the image data to obtain preprocessed image data.

[0118] Step d3: Select an initial image processing algorithm based on the computing power of the automation equipment and the sensor type corresponding to the camera. The initial image processing algorithm includes: edge detection algorithm, feature extraction algorithm or classifier.

[0119] Step d4: Train and optimize the initial image processing algorithm in combination with the computing power of the device to obtain a lightweight image processing algorithm. Through training and optimization, a more efficient lightweight image processing algorithm can be generated to improve the accuracy and speed of image processing.

[0120] Step d5: Use a lightweight image processing algorithm to collect and process images of the target inspection area.

[0121] In this embodiment, the environment under mild abnormal conditions is relatively simple and the demand for computing resources is low. Lightweight image processing algorithms have low demand for computing resources and are suitable for devices with limited computing power. They can efficiently utilize existing resources, avoid waste of resources, and improve the overall performance of automation equipment. During the inspection process, real-time processing of image data can promptly detect and handle abnormal situations, improve the efficiency and response speed of inspections. The computing power and sensor types of different automation equipment may vary. Lightweight algorithms can select the most appropriate algorithm according to the specific situation to ensure the applicability and effectiveness of the algorithm. High-quality image data facilitates subsequent image processing and analysis, thereby improving the accuracy and reliability of lightweight algorithms.

[0122] In the above embodiment, in step 123, a deep learning image recognition algorithm is used to collect and process images of the target inspection area in combination with the computing power and sensor type of the automation equipment, including:

[0123] Step e1: Use the camera on the automation equipment to collect multi-scale images, and use the infrared sensor on the automation equipment to collect sensor data.

[0124] Step e2: Select multiple deep learning models for fusion, introduce knowledge distillation technology in the fusion process, obtain the fusion model, and perform distributed training on the fusion model to obtain a pre-trained fusion model. For moderate abnormal situations, a large amount of training data and a long training time are required. Distributed training can significantly improve training efficiency.

[0125] Step e3: Integrate the multi-stage detection strategy and the pre-trained fusion model into a deep learning image recognition algorithm, and use the deep learning image recognition algorithm to collect and process images of the target inspection area.

[0126] The environment under moderate anomalies is more complex and may involve multiple anomaly types and changing environmental conditions. Deep learning models can process complex images and data, and provide rich environmental information through the fusion of multi-scale images and sensor data. Moderate anomalies require high-precision recognition and classification to ensure timely detection and handling of anomalies. Through the learning of multi-layer neural networks, deep learning models can extract deep features and improve the accuracy and robustness of recognition. Through knowledge distillation and model fusion, the complexity of the model and the demand for computing resources can be reduced while ensuring high performance. In moderate anomalies, multi-stage detection strategies can gradually eliminate interference and ensure that the final detection results are accurate.

[0127] In the above embodiment, in step 124, a semantic segmentation algorithm is used to perform real-time image analysis and processing on the target inspection area in combination with the computing power and sensor type of the automation equipment, including:

[0128] Step f1, collecting images and sensor data. In this embodiment, the camera on the automation equipment can be used to capture images of the target inspection area to ensure that the image clarity and resolution meet the requirements; and / or, various sensors (such as infrared sensors, temperature sensors, humidity sensors, etc.) on the automation equipment can be used to collect environmental data, such as temperature, humidity, light intensity, etc.

[0129] Step f2: pre-process the image and sensor data.

[0130] Step f3, apply the semantic segmentation algorithm to perform real-time image analysis. Specifically, in this embodiment, a suitable semantic segmentation model can be selected, and then a pre-trained semantic segmentation model can be loaded, and then the semantic segmentation model can be optimized in combination with sensor data, and then the optimized semantic segmentation model can be used to perform real-time analysis on the pre-processed image data to generate a semantic segmentation result. The generated semantic segmentation result is post-processed, such as removing small area noise, smoothing boundaries, etc., and the image processing result is visualized.

[0131] The environment in severe abnormal situations is usually very complex and may involve multiple abnormal types and changing conditions. The semantic segmentation algorithm can extract deep features and adapt to complex environments through the learning of multi-layer neural networks. Severe abnormal situations require rapid response and timely processing. The semantic segmentation algorithm can realize real-time image analysis on automated equipment through efficient model design and optimization, improving the efficiency and response speed of inspections. The computing power and sensor types of different devices may vary. The semantic segmentation algorithm can select the most appropriate model and optimization strategy according to the specific situation to ensure the applicability and effectiveness of the algorithm.

[0132] In summary, lightweight image processing algorithms are suitable for handling mild anomalies because they have low computing resource requirements, can achieve real-time processing, and are highly adaptable. Deep learning image recognition algorithms are suitable for handling moderate anomalies because they can handle complex environments and changing conditions and provide high-precision recognition and classification results. The semantic segmentation algorithm can achieve real-time image analysis on automated equipment through efficient model design and optimization, and is suitable for handling severe anomalies. This embodiment can automatically select the most appropriate algorithm according to different degrees of anomalies to improve the efficiency and accuracy of inspections.

[0133] Figure 2 A schematic diagram of a greenhouse inspection system provided in an embodiment of the present application is shown in FIG. Figure 2 As shown, the system includes:

[0134] The acquisition module 21 is used to acquire parameter values ​​of environmental parameters collected by sensors in the greenhouse, plant growth requirement information in the greenhouse, and location information of automation equipment.

[0135] The processing and determination module 22 is used to process the parameter values ​​of the environmental parameters according to the plant growth demand information in the greenhouse using an abnormality detection algorithm to identify abnormal environmental parameters, determine inspection targets according to the abnormal environmental parameters, and determine the initial abnormality degree according to the parameter values ​​of the abnormal environmental parameters.

[0136] A determination generation module 23 is used to determine the target inspection area based on the location information of the automated equipment and the initial abnormality level, determine the corresponding inspection strategy based on the initial abnormality level and the target inspection area, and generate an inspection plan based on the inspection target, the image acquisition and processing method in the inspection strategy, and the target inspection area; different inspection strategies include optimal inspection paths obtained by different algorithms and different image acquisition and processing methods.

[0137] The sending module 24 is used to send the inspection plan to the automation equipment through wireless communication technology, so that the automation equipment can perform inspection along the optimal inspection path according to the inspection plan, process the image according to the image acquisition and processing method during the inspection, obtain the image processing result, and generate the inspection result according to the image processing result.

[0138] Figure 2 The greenhouse inspection system can perform Figure 1 The implementation principle and technical effect of the greenhouse inspection method described in the embodiment are not described in detail. The specific way in which each module and unit performs operations in the greenhouse inspection system in the above embodiment has been described in detail in the embodiment of the method, and will not be described in detail here.

[0139] In one possible design, Figure 2 The inspection system of the greenhouse in the embodiment shown can be implemented as a computing device, such as Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32 .

[0140] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component 32 .

[0141] The processing component 32 is used to: obtain the parameter values ​​of environmental parameters collected by sensors in the greenhouse, plant growth demand information in the greenhouse and the location information of the automation equipment; according to the plant growth demand information in the greenhouse, use an abnormality detection algorithm to process the parameter values ​​of the environmental parameters to identify abnormal environmental parameters, determine the inspection target according to the abnormal environmental parameters, and determine the initial abnormality degree according to the parameter values ​​of the abnormal environmental parameters; determine the target inspection area according to the location information of the automation equipment and the initial abnormality degree, determine the corresponding inspection strategy according to the initial abnormality degree and the target inspection area, and generate an inspection plan according to the inspection target, the image acquisition and processing method in the inspection strategy and the target inspection area; different inspection strategies include optimal inspection paths obtained by different algorithms and different image acquisition and processing methods; send the inspection plan to the automation equipment through wireless communication technology, so that the automation equipment performs inspection along the optimal inspection path according to the inspection plan, processes the image according to the image acquisition and processing method during the inspection process, obtains the image processing result, and generates the inspection result according to the image processing result.

[0142] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented by one or more application-specific integrated circuits (ASIC), digital signal processors (DSP), digital signal processing devices (DSPD), programmable logic devices (PLD), field programmable gate arrays (FPGA), controllers, microcontrollers, microprocessors or other electronic components to perform the above method.

[0143] The storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as random access memory (RAM), static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0144] Of course, the computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0145] The input / output interface provides an interface between the processing component and the peripheral interface module, which may be an output device, an input device, etc.

[0146] The communication component is configured to facilitate, among other things, wired or wireless communications between the computing device and other devices.

[0147] Among them, the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. In this case, the computing device can refer to a cloud server, and the above-mentioned processing components, storage components, etc. can be basic server resources rented or purchased from the cloud computing platform.

[0148] The present application also provides a computer storage medium storing a computer program, wherein the computer program can achieve the above-mentioned Figure 1 The greenhouse inspection method of the illustrated embodiment.

[0149] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0150] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art may understand and implement it without creative work.

[0151] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0152] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A greenhouse inspection method, characterized in that: include: Obtain parameter values ​​of environmental parameters collected by sensors in the greenhouse, information on plant growth requirements in the greenhouse, and location information of automation equipment; According to the plant growth demand information in the greenhouse, an abnormality detection algorithm is used to process the parameter value of the environmental parameter to identify abnormal environmental parameters, an inspection target is determined according to the abnormal environmental parameters, and an initial abnormality degree is determined according to the parameter value of the abnormal environmental parameter; Determine a target inspection area according to the location information and the initial abnormality degree of the automation equipment, determine a corresponding inspection strategy according to the initial abnormality degree and the target inspection area, and generate an inspection plan according to the inspection target, the image acquisition and processing method in the inspection strategy, and the target inspection area; Different inspection strategies include optimal inspection paths obtained by different algorithms and different image acquisition and processing methods; The inspection plan is sent to the automation device through wireless communication technology, so that the automation device performs inspection along the optimal inspection path according to the inspection plan, processes the image according to the image acquisition and processing method during the inspection, obtains the image processing result, and generates the inspection result according to the image processing result; The determining of the corresponding inspection strategy according to the initial abnormality degree and the target inspection area includes: Obtain performance information of automated equipment, including maximum travel speed, maximum operating radius, battery life, computing power, and sensor type; When the initial abnormality is mild, combined with the maximum driving speed and maximum operating radius of the automation equipment, an improved genetic algorithm is used to perform path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the mild degree, and the image acquisition and processing method corresponding to the mild degree is set to combine the computing power and sensor type of the automation equipment, and use a lightweight image processing algorithm to perform image acquisition and processing on the target inspection area; the improved genetic algorithm is a genetic algorithm that introduces a local search algorithm, a heuristic crossover operator and an adaptive mutation operator; When the initial abnormality is moderate, a multi-objective optimization algorithm is used to perform multi-objective comprehensive optimization processing on the target inspection area in combination with the maximum driving speed, maximum operating radius, battery life and computing power of the automation equipment to obtain the optimal inspection path corresponding to the moderate degree, and the image acquisition and processing method corresponding to the moderate degree is set to combine the computing power and sensor type of the automation equipment, and use a deep learning image recognition algorithm to collect and process images of the target inspection area; When the initial abnormality level is severe, combined with the maximum driving speed, maximum operating radius, battery life and obstacle avoidance capability of the automation equipment, an emergency response path planning algorithm based on deep learning is used to perform rapid response path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the severity, and the image acquisition and processing method corresponding to the severity is set to combine the computing power and sensor type of the automation equipment, and a semantic segmentation algorithm is used to perform real-time image analysis and processing on the target inspection area.

2. The method according to claim 1, characterized in that The maximum driving speed and the maximum operating radius of the automated equipment are combined, and an improved genetic algorithm is used to perform path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the light, including: The initial population is randomly generated according to the target inspection area, and each individual represents an initial inspection path; Use the local search algorithm to optimize the initial population and obtain the optimized initial population; The fitness values ​​of individuals in the population are calculated based on a multi-objective fitness evaluation function; the multi-objective fitness evaluation function introduces constraints including a maximum driving speed and a maximum operating radius of the automated equipment; the population is the initial population in the first iteration process and is the updated population in other iteration processes; Use the roulette wheel selection algorithm to perform selection operations based on the fitness values ​​of the individuals to determine the selected individuals; Use a heuristic crossover operator to select individuals from the screened individuals for crossover operation, and use an adaptive mutation operator to mutate the individuals in the population after the crossover operation to obtain an updated population; Determine whether an iteration termination condition is met; the iteration termination condition is that the maximum number of iterations is reached or the fitness value of the optimal individual in the updated population is greater than or equal to a preset threshold; If not, the calculation, selection, crossover and mutation operations are repeated until the iteration termination condition is met; the best individual in the last iteration is the optimal inspection path.

3. The method according to claim 1, characterized in that The above-mentioned multi-objective optimization algorithm is used to perform multi-objective comprehensive optimization processing on the target inspection area in combination with the maximum driving speed, maximum operating radius, battery life and computing power of the automation equipment to obtain the optimal inspection path corresponding to the medium, including: Quantitatively evaluate the maximum driving speed, maximum operating radius, battery life and computing power of automated equipment to form a multi-dimensional performance indicator matrix; A multi-objective optimization model is constructed based on the geographic information and environmental characteristics of the target inspection area; the overall objective function of the multi-objective optimization model includes the following sub-objectives: minimizing the inspection path length, minimizing the inspection time, minimizing the energy consumption, and maximizing the task completion rate; each sub-objective corresponds to a weight, and the weight is obtained by adaptively adjusting the multi-dimensional performance indicator matrix using a preset dynamic weight adjustment strategy; A differential evolution algorithm with memory mechanism, adaptive control parameter strategy, elite retention strategy and neighborhood update mechanism is used to obtain a non-dominated solution set. Decision preference information is obtained, and an optimal inspection path is selected from the non-dominated solution set according to the decision preference information.

4. The method according to claim 1, characterized in that The method combines the maximum driving speed, maximum operating radius, battery life and obstacle avoidance capability of the automated equipment, adopts a deep learning-based emergency response path planning algorithm to perform rapid response path planning processing on the target inspection area, and obtains the optimal inspection path corresponding to the severity, including: Determine the location of the abnormal point based on abnormal environmental parameters; Obtaining map data and obstacle locations of a target inspection area, and integrating the map data, obstacle locations, and abnormal point locations into a multi-channel image; different channels represent different types of information; Using a pre-trained convolutional neural network to extract features from the multi-channel image to obtain a feature map containing terrain features and obstacle features; The position information of the automation equipment, the position of the abnormal point and the characteristic map are used as input sequences, and combined with the maximum driving speed and obstacle avoidance capability of the automation equipment, a recurrent neural network is used to process the input sequence to obtain a predicted inspection path; Combining the maximum operating radius and battery life of the automated equipment, the predicted inspection path is optimized through a reinforcement learning algorithm to obtain the optimal inspection path.

5. The method according to claim 1, characterized in that The method combines the computing power and sensor type of the automation equipment and uses a lightweight image processing algorithm to collect and process images of the target inspection area, including: Use the camera on the automation equipment to collect image data of the target inspection area; Preprocessing the image data to obtain preprocessed image data; Selecting an initial image processing algorithm according to the computing power of the automation equipment and the sensor type corresponding to the camera; the initial image processing algorithm includes: an edge detection algorithm, a feature extraction algorithm or a classifier; The initial image processing algorithm is trained and optimized in combination with the computing power of the device to obtain a lightweight image processing algorithm; A lightweight image processing algorithm is used to collect and process images of the target inspection area.

6. The method according to claim 1, characterized in that The method combines the computing power and sensor type of the automation equipment and uses a deep learning image recognition algorithm to collect and process images of the target inspection area, including: Use cameras on automated equipment to collect multi-scale images and use infrared sensors on automated equipment to collect sensor data; Select multiple deep learning models for fusion, introduce knowledge distillation technology in the fusion process to obtain a fusion model, and perform distributed training on the fusion model to obtain a pre-trained fusion model; The multi-stage detection strategy and the pre-trained fusion model are integrated into a deep learning image recognition algorithm, and the deep learning image recognition algorithm is used to collect and process images of the target inspection area.

7. A greenhouse inspection system, characterized in that: include: An acquisition module is used to obtain parameter values ​​of environmental parameters collected by sensors in the greenhouse, plant growth requirement information in the greenhouse, and location information of automation equipment; A processing and determination module, for processing the parameter values ​​of the environmental parameters using an abnormality detection algorithm according to the plant growth demand information in the greenhouse to identify abnormal environmental parameters, determine inspection targets according to the abnormal environmental parameters, and determine the initial abnormality degree according to the parameter values ​​of the abnormal environmental parameters; A determination generation module is used to determine a target inspection area according to the location information and the initial abnormality degree of the automation equipment, determine a corresponding inspection strategy according to the initial abnormality degree and the target inspection area, and generate an inspection plan according to the inspection target, the image acquisition and processing method in the inspection strategy, and the target inspection area; Different inspection strategies include optimal inspection paths obtained by different algorithms and different image acquisition and processing methods; A sending module, used for sending the inspection plan to the automation device through wireless communication technology, so that the automation device performs inspection along the optimal inspection path according to the inspection plan, performs processing according to the image acquisition and processing method during the inspection, obtains the image processing result, and generates the inspection result according to the image processing result; The determining of the corresponding inspection strategy according to the initial abnormality degree and the target inspection area includes: Obtain performance information of automated equipment, including maximum travel speed, maximum operating radius, battery life, computing power, and sensor type; When the initial abnormality is mild, combined with the maximum driving speed and maximum operating radius of the automation equipment, an improved genetic algorithm is used to perform path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the mild degree, and the image acquisition and processing method corresponding to the mild degree is set to combine the computing power and sensor type of the automation equipment, and use a lightweight image processing algorithm to perform image acquisition and processing on the target inspection area; the improved genetic algorithm is a genetic algorithm that introduces a local search algorithm, a heuristic crossover operator and an adaptive mutation operator; When the initial abnormality is moderate, a multi-objective optimization algorithm is used to perform multi-objective comprehensive optimization processing on the target inspection area in combination with the maximum driving speed, maximum operating radius, battery life and computing power of the automation equipment to obtain the optimal inspection path corresponding to the moderate degree, and the image acquisition and processing method corresponding to the moderate degree is set to combine the computing power and sensor type of the automation equipment, and use a deep learning image recognition algorithm to collect and process images of the target inspection area; When the initial abnormality level is severe, combined with the maximum driving speed, maximum operating radius, battery life and obstacle avoidance capability of the automation equipment, an emergency response path planning algorithm based on deep learning is used to perform rapid response path planning processing on the target inspection area to obtain the optimal inspection path corresponding to the severity, and the image acquisition and processing method corresponding to the severity is set to combine the computing power and sensor type of the automation equipment, and a semantic segmentation algorithm is used to perform real-time image analysis and processing on the target inspection area.

8. A computing device, characterized in that It comprises a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement a greenhouse inspection method as described in any one of claims 1 to 6.

9. A computer storage medium, characterized in that A computer program is stored, and when the computer program is executed by a computer, a greenhouse inspection method as described in any one of claims 1 to 6 is implemented.

Images