Unmanned aerial vehicle intelligent inspection control method and system based on low-altitude grid codes

Abstract

The invention provides an unmanned aerial vehicle intelligent inspection control method and system based on a low-altitude grid code, and relates to the technical field of unmanned aerial vehicle inspection, and the method comprises the steps: planning a grid code atlas based on an inspection point; receiving an inspection task, introducing a heuristic port, executing inspection task interpretation, and adding a random radius disturbance point as a heuristic task cluster; with inspection path-visual parameters as a planning target, according to a depth planning module deployed by an inspection control system, boundary iteration pruning and optimization planning based on heuristic decision are executed for a heuristic task cluster, a pre-planning strategy is determined, an inspection unmanned aerial vehicle and an airborne camera are driven, and a dynamic lightweight module is deployed based on central control of the unmanned aerial vehicle. And executing inspection decision control triggered by the dynamic task. The technical problem that the execution efficiency of the unmanned aerial vehicle inspection task is low in the prior art can be solved, the technical target of adaptive planning based on the inspection task is achieved, and the technical effect of improving the task execution efficiency is achieved.

Description

Technical Field

[0001] This application relates to the field of unmanned aerial vehicle (UAV) inspection technology, and in particular to an intelligent UAV inspection control method and system based on low-altitude grid codes. Background Technology

[0002] In the current technological context of drone inspection systems, with the increasing demand for efficient and automated inspection methods in fields such as power, petrochemicals, and transportation, drone inspection is gradually transitioning from manual operation to intelligent path planning and autonomous perception control.

[0003] Currently, traditional UAV inspection mainly relies on static path planning algorithms and preset shooting parameters for task execution, lacking the ability to flexibly respond to complex environmental changes and dynamic task adjustments. These systems often employ a single-task-driven model, where all flight paths and camera parameters are set once before the task begins. If unexpected events occur during the task, such as sudden weather changes, target occlusion, or image acquisition failure, the inspection strategy often cannot be adjusted in time, leading to problems such as degraded image quality, missing data, or task failure. Especially in scenarios with vast inspection areas, complex target structures, and numerous interference factors, the traditional model suffers from poor path adaptability, low image acquisition success rate, and low efficiency in handling ad-hoc additional tasks.

[0004] In summary, existing technologies suffer from a lack of dynamic perception and response mechanisms, which prevents the drone from adjusting its path and shooting strategy in real time according to environmental changes or mission status during inspections. This further impacts the efficiency of drone inspection missions, the integrity of image acquisition, and the overall intelligence level of the system. Summary of the Invention

[0005] The purpose of this application is to provide a method and system for intelligent inspection control of unmanned aerial vehicles (UAVs) based on low-altitude grid codes, in order to solve the technical problems in the prior art where the lack of dynamic perception and response mechanisms makes it impossible to adjust the path and shooting strategy in real time according to environmental changes or task status during the inspection process, which further affects the execution efficiency of UAV inspection tasks, the integrity of image acquisition, and the overall intelligence level of the system.

[0006] In view of the above problems, this application provides a method and system for intelligent inspection and control of unmanned aerial vehicles based on low-altitude grid codes.

[0007] Firstly, this application provides a UAV intelligent inspection control method based on low-altitude grid codes, implemented through a UAV intelligent inspection control system based on low-altitude grid codes. The method includes: planning a grid code map based on inspection points for the inspection scenario, wherein each inspection point is assigned a grid code associated with location and equipment information; receiving inspection tasks, introducing a heuristic port, executing inspection task interpretation, and adding random radius perturbation points as a heuristic task cluster; using the inspection path-visual parameters as the planning target, and based on the depth planning module deployed in the inspection control system, performing boundary iterative pruning and optimization planning based on heuristic decision-making for the heuristic task cluster, determining a pre-planning strategy and driving the inspection UAV and onboard camera; and executing inspection decision control under dynamic task triggering based on a dynamic lightweight module deployed in the UAV central control unit. The depth planning module and the dynamic lightweight module have the grid code map built-in, and communication interaction is established between the modules.

[0008] Preferably, the UAV intelligent inspection control method based on low-altitude grid code further includes: constructing a first functional node based on semantic interpretation, constructing a second functional node based on heuristic decomposition, and constructing a third functional node based on disturbance addition; cascading the first functional node, the second functional node, and the third functional node to generate the heuristic port.

[0009] Preferably, the UAV intelligent inspection control method based on low-altitude grid codes further includes: the inspection control system receiving the inspection task, triggering the heuristic port, driving the first functional node to perform task interpretation analysis, and determining the interpretation data, wherein the inspection task includes at least one inspection point; triggering the second functional node to identify the interpretation data and locate the heuristic direction, performing task decomposition based on the heuristic direction, and determining the decomposed tasks, wherein the smallest interpretation unit is used as the positioning standard; triggering the third functional node to introduce random radius perturbation points, adding perturbation heuristic directions for each decomposed task, and generating the heuristic direction task cluster, wherein the random radius of the perturbation point is customized.

[0010] Preferably, the UAV intelligent inspection control method based on low-altitude grid codes further includes: the visual parameters at least include camera focal length and shooting spatial angle; according to the inspection task, the range of the inspection points is defined based on the grid code map, and a solution space tree is constructed accordingly; the heuristic task cluster is traversed, and heuristic boundary calculation and pruning based on the solution space tree are performed, and an effective solution space tree is determined through multiple iterations; optimization decision is performed based on the effective solution space tree to determine the pre-planning strategy.

[0011] Preferably, the UAV intelligent inspection control method based on low-altitude grid codes further includes: determining a first heuristic task according to the heuristic task cluster; defining a solution boundary in the solution space tree according to the first heuristic task; pruning the solution space tree according to the solution boundary to determine a solution space tree; determining a second heuristic task; performing a solution boundary definition and pruning of the solution space tree; and so on until the iteration of the heuristic task cluster is completed to determine the effective solution space tree.

[0012] Preferably, the UAV intelligent inspection control method based on low-altitude grid codes further includes: transmitting the pre-planned strategy to the inspection UAV according to communication interaction, and executing strategy-driven inspection control; generating dynamic task instructions, wherein the dynamic task instructions are triggered when a new inspection task is added or an invalid captured image is obtained; and triggering the dynamic lightweight module according to the dynamic task instructions to execute the adjustment decision of the pre-planned strategy, determine and execute the updated inspection strategy.

[0013] Preferably, the UAV intelligent inspection control method based on low-altitude grid codes further includes: introducing a fast discriminator, wherein the validity or invalidity of the captured image is used as the discrimination criterion; as the inspection is executed with the pre-planned strategy, the captured image is received and image judgment based on the fast discriminator is performed; if the judgment result is an invalid captured image, a dynamic task instruction is generated, wherein a backup collection point based on the inspection grid code is located as accompanying information of the instruction.

[0014] Preferably, the UAV intelligent inspection control method based on low-altitude grid codes further includes: determining planning constraints, wherein the optimal acquisition point and at least two backup acquisition points are used as planning constraints for each heuristic task; wherein the optimal acquisition point imposes decision constraints on the depth planning module, and the backup acquisition points impose decision constraints on the dynamic lightweight module.

[0015] Preferably, the UAV intelligent inspection control method based on low-altitude grid codes further includes: acquiring inspection images, wherein the inspection images are associated with grid codes; connecting to a cloud server, uploading the inspection images and arranging them according to the grid code map to determine the inspection map; and performing terminal display and fault analysis and early warning management on the inspection map.

[0016] Secondly, this application also provides a UAV intelligent inspection control system based on low-altitude grid codes, used to execute the UAV intelligent inspection control method based on low-altitude grid codes as described in the first aspect, including: a grid code planning module, used to plan a grid code map based on inspection points for the inspection scenario, wherein each inspection point is assigned a grid code, which is associated with location information and equipment information; an inspection task interpretation module, used to receive inspection tasks, introduce heuristic ports, perform inspection task interpretation and add random radius perturbation points as heuristic task clusters; and an inspection decision control module, used to take the inspection path-visual parameters as the planning target, and according to the deep planning module deployed in the inspection control system, perform boundary iterative pruning and optimization planning based on heuristic decision for the heuristic task clusters, determine the pre-planning strategy and drive the inspection UAV and airborne camera, and perform inspection decision control under dynamic task triggering based on the dynamic lightweight module deployed in the UAV central control, wherein the deep planning module and the dynamic lightweight module have the grid code map built in, and communication interaction is established between the modules.

[0017] The technical solution provided in this application has at least the following technical effects or advantages: by realizing the technical goal of adaptive planning and dynamic control of heuristic inspection tasks based on grid code maps, it achieves the technical effect of improving path adaptability, image acquisition success rate and overall task execution stability in a variable inspection environment.

[0018] The above description is merely an overview of the technical solution of this application. To better understand the technical means of this application and to facilitate its implementation according to the description, and to make the above and other objects, features, and advantages of this application more apparent, specific embodiments of this application are described below. It should be understood that the content described in this section is not intended to identify key or important features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent through the following description. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0020] Figure 1 This is a flowchart illustrating the UAV intelligent inspection control method based on low-altitude grid codes proposed in this application.

[0021] Figure 2 This is a schematic diagram of the structure of the UAV intelligent inspection control system based on low-altitude grid codes in this application.

[0022] Explanation of reference numerals in the attached diagram: Grid code planning module 11, inspection task interpretation module 12, inspection decision control module 13. Detailed Implementation

[0023] This application provides an intelligent inspection control method and system for unmanned aerial vehicles (UAVs) based on low-altitude grid codes. It addresses the technical problem in existing technologies where the lack of dynamic perception and response mechanisms prevents real-time adjustments to paths and shooting strategies based on environmental changes or task status during inspections, thus impacting the efficiency of UAV inspection tasks, the integrity of image acquisition, and the overall system's intelligence level. The method achieves the technical goal of heuristic adaptive planning and dynamic control of inspection tasks based on grid code maps, thereby improving path adaptability, image acquisition success rate, and overall task execution stability in variable inspection environments.

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. It should be understood that this application is not limited to the exemplary embodiments described herein. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application. It should also be noted that, for ease of description, only the parts related to this application are shown in the accompanying drawings, not all of them.

[0025] Example 1, please refer to the appendix. Figure 1 This application provides a UAV intelligent inspection control method based on low-altitude grid codes, which is applied to a UAV intelligent inspection control system based on low-altitude grid codes, and specifically includes the following steps: S1: For inspection scenarios, plan a grid code map based on inspection points. Each inspection point is assigned a grid code that is associated with location information and equipment information.

[0026] Specifically, during the inspection process, a detailed task scenario model is created for the inspection area, clarifying the spatial relationships of each location or piece of equipment to be inspected. Based on geographic information and spatial division rules, each inspection point is marked with an independent grid code. A grid code is a two-dimensional or three-dimensional spatial coding system used to digitize physical locations, facilitating calculation and positioning. For example, a transmission tower in a substation can be assigned a specific grid code to uniquely identify its location.

[0027] Next, each grid code is bound to its corresponding location information, which is high-precision geographic coordinates, such as latitude and longitude or three-dimensional coordinates, so that the specific location of the inspection point on the map can be quickly restored by decoding the grid code.

[0028] Furthermore, each inspection point not only has geographical location information but is also associated with specific equipment information. Equipment information includes equipment type, serial number, maintenance records, and vulnerable components. This association is achieved through database records, facilitating the automatic matching of the equipment type and corresponding technical parameters to be inspected based on task requirements during drone inspections.

[0029] By integrating grid codes, location information, and device information, a grid code map can be constructed, which serves as the basic framework for the entire UAV inspection mission. All task points, location information, and target objects are clearly coded and defined within it.

[0030] S2: Receive inspection tasks, introduce heuristic ports, execute inspection task interpretation and add random radius perturbation points as heuristic task clusters.

[0031] Specifically, receiving an inspection task means that the inspection control system has received the inspection work to be performed from the dispatch center, task platform or manual input interface. This includes multiple inspection points, each corresponding to actual physical equipment, such as transformers, power towers or line joints. Based on the descriptive information included in the inspection task, it is necessary to further identify the spatial location, equipment type and inspection operation requirements of each inspection point.

[0032] Next, the introduction of a heuristic port refers to the system's internal startup of a task processing entry point for subsequent functional nodes to work together, which plays a role in unified scheduling and task distribution in the logical structure.

[0033] Subsequently, the inspection task is interpreted by analyzing and processing the natural language descriptions or structured instructions in the inspection task, converting them into a standardized data format that the system can understand, in order to extract the effective information contained in the inspection task and generate interpreted data.

[0034] Next, random radius perturbation points are added around the original inspection points. These are backup inspection points randomly generated according to the perturbation strategy, used to improve the robustness and flexibility of the task. The radius of the perturbation points is a custom parameter; for example, it can be set to generate 1 to 2 perturbation points within a radius of 5 meters.

[0035] S3: Taking the inspection path-visual parameters as the planning target, based on the deep planning module deployed in the inspection control system, the system performs boundary iterative pruning and optimization planning based on heuristic decision-making for the heuristic task cluster, determines the pre-planning strategy and drives the inspection drone and airborne camera, and executes inspection decision control under dynamic task triggering based on the dynamic lightweight module deployed in the drone central control. The deep planning module and the dynamic lightweight module have the grid code map built in, and communication interaction is established between the modules.

[0036] Specifically, during task planning, both the actual path the unmanned inspection equipment (such as drones or robots) will traverse and the visual parameters required for image acquisition must be considered. Visual parameters include the camera's angle of view, focal length, exposure time, sharpness threshold, and target recognition accuracy requirements, directly determining whether clear and effective image information can be acquired during the inspection. The inspection path refers to the movement trajectory of the equipment from the starting point through all target points sequentially. Planning must ensure the path's continuity, shortest path, and obstacle avoidance capabilities. Therefore, when using the inspection path and visual parameters as the planning objective, while selecting a route, it is also necessary to weigh whether the equipment's camera along that route can meet the imaging requirements of each target point, thus forming a control logic that coordinates and optimizes the path and perception capabilities.

[0037] Next, the achievement of the planning objectives relies on the deep planning module deployed in the inspection and control system. The deep planning module is a core component of an intelligent algorithm, integrating path optimization algorithms (such as A*, RRT, or reinforcement learning strategies), task matching mechanisms, and multi-objective weight adjustment methods. The role of the deep planning module is to simulate the combined results of different path selections and visual parameter settings, and to optimize the selection based on constraints (such as flight altitude, obstacle avoidance requirements, and power consumption) and evaluation metrics (such as coverage and recognition accuracy).

[0038] Boundary iterative pruning is used to perform heuristic decision-making on task clusters. In this process, within the existing solution space, filtering boundaries are gradually set based on task characteristics (such as path length, image quality requirements, camera parameter matching, etc.). Paths that do not meet the current strategy are eliminated. This process is repeated multiple times to ultimately retain high-quality and comprehensive task paths.

[0039] Next, after pruning and optimization, a pre-planned strategy will be generated based on the retained high-quality path combinations. The pre-planned strategy refers to a control scheme, including paths, parameters, and execution order, defined before the formal execution of the task, used to guide the coordinated operation of the inspection drone and the onboard camera. The inspection drone is the device that performs actual path flight and photography tasks and has autonomous navigation capabilities; while the onboard camera is responsible for acquiring image or video data, and its parameter configuration must closely correspond to the task objectives, such as focal length, shooting angle, and exposure time.

[0040] Furthermore, to achieve flexible response and local adaptability, a dynamic lightweight module was deployed in the UAV's central control platform. This module has two core functions: first, local adjustment capability under new tasks, meaning that when a task is dynamically adjusted or a new target point is added, the path and parameters can be regenerated locally without global recalculation; second, rapid response capability in the face of unexpected situations or unsatisfactory acquisition results, such as when images are blurry or the UAV deviates due to wind, the module can immediately activate backup task points, adjust the path and camera parameters, and ensure the integrity of the acquisition task. Since the dynamic lightweight module is deployed in the central control system, it needs to achieve low computational overhead and fast response speed; therefore, a lightweight design is adopted to balance the efficiency and quality of task decision-making.

[0041] The inspection decision control is triggered by dynamic tasks. When a new inspection task is received or an unexpected situation such as invalid image acquisition in the original task is detected, the response process is automatically initiated to reassess and adjust the current inspection strategy. Dynamic tasks are characterized by temporality, uncertainty, and short response time. Therefore, its decision control mechanism needs to balance rapid response with strategy effectiveness.

[0042] The deep planning module and the dynamic lightweight module together constitute the core decision-making unit. The deep planning module undertakes comprehensive optimization of paths and parameters from a global perspective. Its high logical complexity and computational depth make it suitable for handling stable tasks or large-scale path updates. The dynamic lightweight module focuses on local adjustments and rapid responses, such as temporarily adjusting shooting angles or switching backup acquisition points during task execution. It features fast response speed and low computational burden. Both the deep planning and dynamic lightweight modules have built-in grid code maps, which are spatial structure data encoded within the task area. These are used to quickly locate task points, determine the relationship between images and locations, and assist the module in planning calculations. The existence of the grid code map ensures that the module can accurately locate and associate the original task points when making path or angle decisions.

[0043] Furthermore, a communication mechanism is established between the deep planning module and the dynamic lightweight module, allowing them to share status, parameters, and task change information in real time. When the dynamic lightweight module detects a local anomaly, it can immediately notify the deep planning module to synchronize and update its strategy. Conversely, after adjusting the path, the deep planning module can also notify the lightweight module to update its backup point configuration and temporary response strategy, ensuring that internal system information remains consistent during the execution of dynamic tasks, thereby improving response coordination and system stability.

[0044] Furthermore, this application also includes: constructing a first functional node based on semantic interpretation, constructing a second functional node based on heuristic decomposition, and constructing a third functional node based on perturbation addition; cascading the first functional node, the second functional node, and the third functional node to generate the heuristic port.

[0045] Specifically, semantic interpretation refers to the semantic analysis of the task description or instruction content in UAV inspection missions, converting natural language or structured input into operational information that the system can understand, and then constructing a first functional node responsible for recognizing the semantic elements of the task, such as recognizing the geographical direction, equipment type and task verb in "inspect the transformer in the northwest direction", and converting them into structured task units, thereby providing standardized input for subsequent task decomposition and path planning.

[0046] Subsequently, after semantic interpretation, the task is decomposed based on heuristics. A heuristic is a task-heuristic structure that instructs the system to extract multiple possible execution directions or sets of subtasks from the initial task, and then construct a second functional node to decompose the overall task based on spatial characteristics, priority, or task relevance. For example, in a task containing 20 inspection points, the second functional node can divide it into 4 groups of subtasks based on distance or voltage level, with each group containing 5 inspection points.

[0047] Next, to enhance the flexibility and adaptability of task planning, a perturbation mechanism is introduced to enhance the diversity of path planning or obstacle avoidance capabilities. This leads to the construction of a third functional node, which adds a certain number of perturbation points to the edge or gap area of ​​each subtask based on heuristics. For example, 1 to 2 backup points are randomly set within a range of 5 to 15 meters around each inspection subtask set, which can be activated when the main path is not feasible or image acquisition fails.

[0048] To enable three nodes with different responsibilities to work collaboratively, they are cascaded, connecting the first, second, and third functional nodes via data flow or task workflow to form a fully functional heuristic port. The heuristic port acts as a pre-processor for the entire path planning and task scheduling module, capable of outputting a heuristic, perturbable, and clearly hierarchical inspection task cluster from raw task text or structured data. Table 1 shows a partial record of the most recent heuristic port construction process.

[0049] Table 1: Partial record of the most recent heuristic port construction process Node number Construction basis Input type Core Function Description Output content Inter-node connection description Node 1 Semantic interpretation Inspection task semantics and target attribute description Transform task semantics into operational requirements and extract the core intent of the task. Initial path target, visual parameter requirements, etc. The output is connected to the input of node 2. Node 2 Heuristic task decomposition Node 1 output, task cluster characteristics Decompose the task cluster according to spatial characteristics and execution priority. Subtask sets, heuristic constraints, etc. The output is connected to the input of node 3. Node 3 Disturbance addition Node 2 output, perturbation rules (such as random radius) Insert perturbation points into the decomposed task set to improve path diversity. Heuristic path map, alternative shooting nodes, etc. Output to heuristic port generation module Heuristic Port Node cascading Outputs of nodes 1, 2, and 3 By aggregating the outputs of all functional nodes, a heuristic port with perturbation characteristics is constructed. Heuristic task entry data structure Furthermore, this application also includes: the inspection control system receiving the inspection task, triggering the heuristic port, driving the first functional node to perform task interpretation analysis, and determining the interpretation data, wherein the inspection task includes at least one inspection point; triggering the second functional node to identify the interpretation data and locate the heuristic direction, performing task decomposition based on the heuristic direction, and determining the decomposed tasks, wherein the smallest interpretation unit is used as the positioning standard; triggering the third functional node to introduce random radius perturbation points, adding perturbation heuristic directions for each decomposed task, and generating the heuristic direction task cluster, wherein the random radius of the perturbation point is customized.

[0050] Specifically, the inspection control system is the core module for scheduling and controlling inspection tasks, responsible for receiving inspection task instructions from the dispatch center or system platform. An inspection task includes multiple inspection points, each corresponding to the specific location and inspection target of a power transmission device or facility, such as a power tower or transformer on a power transmission line. Upon receiving an inspection task, an initiation port is triggered. This initiation port is the entry point for the task preprocessing module, used to initiate various functional nodes for task structure parsing and processing. Next, the first functional node performs task interpretation and analysis, using natural language processing or task protocol parsing technology to identify the semantic meaning, location information, and task target of each inspection point, thereby generating structured interpretation data.

[0051] Subsequently, a second functional node is triggered to further process the interpreted data. This second functional node locates the heuristic direction and performs task decomposition based on it. The heuristic direction refers to a directional vector in the inspection task that has spatial or logical guiding characteristics, such as the direction of transmission lines, the connection relationships between equipment, or task priority clues. Based on the heuristic direction, the entire inspection task is broken down into multiple smaller tasks, ensuring that each sub-task has spatial continuity or operational consistency, thereby improving path planning efficiency. During decomposition, the smallest interpretation unit is used as the positioning standard; the smallest execution unit of each task segment is an inspection point with an independent spatial location and operational objective. For example, in the original task, every three adjacent inspection points can be considered as a decomposition group unit, thus splitting 10 inspection points into 3 task groups.

[0052] Next, the third functional node is triggered, introducing a perturbation mechanism into each group of tasks after task decomposition. Perturbation points are introduced at locations where there is uncertainty in the task path or execution perspective. The location of the perturbation points is generated based on a customizable random radius. The introduction of perturbation points enhances the flexibility of path planning; for example, if a drone deviates from its original route due to wind or obstacles, it can turn to the nearest perturbation point to complete image acquisition. The random radius of the perturbation points can be set according to different equipment characteristics, such as 3 meters in a substation and 10 meters on a power transmission line in the field. When processing each decomposed task, several perturbation points are added, ultimately generating a complete heuristic task cluster. This cluster includes both the original inspection points and several backup perturbation points.

[0053] Furthermore, this application also includes: the visual parameters at least include camera focal length and shooting spatial angle; according to the inspection task, the range of the task inspection points is defined using the grid code map, and a solution space tree is constructed accordingly; the heuristic task cluster is traversed, and heuristic boundary calculation and pruning based on the solution space tree are performed, and an effective solution space tree is determined through multiple iterations; optimization decision is performed based on the effective solution space tree to determine the pre-planning strategy.

[0054] Specifically, visual parameters refer to a series of parameters that a drone needs to set when performing image acquisition tasks to ensure that the image quality meets the requirements of recognition and analysis. Among these, the camera focal length represents the distance from the optical center of the lens to the imaging plane, determining the image magnification and shooting distance range; the shooting spatial angle refers to the imaging angle range of the camera lens in space, including horizontal, vertical, and tilt angles, directly affecting the coverage area of ​​the photographable region. Therefore, the setting of visual parameters needs to be combined with the actual scene requirements, and the focal length and spatial angle should be reasonably selected to achieve clear acquisition of the target area by the device at different positions and angles.

[0055] After acquiring the inspection task, a grid code map is used to define the scope of the inspection points involved in the task. A grid code map is a regional division structure organized using a geographic grid coding method. Each grid cell corresponds to a spatial region and contains information about the inspection points and equipment within that region. Based on the grid code map, the approximate spatial range of each inspection point can be quickly determined, and a solution space tree can be constructed accordingly. A solution space tree is a hierarchical search structure used for path planning or parameter selection. Each node represents a possible decision or selection path, and its branches reflect different selection directions. The depth and breadth of the solution space tree determine the complexity of the solution space.

[0056] Next, heuristic bounds calculation and pruning are performed sequentially on each inspection task in the heuristic task cluster. Heuristic bounds calculation refers to calculating the reasonable search range allowed for the inspection task in the solution space tree by combining the current inspection task's goal and known conditions; while pruning refers to actively removing branches that exceed the goal limit or are invalid during the search process to reduce computational burden and avoid meaningless attempts. The multi-round iterative process enables the system to gradually converge to the most likely successful solution space region, ultimately selecting a structurally stable and highly efficient effective solution space tree.

[0057] The optimization decision-making process is based on the effective solution space tree. An optimal path or a set of optimal parameter configurations is selected from the effective solution space tree. The optimization process may combine multi-objective optimization algorithms, considering indicators such as image quality, path length, and energy consumption, ultimately outputting a pre-planned strategy. The pre-planned strategy is the system-recommended combination of inspection path and shooting configuration, used to guide the task execution in the subsequent actual execution phase. The effective solution space tree is a search structure that has undergone multiple iterations and heuristic pruning. Each node represents a feasible planning solution, such as a specific inspection path and its corresponding visual parameter combination. Each node in the effective solution space tree has already passed the initial screening of constraints, meeting the basic requirements of the task, such as covering all inspection points, meeting shooting clarity requirements, camera angle allowance, and safe flight radius. Optimization refers to the evaluation method of re-evaluating the relatively optimal solution among all solutions that have met the basic constraints. For this purpose, a fitness function is defined, which may comprehensively consider several objectives, such as path length, flight energy consumption, camera adjustment range, image overlap rate, and spatial angle satisfaction, to quantify and score each feasible solution. The higher the score, the better the solution.

[0058] Furthermore, this application also includes: determining a first heuristic task based on the heuristic task cluster; defining a solution boundary in the solution space tree based on the first heuristic task; pruning the solution space tree based on the solution boundary to determine a solution space tree once; determining a second heuristic task; performing a solution boundary definition and pruning of the solution space tree once; until the iteration of the heuristic task cluster is completed, and determining the effective solution space tree.

[0059] Specifically, a heuristic task cluster refers to a set of alternative task directions generated during task planning by introducing perturbation mechanisms and path diversity analysis. Each task direction represents a possible task execution path or method, including a combination of multiple dimensions such as the specific inspection point sequence, camera angle, and shooting perspective. A heuristic task is randomly selected from the heuristic task cluster as the starting point for processing, called the first heuristic task, which serves as the starting basis for subsequent pruning and optimization.

[0060] Next, after obtaining the first heuristic task, it needs to be placed into the solution space tree for solving. The solution boundary refers to the boundary of the search range set according to the characteristics of the first heuristic task (such as acceptable path length, camera parameter variation range, angle tolerance, etc.), that is, limiting which branches in the solution space are "worth exploring" and which branches can be directly discarded. According to the solution boundary criteria, the solution space tree is pruned, that is, branches that exceed the limits or do not meet the conditions of the first task direction are deleted, and finally a more concise first-order solution space tree is obtained.

[0061] Subsequently, a second heuristic task is randomly selected from the heuristic task cluster and used to further define new solution boundaries. The current solution space tree is pruned again to obtain a more refined solution space structure that is closer to the multi-objective requirements. Heuristic tasks are then continuously introduced to iteratively converge the solution space. Each round further compresses, filters, and optimizes the existing solution space until all heuristic tasks have completed the pruning process, forming the final effective solution space tree.

[0062] Furthermore, this application also includes: transmitting the pre-planned strategy to the inspection drone according to communication interaction, and executing strategy-driven inspection control; generating dynamic task instructions, wherein the dynamic task instructions are triggered when a new inspection task is added or an invalid captured image is obtained; and triggering the dynamic lightweight module according to the dynamic task instructions to execute the adjustment decision of the pre-planned strategy, determine and execute the updated inspection strategy.

[0063] Specifically, communication and interaction are achieved through a data communication link established between the ground control system and the UAV. Based on wireless protocols such as 5G, WiFi, or a dedicated low-latency data link, real-time synchronization of control commands is ensured. The pre-planned strategy is transmitted to the inspection UAV, sending pre-defined flight paths, camera parameters, and shooting locations to the UAV's execution terminal. The pre-planned strategy is the optimal inspection plan selected from multiple candidate paths in the early stages of the mission, encompassing control parameters such as the inspection point sequence, flight altitude, and shooting attitude. After receiving the pre-planned strategy, the inspection UAV executes its flight path and image acquisition tasks according to the predetermined plan.

[0064] Following this, during actual flight, mission changes or unexpected anomalies may occur, at which point dynamic mission commands will be generated. Dynamic mission commands are control commands temporarily added during the execution of an inspection mission to respond to environmental changes, image failures, or other uncertainties. Adding a new inspection mission is, for example, adding a power tower inspection point. Invalid images refer to images whose quality is below the set standard (e.g., less than 75% clarity) due to issues such as camera shake or insufficient lighting. Once these situations occur, corresponding dynamic mission commands will be automatically generated to trigger subsequent mission strategy adjustments. For example, if a relatively urgent temporary mission is assigned to the inspection drone, the built-in module will adjust the trajectory locally based on the original trajectory strategy, the spatial relationship of the grid code, and the acquisition requirements. While the pre-planned trajectory provides the optimal shooting point and already considers disturbance factors, unavoidable small-probability accidental factors or poor acquisition results may occur. Therefore, it is necessary to combine backup point acquisition to achieve the desired effect. The above decomposition is not based on the decomposition of inspection points, but on a task-oriented hierarchy. An inspection point may contain multiple heuristic tasks, and each heuristic task corresponds to an optimal collection point and two backup collection points. The backup collection points are dynamically triggered to make decisions when the effect is not good during the actual collection process, so as to avoid collection redundancy and improve the dynamic flexibility and adaptability of the overall inspection system.

[0065] Therefore, after a dynamic task instruction is generated, a dynamic lightweight module deployed within the UAV's central control system is triggered. This module quickly determines whether the current task status needs adjustment and modifies the inspection plan accordingly. The dynamic lightweight module executes adjustment decisions based on the pre-planned strategy, introducing new points, deleting invalid paths, or modifying camera configurations to adapt to the new situation. Finally, an updated inspection strategy is generated, and the UAV is directed to continue executing the adjusted task flow.

[0066] Furthermore, this application also includes: introducing a fast discriminator, wherein the validity or invalidity of the captured image is used as the discrimination criterion; as the inspection is executed with the pre-planned strategy, the captured image is received and image judgment based on the fast discriminator is performed; if the discrimination result is an invalid captured image, a dynamic task instruction is generated, wherein a backup collection point based on the inspection grid code is located as accompanying information of the instruction.

[0067] Specifically, a fast discriminator is introduced to quickly identify image quality. The main function of the fast discriminator is to perform real-time analysis of images captured by inspection drones or other inspection equipment. Its discrimination criteria are based on whether the image is valid. Valid images meet technical requirements such as sharpness, brightness, compositional integrity, and coverage of key areas, while invalid images may be unusable for subsequent identification and analysis due to issues such as camera shake, insufficient lighting, focal length deviation, or occlusion. The fast discriminator, based on a lightweight image classification model or feature matching algorithm, can provide results within hundreds of milliseconds after image acquisition, thus improving the real-time response capability of the entire system.

[0068] Next, as the pre-planned inspection strategy is executed—that is, after the inspection drone completes its shooting task at each inspection point according to the planned path—each acquired image is transmitted to the system for quality analysis. At this point, the rapid discriminator intervenes and performs image judgment. If the judgment result is an invalid captured image, it means that the current image cannot be used for the identification or analysis of the inspection target. In this case, a new dynamic task instruction is generated immediately to compensate for the information loss caused by the invalid data.

[0069] Meanwhile, the dynamic mission command includes a backup acquisition point for image re-acquisition. The backup acquisition point is located based on the inspection grid code; that is, the entire inspection area has been spatially divided according to a unified grid code, and each inspection point has its corresponding unique grid code. The backup acquisition point is a set of pre-set shooting candidate locations in the vicinity of the original mission point, no more than 10 meters from the origin and with better viewing conditions than the initial point. The backup acquisition point can quickly replace the original failed point as a new shooting target, thereby reducing mission delay. The accompanying information refers to the information about the backup acquisition point attached to the dynamic mission command, used for rapid execution after the UAV receives the command.

[0070] Furthermore, this application also includes: determining planning constraints, wherein the optimal acquisition point and at least two backup acquisition points are used as planning constraints for each heuristic task; wherein the optimal acquisition point imposes decision constraints on the depth planning module, and the backup acquisition points impose decision constraints on the dynamic lightweight module.

[0071] Specifically, defining planning constraints means clarifying various limitations and objectives when formulating execution strategies for inspection tasks, thereby ensuring the stability and flexibility of the inspection tasks. Each heuristic task requires a set of constraints, including an optimal acquisition point and at least two backup acquisition points. A heuristic task is a sub-task with heuristic weights generated based on task interpretation results, inspection target characteristics, spatial perspective requirements, and other factors. The optimal acquisition point is the best shooting location selected after comprehensive evaluation of multiple factors such as angle, field of view, lighting, and focal length, ensuring the clarity and completeness of the acquired images. Backup acquisition points are several pre-calculated alternative shooting points within the scope of the task path and spatial conditions, used to cope with situations where shooting cannot be completed due to interference or equipment malfunction during execution.

[0072] Furthermore, the optimal acquisition point serves as a decision constraint for the depth planning module. When calculating the inspection path, camera angle, and UAV flight trajectory, the depth planning module prioritizes satisfying the positioning and execution requirements of the optimal acquisition point. The depth planning module typically relies on graph search algorithms, 3D modeling, or reinforcement learning mechanisms to seek the optimal execution solution globally. Therefore, the accuracy and stability of the optimal acquisition point directly affect the quality of path planning. For example, in power transmission line inspection, if the optimal acquisition point is located 20 meters west of the tower and there is no obstruction to the view at that location, the path planning must ensure that the UAV hovers and takes pictures at that point.

[0073] In contrast, backup acquisition points serve as decision constraints for the dynamic lightweight module. The dynamic lightweight module is a real-time scheduling and adaptive module deployed in the UAV's central control system. Its main function is to handle sudden changes during mission execution, such as image recognition failures, wind speed disturbances, and sudden changes in lighting. In the event of an initial shooting failure, the depth planning module will quickly evaluate and select one of the backup acquisition points to perform a reshoot. Because the depth planning module has high requirements for response time and resource consumption, the backup acquisition points undergo lightweight modeling and viewpoint redundancy optimization to facilitate rapid switching decisions. For example, if the original shooting task fails, backup point 1 is located 10 meters north-east of the origin, and backup point 2 is located 8 meters south-west. The dynamic lightweight module will select the better task to execute within 300 milliseconds based on the environmental conditions.

[0074] Furthermore, this application also includes: acquiring inspection images, wherein the inspection images are associated with grid codes; connecting to a cloud server, uploading the inspection images and arranging them based on the grid code map to determine the inspection map; and performing terminal display and fault analysis and early warning management on the inspection map.

[0075] Specifically, acquiring inspection images refers to the image data collected by inspection drones using their onboard cameras during missions. This records the visual information of target equipment or facilities at specific locations and times, which is then used for subsequent detection, identification, and analysis. After acquiring the inspection images, they are associated with their corresponding grid codes. A grid code is a spatial coding identifier used to divide the actual geographic space into manageable grid areas, ensuring that each inspection image accurately corresponds to a specific location. By binding inspection images with grid codes, the shooting point can be quickly traced and located, improving the spatial organization efficiency of the images.

[0076] Next, the inspection images are uploaded to a cloud server via a network connection. The cloud server is a high-performance data processing center with capabilities for image storage, analysis, scheduling, and visualization. During the upload process, the images are arranged using an established grid code map. A grid code map is an image distribution map organized according to geographic grid units, forming a structured spatial representation of images within the task area. Images are categorized and mapped to corresponding areas according to grid numbers, thus forming a comprehensive and clearly defined inspection map. The grid code map not only displays the current task's coverage area but also provides spatial evidence for subsequent identification of abnormal areas and comparison of changing trends.

[0077] Subsequently, the inspection map is used for terminal display and fault analysis and early warning management. Terminal display refers to presenting the map structure on the user interface (such as a monitoring platform, tablet terminal, or operator station), allowing maintenance personnel to clearly view the image data and status results of each grid area. Fault analysis and early warning management refers to using intelligent recognition algorithms to determine whether there are potential anomalies based on the feature information in the grid code image (such as cracks, oil stains, water stains, breakpoints, etc. on the equipment surface), and generating risk warnings in combination with historical records. If a grid cell shows image anomalies consecutively, the risk level can be automatically marked and maintenance personnel can be alerted to intervene. For example, in power line inspection, if the grid location of a tower numbered T47 shows insulator damage in two consecutive images, the area will be marked in red on the map and a level three early warning notification will be issued.

[0078] In summary, the UAV intelligent inspection control method based on low-altitude grid codes provided in this application has the following technical effects: by realizing the technical goal of adaptive planning and dynamic control of inspection tasks based on grid code maps, it achieves the technical effects of improving path adaptability, image acquisition success rate and overall task execution stability in a variable inspection environment.

[0079] Example 2: Based on the same inventive concept as the UAV intelligent inspection control method based on low-altitude grid codes in the previous examples, this application also provides a UAV intelligent inspection control system based on low-altitude grid codes. Please refer to the appendix. Figure 2The system includes: a grid code planning module 11, used to plan a grid code map based on inspection points for inspection scenarios, wherein each inspection point is assigned a grid code associated with location information and equipment information; an inspection task interpretation module 12, used to receive inspection tasks, introduce heuristic ports, perform inspection task interpretation, and add random radius perturbation points as heuristic task clusters; and an inspection decision control module 13, used to perform boundary iterative pruning and optimization planning based on heuristic decision for the heuristic task clusters, with inspection path-visual parameters as the planning target, according to the deep planning module deployed in the inspection control system, to determine the pre-planning strategy and drive the inspection drone and airborne camera, and to perform inspection decision control under dynamic task triggering based on the dynamic lightweight module deployed in the drone central control, wherein the deep planning module and the dynamic lightweight module have the grid code map built in, and communication interaction is established between the modules.

[0080] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used to: construct a first functional node based on semantic interpretation, construct a second functional node based on heuristic decomposition, and construct a third functional node based on disturbance addition; cascade the first functional node, the second functional node, and the third functional node to generate the heuristic port.

[0081] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used for: receiving the inspection task, triggering the revelation port, driving the first functional node to perform task interpretation analysis, and determining the interpretation data, wherein the inspection task includes at least one inspection point; triggering the second functional node to identify the interpretation data and locate the heuristic direction, performing task decomposition based on the heuristic direction, and determining the decomposed tasks, wherein the smallest interpretation unit is used as the positioning standard; triggering the third functional node to introduce random radius perturbation points, adding perturbation heuristic directions for each decomposed task, and generating the heuristic direction task cluster, wherein the random radius of the perturbation points is customized.

[0082] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used for: the visual parameters including at least camera focal length and shooting spatial angle; according to the inspection task, defining the range of the task inspection points based on the grid code map, and constructing a solution space tree accordingly; traversing the heuristic task cluster, performing heuristic boundary calculation and pruning based on the solution space tree, and determining the effective solution space tree through multiple iterations; and performing optimization decision-making based on the effective solution space tree to determine the pre-planning strategy.

[0083] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used for: determining a first heuristic task based on the heuristic task cluster; defining solution boundaries in the solution space tree based on the first heuristic task; pruning the solution space tree based on the solution boundaries to determine a solution space tree; determining a second heuristic task; performing solution boundary definition and pruning of the solution space tree once; until the iteration of the heuristic task cluster is completed, and determining the effective solution space tree.

[0084] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used for: transmitting the pre-planned strategy to the inspection UAV according to communication interaction, and executing strategy-driven inspection control; generating dynamic task instructions, wherein the dynamic task instructions are triggered when a new inspection task is added or an invalid captured image is obtained; and triggering the dynamic lightweight module according to the dynamic task instructions to execute the adjustment decision of the pre-planned strategy, determine the updated inspection strategy, and execute it.

[0085] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used to: introduce a rapid discriminator, wherein the validity or invalidity of the captured image is used as the discrimination criterion; as the inspection is executed according to the pre-planned strategy, the captured image is received and image judgment based on the rapid discriminator is performed; if the judgment result is an invalid captured image, a dynamic task instruction is generated, wherein a backup collection point based on the inspection grid code is located as accompanying information of the instruction.

[0086] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used to: determine planning constraints, wherein the optimal acquisition point and at least two backup acquisition points are used as planning constraints for each heuristic task; wherein the optimal acquisition point imposes decision constraints on the depth planning module, and the backup acquisition points impose decision constraints on the dynamic lightweight module.

[0087] Furthermore, the UAV intelligent inspection control system based on low-altitude grid codes is also used for: acquiring inspection images, wherein the inspection images are associated with grid codes; connecting to a cloud server, uploading the inspection images and arranging them based on the grid code map to determine the inspection map; and performing terminal display and fault analysis and early warning management on the inspection map.

[0088] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The UAV intelligent inspection control method and specific examples based on low-altitude grid codes in the aforementioned Embodiment 1 are also applicable to the UAV intelligent inspection control system based on low-altitude grid codes in this embodiment. Through the foregoing detailed description of the UAV intelligent inspection control method based on low-altitude grid codes, those skilled in the art can clearly understand the UAV intelligent inspection control system based on low-altitude grid codes in this embodiment. Therefore, for the sake of brevity, it will not be described in detail here.

[0089] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0090] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of this application and its equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for intelligent inspection and control of unmanned aerial vehicles based on low-altitude grid codes, characterized in that, The method includes: For inspection scenarios, a grid code map based on inspection points is planned, in which each inspection point is assigned a grid code that is associated with location information and equipment information; Receive inspection tasks, introduce heuristic ports, execute inspection task interpretation and add random radius perturbation points as heuristic task clusters; Using the inspection path-visual parameters as the planning target, and based on the deep planning module deployed in the inspection control system, the system performs boundary iterative pruning and optimization planning based on heuristic decision-making for the heuristic task cluster, determines the pre-planning strategy and drives the inspection drone and airborne camera. Based on the deployment of a dynamic lightweight module in the drone's central control unit, the system executes inspection decision control under dynamic task triggering. The deep planning module and the dynamic lightweight module have the grid code map built in, and communication interaction is established between the modules.

2. The UAV intelligent inspection and control method based on low-altitude grid codes as described in claim 1, characterized in that, Introducing heuristic ports, including: The first functional node is constructed based on semantic interpretation, the second functional node is constructed based on heuristic decomposition, and the third functional node is constructed based on perturbation addition. The first functional node, the second functional node, and the third functional node are cascaded to generate the heuristic port.

3. The UAV intelligent inspection control method based on low-altitude grid codes as described in claim 2, characterized in that, Perform inspection tasks to interpret and add random radius perturbation points as heuristic task clusters, including: The inspection control system receives the inspection task, triggers the revelation port, drives the first functional node to perform task interpretation and analysis, and determines the interpretation data. The inspection task includes at least one inspection point. The second functional node is triggered to identify the interpreted data and locate the heuristic direction, perform task decomposition based on the heuristic direction, and determine the decomposed task, wherein the smallest interpretation unit is used as the positioning standard; The third functional node is triggered, a random radius perturbation point is introduced, and the perturbation heuristic is added to the task decomposed step by step to generate the heuristic task cluster, wherein the random radius of the perturbation point is customized.

4. The UAV intelligent inspection and control method based on low-altitude grid codes as described in claim 1, characterized in that, The visual parameters include at least the camera focal length and the shooting spatial angle; Based on the inspection task, the range of the inspection points is defined using the grid code map, and a solution space tree is constructed accordingly. Traverse the heuristic task cluster, perform heuristic boundary calculation and pruning based on the solution space tree, and determine the effective solution space tree through multiple rounds of iteration; Perform optimization decisions based on the effective solution space tree to determine the pre-planning strategy.

5. The UAV intelligent inspection control method based on low-altitude grid codes as described in claim 4, characterized in that, For the heuristic approach, perform heuristic decision-based bounded iterative pruning and optimization planning on the task cluster to determine the pre-planning strategy, including: Based on the heuristic task cluster, determine the first heuristic task; Based on the first heuristic task, a solution boundary is defined in the solution space tree, and the solution space tree is pruned according to the solution boundary to determine a first solution space tree; Determine the second heuristic task, perform a solution bound definition and pruning of the solution space tree once, until the iteration of the heuristic task cluster is completed, and determine the effective solution space tree.

6. The UAV intelligent inspection control method based on low-altitude grid codes as described in claim 1, characterized in that, Execution of inspection decision control triggered by dynamic tasks includes: Based on the communication interaction, the pre-planned strategy is transmitted to the inspection drone to execute strategy-driven inspection control; Generate dynamic task instructions, wherein the dynamic task instructions are triggered when a new inspection task is added or an invalid captured image is obtained; Based on the dynamic task instruction, the dynamic lightweight module is triggered to execute the adjustment decision of the pre-planned strategy, determine the update inspection strategy, and execute it.

7. The UAV intelligent inspection and control method based on low-altitude grid codes as described in claim 6, characterized in that, Generate dynamic task instructions, which are triggered upon receiving invalid captured images, including: A fast discriminator is introduced, in which the validity or invalidity of the captured image is used as the discrimination criterion; As the pre-planned strategy is executed during inspection, captured images are received and image determination is performed based on the fast discriminator. If the determination result is an invalid captured image, a dynamic task instruction is generated, wherein a backup collection point based on the inspection grid code is located as accompanying information of the instruction.

8. The UAV intelligent inspection control method based on low-altitude grid codes as described in claim 1, characterized in that, Determine the planning constraints, wherein the optimal data collection point and at least two backup data collection points are used as the planning constraints for each heuristic task. The optimal acquisition point serves as a decision constraint for the depth planning module, while the backup acquisition point serves as a decision constraint for the dynamic lightweight module.

9. The UAV intelligent inspection control method based on low-altitude grid codes as described in claim 1, characterized in that, After executing inspection decision control triggered by dynamic tasks, the following is included: Acquire inspection images, wherein the inspection images are associated with grid codes; Connect to the cloud server, upload the inspection images and arrange them based on the grid code map to determine the inspection map; The inspection map is displayed on the terminal and managed for fault analysis and early warning.

10. A UAV intelligent inspection and control system based on low-altitude grid codes, characterized in that, The steps for implementing the UAV intelligent inspection control method based on low-altitude grid codes according to any one of claims 1 to 9 include: The grid code planning module is used to plan a grid code map based on inspection points for inspection scenarios. Each inspection point is assigned a grid code that is associated with location information and equipment information. The inspection task interpretation module is used to receive inspection tasks, introduce heuristic ports, perform inspection task interpretation, and add random radius perturbation points as heuristic task clusters. The inspection decision control module is used to perform heuristic decision-based boundary iterative pruning and optimization planning for the heuristic task cluster, based on the inspection path-visual parameters as the planning target and the deep planning module deployed in the inspection control system. It determines the pre-planning strategy and drives the inspection drone and airborne camera. Based on the dynamic lightweight module deployed in the drone's central control, it performs inspection decision control under dynamic task triggering. The deep planning module and the dynamic lightweight module have the grid code map built in, and the modules establish communication interaction.

Images