A water cushion pond inspection method and system

Abstract

The application discloses a water cushion pond inspection method and system, and relates to the field of water cushion pond robot inspection.The method comprises the following steps: firstly, a communication connection is established with an underwater robot; secondly, map information of a target water cushion pond is acquired; thirdly, a target inspection route is determined, and the underwater robot is controlled to perform inspection; and finally, an inspection result is received and the inspection result is analyzed.Through the above design, the specific inspection depth is positioned through a historical water cushion pond hierarchical diagram, and inspection is performed at multiple depths, so that comprehensive monitoring and management of the water cushion pond are effectively realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of water cushion pond robot inspection, in particular to a water cushion pond inspection method and system. BACKGROUND

[0002] Traditional water cushion pond inspection methods often require manual operation, which is low in efficiency and prone to safety accidents. Especially for the depth inspection of water cushion ponds, due to the lack of effective equipment and methods, it is often difficult to achieve accuracy and comprehensiveness.

[0003] In addition, the traditional method often cannot conduct detailed inspection for each depth in the water cushion pond, resulting in some potential problems that may be ignored. Therefore, it is particularly important to develop a new type of water cushion pond inspection method that can conduct comprehensive inspection at multiple depths. SUMMARY

[0004] The purpose of the present application is to provide a water cushion pond inspection method and system.

[0005] In a first aspect, the present application provides a water cushion pond inspection method, the method comprising:

[0006] In response to a patrol task start instruction for a target water cushion pond, a communication connection with an underwater robot is established;

[0007] Obtain water cushion pond map information corresponding to the target water cushion pond, the water cushion pond map information comprising a plurality of depth historical water cushion level maps;

[0008] According to the historical water cushion level map corresponding to the target depth, determine a target inspection route and control the underwater robot to conduct inspection, the target depth being any depth in the plurality of depths;

[0009] Receive the inspection results of the underwater robot for the plurality of depths;

[0010] Analyze the inspection results to obtain an inspection analysis report of the inspection results.

[0011] In a second aspect, the present application provides a server system comprising a server, the server being configured to execute the method of the first aspect.

[0012] The present application has the following beneficial effects compared with the prior art:

[0013] The application provides a water cushion pond inspection method and system, which establishes a communication connection with an underwater robot, obtains map information of a target water cushion pond, determines a target inspection route, controls the underwater robot to perform inspection, and finally receives and analyzes the inspection result. Thus, the specific inspection depth is positioned through the historical water cushion pond hierarchical map, and the inspection is performed at multiple depths, so that the comprehensive monitoring and management of the water cushion pond are effectively realized. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed to be used in the embodiments. It should be understood that the following drawings only show some of the embodiments of the application, and therefore should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can also be obtained without creative labor.

[0015] Figure 1 The water cushion pond inspection method provided by the embodiment of the application is shown in the flowchart.

[0016] Figure 2 The structure schematic diagram of the computer device provided by the embodiment of the application is shown in the block diagram. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the application more clear, the following will combine the drawings in the preferred embodiments of the application to describe the technical solutions in the embodiments of the application in more detail. In the drawings, the same or similar notations represent the same or similar parts or parts with the same or similar functions throughout. The described embodiments are part of the embodiments of the application, not all of the embodiments. The embodiments described below by referring to the drawings are exemplary, and are intended to explain the application, and cannot be understood as a limitation on the application. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0018] The embodiments of the application will be described in detail below with reference to the drawings.

[0019] In order to solve the technical problems in the foregoing background art, Figure 1 The flowchart of the water cushion pond inspection method provided by the embodiment of the application is shown, and the following will introduce the water cushion pond inspection method in detail.

[0020] Step S201, in response to the inspection task starting instruction for the target water cushion pond, a communication connection with the underwater robot is established;

[0021] In step S202, water mat pond map information corresponding to the target water mat pond is acquired, and the water mat pond map information includes a plurality of historical water mat pond level maps of different depths.

[0022] In step S203, a target inspection route is determined according to the historical water mat pond level map corresponding to a target depth, and the underwater robot is controlled to perform inspection, and the target depth is any one of the plurality of depths.

[0023] In step S204, the inspection results of the underwater robot for the plurality of depths are received.

[0024] In step S205, the inspection results are analyzed to obtain an inspection analysis report of the inspection results.

[0025] In the embodiment of the invention, the water basin manager specifies the target water basin that needs to be inspected in the inspection plan and sends the start instruction of the inspection task through the control panel or mobile application. After receiving the instruction, the underwater robot establishes a stable communication connection with the control center using a wireless communication device. This may involve the use of Bluetooth, Wi-Fi or special communication protocols and other technologies. The underwater robot is equipped with various sensors and scanning devices such as sonar, laser radar and camera, etc. Before the inspection begins, the underwater robot will start these sensors and devices to scan and measure the water basin. The data obtained by scanning includes the geometric structure of the underwater environment, the distribution of obstacles and the water quality, etc. These data will be processed and integrated to form the water basin map information, which includes the historical records of different depths, providing the hierarchical structure and trend of the water basin. Based on the target depth, the underwater robot uses the data of the historical water basin hierarchical map to plan the inspection route. For example, if the target depth is 10 meters, the underwater robot will determine the inspection path according to the historical hierarchical map corresponding to this depth to cover the area within this depth range. The inspection path may pass through known underwater attractions, key areas or potential problem areas to ensure comprehensive and effective inspection of the target depth. The underwater robot starts to execute the task according to the planned inspection route, while using sensors and devices to monitor the water basin in real time. During the inspection process, the underwater robot records and transmits the inspection result data to the control center. The inspection result may include measured water quality parameters, detected obstacle positions, image or video records, etc. For each depth, the robot will send back the corresponding inspection result for further analysis and processing. After receiving the inspection result, the control center or operator will conduct in-depth analysis and interpretation of the data. The analysis process may include the following aspects: water quality analysis: based on the measurement data collected by the underwater robot, such as pH value, dissolved oxygen, turbidity, etc., the water quality of the water basin is evaluated. These data can be compared with the preset standards or historical data to determine whether there are water quality abnormalities or pollution problems. Obstacle detection: according to the sensor data obtained by the underwater robot, obstacles in the water basin such as water grass, floating objects or other potential objects are identified and analyzed. This helps to evaluate the overall cleanliness and navigation safety of the water basin. Topographic analysis: using the topographic data recorded by the underwater robot, a water bottom topographic map is generated, and the undulations, slopes and unevenness of the water basin bottom are analyzed. This can be used to evaluate the geomorphic features of the water basin and detect potential sediment or soil erosion. The inspection analysis report will include the evaluation results of each aspect, the indication of abnormal conditions and the recommended improvement measures. The report may contain charts, data visualization and textual descriptions to enable the water basin manager to fully understand the inspection results and make corresponding decisions. In addition, the report can record historical inspection data, trend analysis and long-term monitoring recommendations to support the maintenance and management of the water basin.

[0026] In the embodiments of the present application, the foregoing step S203 can be implemented by the following detailed implementation.

[0027] (1) Based on iteration of the historical water sump level diagram, a to-be-inspected water sump level diagram is obtained, wherein the to-be-inspected water sump level diagram is the historical water sump level diagram after iteration, and the inspection route of the historical water sump level diagram is the historical inspection route;

[0028] (2) In the process of operating the underwater robot on the to-be-inspected water sump level diagram according to the historical inspection route, the non-passable area is identified to obtain a target non-passable area;

[0029] (3) According to the detour starting position, a to-be-determined non-passable range is determined;

[0030] (4) According to at least one of the first spatial span and the second spatial span, a region unit detour cost of each current region unit in the to-be-determined non-passable range is determined, wherein the first spatial span is a spatial span between the current region unit and the detour starting position, and the second spatial span is a spatial span between the current region unit and the target non-passable area;

[0031] (5) In the to-be-determined non-passable range, a current region unit group on each candidate inspection strategy in a plurality of candidate inspection strategies is obtained;

[0032] (6) According to the region unit detour cost, a region unit detour cost array corresponding to the current region unit group is determined;

[0033] (7) The region unit detour cost array is data-regularized to obtain a detour cost corresponding to each candidate inspection strategy;

[0034] (8) According to the detour cost, at least one detour strategy is selected from the plurality of candidate inspection strategies;

[0035] (9) According to the at least one detour strategy, a current candidate range in the to-be-determined non-passable range is determined;

[0036] (10) In the current candidate range, a passable region unit group adjacent to a current starting position region unit is traversed, wherein the current starting position region unit is a current region unit corresponding to the detour starting position;

[0037] (11) An inspection cost corresponding to each current region unit in the passable region unit group is obtained, and a latest traversed current region unit corresponding to the obtained inspection cost is determined as a previous level region unit, wherein the inspection cost is a cost from the current starting position region unit to a current exit position region unit through the corresponding current region unit;

[0038] (12) traversing a next set of passable region cells adjacent to a target current region cell until a current departure location region cell is included in the traversed set of passable region cells, terminating the traversal and performing a previous region cell backtracking from the current departure location region cell to a current origin location region cell to obtain a target detour route between the detour origin location and the detour departure location, wherein the target current region cell is a current region cell that has not been traversed and has an optimal inspection cost, and wherein the target detour route is used to detour around the target impassable region from the detour origin location to the detour departure location, the detour origin location being a detour start on the historical inspection route, and the detour departure location being a detour end on the historical inspection route;

[0039] (13) adjusting the historical inspection route according to the target detour route to obtain a target inspection route;

[0040] (14) operating the underwater robot to perform the inspection on the to-be-inspected water mat level map according to the target inspection route.

[0041] In the embodiments of the present application, it is assumed that the historical water mat level map is used as an initial map for iterative processing in a water mat. By analyzing the shape, depth, and obstacles of the current water mat, the to-be-inspected water mat level map is updated to more accurately plan the inspection path. When the underwater robot performs the inspection in the water mat, it uses sensors to detect the underwater environment. During the detection process, if the robot detects rocks, branches, or other obstacles, these detected regions will be marked as target impassable regions. The detour origin location can be the starting position of the robot or the starting point required by a specific task. A boundary is set around the target impassable region, which forms a tentative impassable range. The spatial span can be defined according to the physical characteristics in the water mat, such as the distance or path length between two region cells. When calculating the region cell detour cost, other factors such as water flow speed, energy consumption, etc. may also be considered. The candidate inspection strategy can be a set of predefined rules for selecting the next move of the robot during the inspection process. The current region cell set is a set of region cells adjacent to the current position of the robot and passable, determined by the candidate inspection strategy. For each region cell in the current region cell set, its detour cost is calculated and stored in an array. The detour cost array reflects the detour cost of each region cell, which helps to evaluate the pros and cons of different region cell sets. Based on the region cell detour cost array, the region cell set with the lowest detour cost is selected as one of the detour strategies. This ensures that the robot selects the most economical and efficient path during the inspection process.

[0042] Assume that the detour strategy A is selected as the current detour strategy. The detour strategy A requires the robot to select a specific set of adjacent and passable region cells to the current location. Based on this strategy, the current candidate range that meets the conditions in the pending impassable range is determined. In the current candidate range, there are multiple adjacent sets of passable region cells. The underwater robot will traverse these adjacent sets of region cells in turn to evaluate the inspection cost of each set of region cells and continue to explore the next step. For the current region cell in each adjacent set of passable region cells, the underwater robot obtains the inspection cost of reaching the region cell. The inspection cost is obtained by calculating the time, energy consumption or other indicators required to reach the current region cell from the current starting location region cell. After obtaining the inspection cost, the robot updates the current region cell to the previous level region cell and continues to backtrack to the next step.

[0043] In the traversal process, when the set of passable region cells contains the current departure location region cell, the robot terminates the traversal operation. At this time, the robot starts from the current departure location region cell to the current starting location region cell to backtrack the previous level region cell to form a complete detour route. For example, assume that there is a historical inspection route consisting of the following region cells: A→B→C→D. When performing inspection, the robot detects that there is an obstacle between region cells B and C, which is marked as the target impassable region. According to the historical inspection route, A is selected as the detour starting location and D is selected as the detour departure location. According to the detour strategy, a set of adjacent and passable region cells to the current location is selected. Assume that the current robot is located at region cell B, then the current candidate range includes adjacent region cells A, C and D. The robot starts from region cell B to traverse the adjacent set of passable region cells: A, C and D. The robot calculates the inspection cost of reaching the current region cell (A, C or D) from the starting location (A) and selects the optimal current region cell. Assume that the inspection cost is as follows: A: 10 minutes; C: 15 minutes; D: 12 minutes. According to the lowest inspection cost, the robot selects region cell A as the target current region cell. Since the target current region cell (A) contains the detour departure location (D) in the traversal process, the traversal operation is terminated. Then, the robot starts from the detour departure location (D) to the detour starting location (A) to backtrack the previous level region cell. Through the backtracking operation, the robot obtains the target detour route: detour starting location: A, detour departure location: D, target detour route: A→B→C→D. The robot forms the target detour route between the detour starting location and the detour departure location by traversing the adjacent set of passable region cells and selecting the current region cell with the lowest inspection cost. This detour route can help the robot avoid the target impassable region and perform the task according to the historical inspection route.

[0044] Based on the robot computing of the bypass route and the backtracking process, the target bypass route is inserted into the historical inspection route. In this way, a new target inspection route is obtained, which combines historical data and optimized bypass paths. Finally, according to the target inspection route, the robot performs specific inspection tasks on the water cushion pond level map to be inspected. The robot will move in the order indicated by the target inspection route and perform corresponding inspection operations on each target area unit to collect data or perform specific tasks.

[0045] In the embodiments of the present application, when the at least one bypass strategy is a plurality of bypass strategies, the aforementioned step of determining the current candidate range in the pending impassable range according to the at least one bypass strategy can be implemented by the following examples.

[0046] A plurality of local candidate ranges corresponding to a plurality of bypass strategies are obtained, and the plurality of local candidate ranges are combined to form the current candidate range in the pending impassable range. Alternatively, a target bypass strategy is determined from the plurality of bypass strategies according to the current departure location area unit, and a candidate range on the target bypass strategy is determined as the current candidate range in the pending impassable range, wherein the current departure location area unit is a current area unit corresponding to the bypass departure location.

[0047] In the embodiments of the present application, in the inspection task, there can be a plurality of bypass strategies to choose from. These strategies can be different according to specific needs and scenarios. For example: shortest path strategy: select the nearest passable area to the current position as the next target. Optimal cost strategy: select the lowest cost passable area as the next target according to factors such as inspection time or energy consumption. Random strategy: randomly select a passable area as the next target. The current candidate range is determined according to the bypass strategy, which includes the local candidate range and the pending impassable range.

[0048] a) Obtain the local candidate range corresponding to the plurality of bypass strategies: Assume that there are two bypass strategies, the shortest path strategy and the optimal cost strategy. For the shortest path strategy, the local candidate range is area units A, C and D (adjacent area units). For the optimal cost strategy, the local candidate range is area units B, C and D. The two local candidate ranges are combined to form the current candidate range in the pending impassable range, i.e. A, B, C and D.

[0049] b) Determine the target bypass strategy according to the current departure location area unit: Assume that the recorded historical inspection route is A→B→C→D, and D is the current departure location area unit. According to the historical inspection route, the target bypass strategy can be determined as the shortest path strategy, because it needs to choose the shortest path to return from D to A. Then, the candidate range (A, B, C) on the target bypass strategy is determined as the current candidate range in the pending impassable range.

[0050] If multiple detour strategies are available, the corresponding local candidate ranges of each strategy can be obtained according to each strategy, and these local candidate ranges are combined together to form the current candidate range. In addition, a target detour strategy can also be selected from the multiple detour strategies according to the current departure location area unit, and the candidate range of the target detour strategy is taken as the current candidate range. In this way, the robot can perform path planning and decision-making according to the current candidate range to bypass the target impassable area and complete the inspection task.

[0051] In the embodiments of the present application, before the step of performing the aforementioned iteration based on the historical water sump hierarchical map to obtain the water sump hierarchical map to be inspected, the following implementation can also be included.

[0052] (1) The historical water sump hierarchical map is segmented into area units to obtain an area unit set, wherein the area unit set includes a plurality of historical area units;

[0053] (2) In the area unit set, the historical starting location area unit is taken as the starting point, and a plurality of sensing paths corresponding to a plurality of preset deflection amounts are determined according to a preset sensing interval;

[0054] (3) According to a preset rule, a reliable sensing path array is extracted from the plurality of sensing paths, wherein the preset rule is determined according to at least one of a sensing distance and a preset command of the target underwater robot;

[0055] (4) A preset number of reliable sensing paths adjacent in a preset deflection amount order are obtained from the reliable sensing path array, and a sensing path range corresponding to the preset number of reliable sensing paths is obtained;

[0056] (5) The historical area units within the sensing path range greater than the preset range are combined to form a first historical area unit group connected to the historical starting location area unit;

[0057] (6) When the first historical area unit group does not include the historical departure location area unit, the following processing is continuously performed:

[0058] Distance sensing of a target sensing interval is performed from the current historical area unit group as the starting point to obtain a next historical area unit group connected to the current historical area unit group, wherein the connection coincidence rate of the distance sensing and the target sensing interval has a positive feedback relationship;

[0059] (7) When the obtained next historical area unit group includes the historical departure location area unit, the distance sensing is ended, and the historical starting location area unit and all historical area unit groups are combined to form a connected preset number of historical area units;

[0060] (8) The preset number of historical area units are determined as the historical candidate range;

[0061] (9) According to the historical candidate range, the historical starting position area unit and the historical leaving position area unit are routed to obtain a historical inspection route, wherein the historical starting position area unit is a historical area unit corresponding to a starting inspection position of the historical water sump hierarchical map, the historical leaving position area unit is a historical area unit corresponding to a leaving inspection position of the historical water sump hierarchical map, and the preset number of historical area units at least includes the historical starting position area unit and the historical leaving position area unit.

[0062] In the embodiment of the present application, the preparation work before iteration of the historical water sump hierarchical map can include:

[0063] a) Area unit segmentation: performing area unit segmentation on the historical water sump hierarchical map to obtain a set of area unit collections. For example, the water sump is divided into regular grids, and each grid unit represents an area unit.

[0064] b) Preset sensing path: in the area unit collection, taking the historical starting position area unit as the starting point, a sensing path corresponding to each preset deflection amount is determined according to a preset sensing interval. These sensing paths can be used to detect information of different area units. For example, starting from the starting point area unit, according to the preset sensing interval, the area unit of the next sensing path is selected along the preset deflection amount in turn.

[0065] c) Extracting reliable sensing path array: according to the preset rule, the reliable sensing path array is extracted from the plurality of sensing paths. The preset rule can be determined based on the sensing distance or the preset command of the target underwater robot. For example, according to the set minimum sensing distance and the communication quality requirement of the target robot, the path with good communication quality of the target robot in the sensing path array is selected.

[0066] d) Obtaining sensing path range: obtaining a preset number of reliable sensing paths adjacent in the preset deflection amount order from the reliable sensing path array, and obtaining the sensing path range corresponding to the sensing paths. For example, if the preset number is 3 and the reliable sensing path array is [A, B, C, D, E], the sensing path range may be [A, B, C], [B, C, D] and [C, D, E].

[0067] The iteration process of the historical water sump hierarchical map can include:

[0068] a) Target Sensing Interval Distance Sensing: Starting from the current historical region unit group, distance sensing is performed according to the target sensing interval. This allows the robot to identify the next historical region unit group connected to the current one. The distance sensing results guide the robot in selecting the next region unit group. For example, based on the set distance sensing interval, the robot starts from the current region unit group and senses the next connected region unit group within a certain range.

[0069] b) Determine if the next historical region unit group includes the historical departure location region unit: If the obtained next historical region unit group includes the historical departure location region unit, then the distance sensing ends. At this time, the historical origin location region unit and all historical region unit groups are combined to form a preset number of connected historical region units.

[0070] Determining the historical candidate range may include identifying a preset number of historical region units as the historical candidate range. These historical region units will serve as candidate paths for the robot's inspection. For example, if the preset number is 5, then several historical region units related to the historical starting position and historical departure position can be selected as the historical candidate range.

[0071] The layout of historical inspection routes can include arranging detour routes for historical starting position area units and historical departure position area units based on historical candidate ranges to obtain historical inspection routes. The historical starting position area unit corresponds to the inspection starting position in the historical stilling basin hierarchy map, while the historical departure position area unit corresponds to the inspection departure position in the historical stilling basin hierarchy map. A preset number of historical area units include at least historical starting position area units and historical departure position area units. For example, based on the selection of historical candidate ranges, paths are planned between area units to form historical inspection routes. For instance, suppose there is an inspection task for a stilling basin, which is divided into the following area units (represented by letters): A, B, C, D, E, F, G, H, I. The historical starting position area unit is A, and the historical departure position area unit is I.

[0072] Now the historical inspection route is planned according to the historical candidate range. Assuming that the preset number is 5, the two historical region units before and after the historical start location and the historical departure location are selected as the historical candidate range, i.e. [A, B, C, D, E, I]. Next, the detour route is laid out according to the historical candidate range to form the historical inspection route. The specific steps are as follows: starting from the historical start location region unit A, the region unit B adjacent to it is selected as the next inspection point. From B, the region unit C adjacent to it is selected as the next inspection point. Then, from C to D, D is selected as the next inspection point. Next, from D, E is selected as the next inspection point. Finally, from E to I, I is selected as the last point of the inspection. According to the above steps, the historical inspection route obtained is A→B→C→D→E→I.

[0073] In this example, the two region units before and after the historical start location and the historical departure location are selected as the limited range of the inspection path according to the historical candidate range. Then, according to the connection relationship between the region units, the next inspection point is selected step by step from the start location until the departure location is reached. In this way, a historical inspection route can be planned to meet the requirements of the inspection task.

[0074] In the embodiment of the present application, before the step of performing the distance sensing of the target sensing interval with the current historical region unit group as the starting point to obtain the next historical region unit group connected to the current historical region unit group, the following examples can also be included.

[0075] (1) Obtain the number of identified region units and the number of connected region units, wherein the number of identified region units is the number of historical region units identified before the distance sensing of the current round, and the number of connected region units is the number of connected historical region units identified before the distance sensing of the current round.

[0076] (2) Obtain the target sensing interval which has a negative feedback relationship with the number of identified region units and a positive feedback relationship with the number of connected region units.

[0077] In an embodiment of the present application, it is assumed that the robot starts the inspection from the area unit A. During the inspection, the robot uses sensors for target sensing and records the next set of historical area units (e.g., B, C, D) connected to the current set of historical area units (i.e., A). Through distance sensing, the robot can determine the distance between each connected set of area units and the current set of area units. Before the target sensing interval, the robot has already identified some historical area units. It is assumed that before the current round, the robot has identified historical area units A, B, and C. At this time, the number of identified area units is 3. At the same time, during the target sensing interval, according to the distance sensing result, the robot also finds area units B, C, and D connected to the current set of historical area units. Therefore, the number of connected area units is 3. For the determination of the target sensing interval, there are two key factors: the number of identified area units and the number of connected area units. The robot adjusts the frequency of target sensing according to these factors.

[0078] If the robot observes an increase in the number of identified area units, the probability of discovering actual targets is low, and a strategy of reducing the frequency of target sensing can be adopted. For example, when the robot has identified multiple historical area units, continuing target sensing may not have much benefit, as the probability of detecting more repeated targets is high. (Negative feedback)

[0079] However, when the robot finds that the number of newly connected area units is large, it indicates that there may be more unexplored targets. In this case, the robot can adopt a strategy of increasing the frequency of target sensing. By shortening the target sensing interval time, the robot can discover potential targets faster and take appropriate action in a timely manner. (Positive feedback)

[0080] In the inspection task of the water cushion pond, the robot adjusts the interval time of target sensing flexibly according to the feedback relationship between the number of identified area units and the number of connected area units. This can optimize the efficiency of the inspection task and improve the target discovery rate, enabling the robot to better adapt to the differences between different area units and target distribution conditions.

[0081] In an embodiment of the present application, before the step of obtaining the water cushion pond level map to be inspected by performing the aforementioned iteration based on the historical water cushion pond level map, the following examples are provided.

[0082] (1) According to the historical inspection route, the underwater robot is guided to perform inspection on the historical water cushion pond level map;

[0083] (2) In the process of inspection on the historical water cushion pond level map, the environmental changes of the historical water cushion pond level map are identified;

[0084] (3) When it is identified that the historical water mat pond level map has environmental changes, and the underwater robot has not reached the inspection departure position, it is determined that the historical water mat pond level map has been adjusted; or, based on an active adjustment command, it is determined that the historical water mat pond level map has been adjusted.

[0085] In the implementation of the present application, it is assumed that the historical inspection route is in the order of A→B→C→D→E→I. The underwater robot first arrives at the area unit A, and then inspects the area units B, C, D, and E in turn along the planned path, and finally arrives at the inspection departure position I. The robot completes the inspection task along the path by adjusting its own movement according to the pre-set movement and navigation strategy. During the inspection process, the underwater robot uses sensors and vision systems to perceive changes in the surrounding environment. For example, the robot may detect changes in water quality, such as an increase in turbidity or temperature, through water quality sensors; or identify the appearance of new damage, cracks, or obstacles on the water mat pond mat surface through a camera. These identified environmental changes can be used as a basis for the robot to determine whether the historical water mat pond level map needs to be adjusted. It is assumed that during the inspection process, the underwater robot detects the appearance of a new area unit F, and the robot has not yet reached the pre-determined inspection departure position I. In this case, it can be determined that the historical water mat pond level map needs to be adjusted, and the newly added area unit F is added to the level map, and the connection relationship with other area units is updated. This can ensure that the robot can cover the newly added area unit F in the next inspection task. Another scenario is that the command center may send a command to the underwater robot to adjust the historical water mat pond level map according to specific requirements. For example, if a report about a safety problem in a certain area is received, the command center can issue a command to the robot to increase the coverage of the area in the next inspection task, and accordingly modify the historical water mat pond level map. This can ensure that the robot inspects according to the new instructions to meet the changing requirements of the task. Through the above steps, the underwater robot can be controlled according to the historical inspection route, identify environmental changes during the inspection process, and adjust the historical water mat pond level map. This enables the robot to adapt to changes in the water mat pond and flexibly respond to changes in task requirements.

[0086] In the embodiment of the present application, after the aforementioned step of manipulating the underwater robot to inspect the water mat pond level map to be inspected according to the target inspection route, the following implementation is provided.

[0087] (1) During the process of manipulating the underwater robot to inspect the water mat pond level map to be inspected according to the target inspection route, a new water mat pond level map is obtained based on the iteration of the water mat pond level map to be inspected, wherein the new water mat pond level map is the water mat pond level map to be inspected after iteration;

[0088] (2) In the process of operating the underwater robot to patrol the new water pad pond level map according to the target patrol line, the target patrol line is updated according to the bypass line of the identified impassable area, and a new patrol line is obtained;

[0089] (3) Operating the underwater robot to patrol the new water pad pond level map according to the new patrol line.

[0090] Assuming that the underwater robot patrols according to the preset target patrol line A→B→C→D→E and successfully completes the inspection of the water pad pond to be inspected. At this time, the robot has collected data and information about each area unit (pad surface) of the water pad pond. During the inspection, the underwater robot uses sensors and vision systems to obtain data about the pad surface state, water quality, temperature, etc. With these new data, iteration of the water pad pond level map to be inspected can be performed, i.e. updating the state or attributes of the pad surface. For example, if the robot detects that the pad surface of a certain area unit is severely worn, the area unit can be marked as needing repair or replacement in the new water pad pond level map. During the inspection of the new water pad pond level map, the underwater robot may encounter some impassable areas, such as obstacles or damaged pad surfaces. Through sensors and vision systems, the robot can detect these impassable areas and identify bypass lines to avoid these obstacles. Based on the information of the identified impassable areas, the target patrol line can be updated to generate a new patrol line, so that the robot can bypass the impassable areas and continue to patrol other area units. Based on the updated patrol line, the underwater robot patrols the new water pad pond level map according to the new instructions. For example, the robot may patrol the area units in the order A→C→D→B→E according to the new line. By patrolling the new water pad pond level map, the robot can collect updated data and information to provide more accurate references for maintenance, management or the next inspection task.

[0091] Through the above steps, the underwater robot can control itself to patrol on the to-be-patrolled water pad pond level map according to the target patrol route, and iteratively obtain a new water pad pond level map during the patrol. At the same time, according to the information of the identified impassable area, the target patrol route is updated, so that the robot can avoid obstacles and successfully complete the patrol task. For example, in an actual scenario, assuming that the to-be-patrolled water pad pond level map includes region units A, B, C and D. The initial target patrol route is A→B→C→D. During the patrol, the underwater robot finds that there is a damaged pad surface in region unit C, which causes it to be unable to pass through. The robot detects this situation through the sensor and plans a path to bypass region unit C. According to the bypass route of the identified impassable area, the robot can update the target patrol route. In this example, the updated patrol route may become A→B→D to avoid the damaged region unit C. Subsequently, the underwater robot manipulates itself to patrol on the new water pad pond level map according to the new patrol route, and checks the corresponding region units in the order of A→B→D.

[0092] In the embodiment of the present application, the foregoing step S205 can be implemented by the following manner.

[0093] (1) performing a feature learning operation on the patrol result to obtain a feature matrix of the patrol result, the feature matrix being used to indicate the density of the patrol video content at each region in the patrol result;

[0094] (2) determining a sample selection frequency of the corresponding region according to the density of the patrol video content at each region in the feature matrix;

[0095] (3) performing sample selection on the patrol result according to the sample selection frequency to obtain a plurality of first patrol video segments, the patrol video content being determined according to the image elements in the patrol result, and the density of the patrol video content and the sample selection frequency having a positive feedback relationship;

[0096] (4) for any first patrol video segment, performing a cutting operation on the first patrol video segment to obtain a plurality of second patrol video segments;

[0097] (5) obtaining feature vectors of the plurality of second patrol video segments, the feature vectors of the plurality of second patrol video segments being determined by merging the video attributes of the plurality of second patrol video segments and the spatial attributes of the plurality of second patrol video segments;

[0098] (6) standardizing the feature vectors of the plurality of second patrol video segments according to the first standardization module;

[0099] (7) processing the standardized feature vectors of the plurality of second patrol video segments according to the long short-term memory network configured with the preset static filter to obtain first transition feature vectors;

[0100] (8) determining a second transition feature vector according to the feature vectors of the plurality of second inspection video segments and the first transition feature vector;

[0101] (9) normalizing the second transition feature vector according to a second normalization module;

[0102] (10) processing the second transition feature vector according to a deep fully connected network to obtain intermediate inspection video vectors of the plurality of second inspection video segments;

[0103] (11) extracting features of the intermediate inspection video vectors of the plurality of second inspection video segments according to a long short-term memory network configured with a preset dynamic filter, a normalization module, and a deep fully connected network to obtain an inspection video vector of the first inspection video segment, and reconstructing features of the inspection video vector of the first inspection video segment to obtain abnormal levels of a plurality of image elements in the first inspection video segment for any first inspection video segment;

[0104] (12) generating an inspection video frame of the first inspection video segment according to the abnormal levels of the plurality of image elements in the first inspection video segment, wherein an element feature of each image element in the inspection video frame is used to represent a corresponding abnormal level, and each inspection video frame is configured with an abnormal level;

[0105] (13) performing an identification operation according to the inspection video frames of the plurality of first inspection video segments to obtain a danger level of each first inspection video segment;

[0106] (14) evaluating a danger level of an inspection result according to the danger level of each first inspection video segment to obtain an inspection analysis report of the inspection result.

[0107] In embodiments of the present application, computer vision techniques can be used to analyze the video data of the water cushion pond inspection. For example, image processing algorithms are used to detect defects, cracks, or corrosion problems, and extract features such as color, texture, shape, etc. of these areas. According to the feature matrix in the inspection result, the abnormal density of each area can be determined. For example, by calculating the average number of abnormal pixels or the average degree of abnormality in each area, the abnormal density (i.e. problem severity) of the area can be determined. A higher density indicates that the area may have more serious problems, so the area needs to be selected more frequently as a sample for further analysis and evaluation. Based on the sample selection frequency, multiple first inspection video segments are selected from the inspection results as samples. For example, assuming that 10 areas with high abnormal density are found in the inspection results, one first inspection video segment in each area can be selected as a sample, and a total of 10 first inspection video segments are selected. For each selected first inspection video segment, it is cut into multiple second inspection video segments. For example, if the length of the first inspection video segment is 10 minutes, it can be divided into 10 second inspection video segments each lasting 1 minute. In this way, each second inspection video segment can contain specific events or abnormal conditions. For each second inspection video segment, its feature vector is extracted. For example, for image features, color histogram algorithm can be used to extract color distribution information; for texture features, local binary pattern (LBP) can be used to extract texture information; for shape features, edge detection algorithm can be used to extract edge information. In this way, the feature information of each second inspection video segment can be captured from different angles. The feature vectors extracted from multiple second inspection video segments are standardized. This can include normalizing the feature vectors to similar scales and distributions to ensure their comparability and reduce the bias between different features. For example, Z-score standardization can be used to convert the feature vectors to values with zero mean and unit variance. The standardized feature vectors are processed using a pre-configured long short-term memory (LSTM) network, a standardization module, and a deep fully connected (DNN) network. For example, the LSTM network can capture long-term dependencies in the feature sequence and generate a first transition feature vector for further feature extraction and analysis. For each first inspection video segment, a feature reconstruction operation is performed to restore the feature representation of the original image. Self-encoder or other reconstruction models can be used to achieve this. By comparing with the original image, the abnormal level of each image element is calculated. For example, by calculating the pixel-level error or reconstruction loss, it can be determined whether the element is abnormal and assign it a corresponding abnormal level. Based on the abnormal level result of the first inspection video segment, an inspection video frame is generated. In these frames, each image element is labeled with a corresponding abnormal level to reflect its abnormality degree or classification. The abnormal level can be added to the inspection video frame in the form of color mapping or symbolic labeling, so that the abnormality of different positions in the pipeline can be visualized.The generated inspection video frames are subjected to recognition operations using appropriate algorithms or models. By analyzing the abnormality levels and distribution of image elements in each first inspection video segment, the corresponding risk level can be determined. For example, the inspection results are classified into different risk levels, such as low, medium, high, or the risk level is determined by comparing with pre-defined thresholds, according to the number, density and distribution of abnormality levels. According to the risk level of each first inspection video segment, the rating results of all inspection video segments are considered comprehensively to evaluate the risk level of the overall inspection result. The inspection results of different levels can be weighted or aggregated to determine the overall risk level of the inspection result, and a corresponding inspection analysis report is generated. In addition, a detailed inspection report can also be generated based on the risk level evaluation and analysis of the inspection results. The report should include the abnormal conditions, risk levels and possible cause analysis of each area. In addition, a recommended maintenance plan can also be provided, including repair measures, preventive measures and optimization schemes, to reduce the occurrence of potential problems in the pipeline. In order to ensure the safety and reliability of the water cushion pond, repeated inspection and monitoring need to be carried out regularly. This can be achieved by setting up an inspection plan and re-inspecting the pipeline at certain time intervals. The new inspection data will be compared with the historical data to detect any new abnormalities or changes and take appropriate measures to handle them. By continuously collecting and analyzing inspection data, monitoring data and implementing maintenance plans, potential problems and improvement opportunities in the system can be identified. These feedback information can be used to improve the inspection process, optimize algorithms and models, and improve the accuracy and efficiency of water cushion pond inspection.

[0108] In the embodiments of the present application, the step of performing recognition operations on the inspection video frames of the plurality of first inspection video segments to obtain the risk level of each first inspection video segment can be implemented by the following examples.

[0109] (1) For the inspection video frames of any first inspection video segment, the number of image elements corresponding to each abnormality level in the inspection video frames of the first inspection video segment is obtained;

[0110] (2) According to the number of image elements corresponding to each abnormality level, the risk level of the first inspection video segment is determined;

[0111] The step of performing recognition operations on the inspection video frames of the plurality of first inspection video segments to obtain the risk level of each first inspection video segment can also be implemented by the following examples.

[0112] (1) For the inspection video frames of any first inspection video segment, the abnormality levels corresponding to the plurality of image elements in the inspection video frames are corrected according to the relative layout between the plurality of image elements in the inspection video frames;

[0113] (2) According to the number of image elements corresponding to each abnormality level in the modified inspection video frame, the danger level of the first inspection video segment is obtained.

[0114] In an embodiment of the present application, it is assumed that the inspection video frame of a certain first inspection video segment contains image elements of three abnormality levels: high, medium and low. By analyzing the image elements corresponding to each abnormality level in the inspection video frame, the specific number can be obtained. For example, there are 30 image elements of high-level abnormality, 50 image elements of medium-level abnormality, and 100 image elements of low-level abnormality. Based on the above scenario, according to the number of image elements of different abnormality levels, the danger level of the first inspection video segment can be calculated. For example, if it is stipulated that more than 50 image elements of high-level abnormality are determined as high danger level, image elements of medium-level abnormality between 20-50 are determined as medium danger level, and less than 20 image elements of low-level abnormality are determined as low danger level. According to these rules, the danger level of the first inspection video segment can be determined as medium danger level.

[0115] In another embodiment of the present application, in the inspection video frame of a certain first inspection video segment, there is a concentrated distribution of abnormal areas, some of which are misclassified as low-level abnormality. By analyzing the relative layout between multiple image elements in the inspection video frame, this concentrated distribution of abnormal areas can be detected, and the abnormality level of the related image elements is corrected to improve its level. For example, the image elements originally misclassified as low-level may be reclassified as medium-level abnormality after correction. Based on the above scenario, after the inspection video frame of the first inspection video segment is modified, the number of image elements corresponding to each abnormality level is recalculated. For example, in the modified inspection video frame, there are 25 image elements of high-level abnormality, 40 image elements of medium-level abnormality, and 90 image elements of low-level abnormality. According to the new number, the danger level of the first inspection video segment can be re-determined. If the previous determination rule is followed, the danger level of the first inspection video segment may be corrected to low danger level.

[0116] An embodiment of the present application provides a computer device 100, which comprises a processor and a non-volatile memory storing computer instructions. When the computer instructions are executed by the processor, the computer device 100 executes the water cushion pond inspection method described above. As shown in FIG. 1, the computer device 100 comprises a processor 101 and a non-volatile memory 102. The non-volatile memory 102 stores computer instructions. When the computer instructions are executed by the processor 101, the computer device 100 executes the water cushion pond inspection method described above. Figure 2 Figure 2 ​A structural block diagram of the computer device 100 is provided for the embodiments of the present application. The computer device 100 comprises a memory 111, a processor 112 and a communication unit 113. In order to realize the transmission or interaction of data, the memory 111, the processor 112 and the communication unit 113 are electrically connected with each other directly or indirectly. For example, the electrical connection between these elements can be realized by one or more communication buses or signal lines.

[0117] The above merely provides the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement and improvement made within the spirit and principle of the present application shall fall within the scope of the present application.

Claims

1. A method for inspecting a water cushion basin, characterized by, The method comprises: in response to a patrol task starting instruction for a target water pad pond, establishing a communication connection with an underwater robot; obtain the water pad pond map information corresponding to the target water pad pond, the water pad pond map information at least includes a plurality of depth history water pad level map; according to the target depth corresponding to the history water pad level map, determine the target patrol route and control the underwater robot to patrol, the target depth is any depth in the plurality of depths; receiving the patrol results of the underwater robot for the plurality of depths; analyze the patrol results to obtain the patrol analysis report of the patrol results; the according to the target depth corresponding to the history water pad level map, determine the target patrol route and control the underwater robot to patrol, comprising: based on the iteration of the history water pad level map, obtain the water pad level map to be patrolled, wherein the water pad level map to be patrolled is the history water pad level map after iteration, and the patrol line of the history water pad level map is the history patrol line; in the process of controlling the underwater robot to patrol on the water pad level map to be patrolled according to the history patrol line, the unpassable area is identified to obtain the target unpassable area; determine the unpassable range to be determined according to the detour starting position; determine the area unit detour cost of each current area unit in the unpassable range to be determined according to at least one of the first spatial span and the second spatial span, wherein the first spatial span is the spatial span between the current area unit and the detour starting position, and the second spatial span is the spatial span between the current area unit and the target unpassable area; in the unpassable range to be determined, obtain a current area unit group on each of a plurality of candidate patrol strategies; determine the area unit detour cost array corresponding to the current area unit group according to the area unit detour cost; data regularization is carried out on the area unit detour cost array to obtain the detour cost corresponding to each of the candidate patrol strategies; select at least one detour strategy from a plurality of the candidate patrol strategies according to the detour cost; determine the current candidate range in the unpassable range to be determined according to at least one of the detour strategies; in the current candidate range, traverse a group of passable area units adjacent to the current starting position area unit, wherein the current starting position area unit is the current area unit corresponding to the detour starting position; obtain the patrol cost corresponding to each of the current area units in the group of passable area units, and determine the current area unit corresponding to the latest traversal of the patrol cost as the last level area unit, wherein the patrol cost is the cost from the current starting position area unit to the current leaving position area unit through the corresponding current area unit; traverse a next set of passable region cells adjacent to a target current region cell until a current departure location region cell is included in the set of passable region cells, terminate the traversal, and backtrack from the current departure location region cell to a current starting location region cell to obtain a target detour route between a detour starting location and a detour departure location, wherein the target current region cell is a current region cell that has not been traversed and has an optimal patrol cost, the target detour route is used to bypass a target impassable region from the detour starting location to the detour departure location, the detour starting location is a detour start on the historical patrol route, and the detour departure location is a detour end on the historical patrol route; adjust the historical patrol route according to the target detour route to obtain a target patrol route; and maneuver the underwater robot to patrol on the to-be-patrolled water sump hierarchical graph according to the target patrol route.

2. The method of claim 1, wherein, When the at least one detour strategy is a plurality of detour strategies, the determining the current candidate range in the to-be-determined impassable range according to the at least one detour strategy includes: obtaining a plurality of local candidate ranges corresponding to the plurality of detour strategies, and combining the plurality of local candidate ranges to form the current candidate range in the to-be-determined impassable range; or determining a target detour strategy from the plurality of detour strategies according to a current departure location region cell, and determining a candidate range on the target detour strategy as the current candidate range in the to-be-determined impassable range, wherein the current departure location region cell is a current region cell corresponding to the detour departure location.

3. The method of claim 1, wherein, Before the obtaining the to-be-patrolled water sump hierarchical graph based on the iteration of the historical water sump hierarchical graph, the method further includes: segmenting the historical water sump hierarchical graph into region cells to obtain a region cell set, wherein the region cell set includes a plurality of historical region cells; taking a historical starting location region cell as a starting point and determining a plurality of sensing paths corresponding to a plurality of preset deflection amounts according to a preset sensing interval in the region cell set; extracting a reliable sensing path array from the plurality of sensing paths according to a preset rule, wherein the preset rule is determined according to at least one of a sensing distance and a preset command of a target underwater robot; obtaining a preset number of reliable sensing paths adjacent in a preset deflection amount order from the reliable sensing path array, and obtaining sensing path ranges corresponding to the preset number of reliable sensing paths; combining the historical region cells in the sensing path ranges greater than a preset range to form a first historical region cell set connected to the historical starting location region cell; when the first historical region cell set does not include a historical departure location region cell, continuously performing the following processing: The distance sensing of the target sensing interval is performed from the current historical region unit group as a starting point, and a next historical region unit group connected with the current historical region unit group is obtained, wherein the target sensing interval has a positive feedback relationship with the connection sensing rate of the distance sensing; When the obtained next historical region unit group includes the historical departure location region unit, the distance sensing is ended, and the historical starting location region unit and all historical region unit groups are jointly formed into a preset number of connected historical region units; The preset number of historical region units is determined as a historical candidate range; The historical starting location region unit is the historical region unit corresponding to the starting location of the historical sump hierarchical graph, the historical departure location region unit is the historical region unit corresponding to the departure location of the historical sump hierarchical graph, and the preset number of historical region units at least includes the historical starting location region unit and the historical departure location region unit.

4. The method of claim 3, wherein, Before the distance sensing of the target sensing interval is performed from the current historical region unit group as a starting point to obtain a next historical region unit group connected with the current historical region unit group, the method further includes: acquiring a number of identified region units and a number of connected region units, wherein the number of identified region units is the number of historical region units identified before the distance sensing of the current round, and the number of connected region units is the number of connected historical region units identified before the distance sensing of the current round; acquiring the target sensing interval having a negative feedback relationship with the number of identified region units and a positive feedback relationship with the number of connected region units.

5. The method of claim 1, wherein, Before the iteration based on the historical sump hierarchical graph is performed to obtain the sump hierarchical graph to be inspected, the method further includes: controlling the underwater robot to perform inspection on the historical sump hierarchical graph according to the historical inspection route; identifying an environmental change of the historical sump hierarchical graph during the inspection of the historical sump hierarchical graph by the underwater robot; determining that the historical sump hierarchical graph is adjusted when the environmental change of the historical sump hierarchical graph is identified and the underwater robot does not reach the departure location for inspection; or determining that the historical sump hierarchical graph is adjusted based on an active adjustment command.

6. The method of claim 1, wherein, After the underwater robot is controlled to perform inspection on the sump hierarchical graph to be inspected according to the target inspection route, the method further includes: during the inspection of the sump hierarchical graph to be inspected by the underwater robot according to the target inspection route, a new sump hierarchical graph is obtained based on the iteration of the sump hierarchical graph to be inspected, wherein the new sump hierarchical graph is the sump hierarchical graph to be inspected after iteration; In the process of operating the underwater robot to patrol the new water mat pond level map according to the target patrol line, the target patrol line is updated according to the bypass line of the identified impassable area, and a new patrol line is obtained; According to the new patrol line, the underwater robot is operated to patrol the new water mat pond level map.

7. The method of claim 1, wherein, The analysis of the patrol result includes: Performing feature learning operations on the patrol result to obtain a feature matrix of the patrol result, the feature matrix being used to indicate the density of patrol video content at each region in the patrol result; According to the density of patrol video content at each region in the feature matrix, determine the sample selection frequency of the corresponding region; According to the sample selection frequency, perform sample selection on the patrol result to obtain a plurality of first patrol video segments, the patrol video content being determined according to image elements in the patrol result, and the density of the patrol video content and the sample selection frequency having a positive feedback relationship; For any first patrol video segment, perform cutting operations on the first patrol video segment to obtain a plurality of second patrol video segments; Obtain feature vectors of the plurality of second patrol video segments, the feature vectors of the plurality of second patrol video segments being determined by merging video attributes of the plurality of second patrol video segments and spatial attributes of the plurality of second patrol video segments; According to a first standardization module, standardize the feature vectors of the plurality of second patrol video segments; According to a long short-term memory network configured with a preset static filter, process the standardized feature vectors of the plurality of second patrol video segments to obtain a first transition feature vector; According to the feature vectors of the plurality of second patrol video segments and the first transition feature vector, determine a second transition feature vector; According to a second standardization module, standardize the second transition feature vector; According to a deep fully connected network, process the second transition feature vector to obtain an intermediate patrol video vector of the plurality of second patrol video segments; According to a long short-term memory network configured with a preset dynamic filter, a standardization module, and a deep fully connected network, perform feature extraction on the intermediate patrol video vector of the plurality of second patrol video segments to obtain a patrol video vector of the first patrol video segment, and for any first patrol video segment, perform feature reconstruction on the patrol video vector of the first patrol video segment to obtain abnormal levels corresponding to a plurality of image elements in the first patrol video segment; According to the abnormal levels corresponding to the plurality of image elements in the first patrol video segment, generate patrol video frames of the first patrol video segment, element features of each image element in the patrol video frames being used to represent the corresponding abnormal level, and each patrol video frame being configured with an abnormal level; According to the patrol video frames of the plurality of first patrol video segments, perform identification operations to obtain a danger level of each first patrol video segment; According to the danger level of each first patrol video segment, evaluate a danger level of the patrol result to obtain a patrol analysis report of the patrol result.

8. The method of claim 7, wherein, The identification operation is performed according to the inspection video frames of the plurality of first inspection video segments, and a danger level of each of the first inspection video segments is obtained, including: For the inspection video frame of any first inspection video segment, the number of image elements corresponding to each abnormality level in the inspection video frame of the first inspection video segment is obtained; According to the number of image elements corresponding to each abnormality level, the danger level of the first inspection video segment is determined; The identification operation is performed according to the inspection video frames of the plurality of first inspection video segments, and a danger level of each of the first inspection video segments is obtained, further including: For the inspection video frame of any first inspection video segment, according to the relative layout between a plurality of image elements in the inspection video frame, the abnormality levels corresponding to the plurality of image elements in the inspection video frame are corrected; According to the number of image elements corresponding to each abnormality level in the corrected inspection video frame, the danger level of the first inspection video segment is obtained.

9. A server system, characterized by The server is configured to execute the method of any one of claims 1-8.

Images