An unmanned aerial vehicle re-provisioning decision method and system based on dynamic revisit

Abstract

The application discloses a kind of unmanned aerial vehicle based on dynamic Revisit's supplementary investment decision-making method and system, this method first obtains the overhead image of offshore net cage culture field, constructs historical space topology and current local topology, determines the tracking ID of net cage to be operated.Second, according to the feeding time and target waiting interval of each net cage, a candidate set of revisit is established, and the revisit order is determined.Then the unmanned aerial vehicle reaches the candidate net cage and reconfirms the tracking ID of net cage, judges whether to execute supplementary investment, and updates the next target waiting interval of the corresponding net cage.Finally, the number of effective supplementary investment requirements generated by each net cage within the effective operation day is counted, the rolling feeding state index is calculated, the feeding state grade and abnormal prompt are generated.The application realizes the stable tracking of the same net cage across batches, solves the ID mismatch caused by positioning deviation, net cage drift and other problems, and solves the problems of insufficient feeding and bait waste caused by fixed cycle supplementary investment.

Description

Technical Field

[0001] This invention belongs to the field of intelligent operation technology of drones for marine cage aquaculture, and particularly relates to a drone refeeding decision-making method and system based on dynamic Revisit, which is based on cage topology ID matching, residual bait image detection and fish health value inference. Background Technology

[0002] Offshore cage aquaculture is typically characterized by a large number of cages, a wide distribution range, and an operating environment significantly affected by wind, waves, and currents. Feeding requires monitoring multiple cages across the entire area and adjusting the feeding amount promptly based on the fish's feeding behavior. Traditional manual feeding or fixed-route feeding methods struggle to consistently obtain accurate feeding feedback from each cage, often leading to feeding imbalances: some units receive insufficient feed, while others accumulate uneaten feed, posing a risk to the aquaculture industry.

[0003] In existing drone inspection or feeding solutions, feeding, inspection, and health monitoring are typically treated as relatively independent tasks. Because net cages look similar and are prone to drifting in waves, relying solely on absolute coordinates, flight path sequence, or single-frame visual appearance for cage identification after multiple drone returns to the work area can easily lead to inconsistent tracking IDs for the same physical cage across different batches. This unstable cage ID further complicates the accurate matching of feeding records, uneaten bait observation records, and health status records.

[0004] Furthermore, existing methods for observing uneaten feed after feeding often rely on fixed-period reviews or manual judgment, lacking quantitative expressions for uneaten feed area, uneaten feed density, fish feeding activity, and health values. They also lack the calculation relationship to directly update the re-feeding amount and next round of feeding parameters based on these monitoring indicators. Therefore, a UAV-based revisit and re-feeding decision-making method is needed that clearly defines the image observation object, monitoring indicators, threshold determination, and control update formula. Summary of the Invention

[0005] The present invention aims to solve the aforementioned technical problems existing in the prior art by providing a method and system for drone re-deployment decision-making based on dynamic revisit.

[0006] In one aspect, this invention provides a UAV re-deployment decision-making method based on dynamic revisit, comprising the following steps: Step 1: Obtain an overhead view image of the offshore cage aquaculture farm, perform cage target detection on the overhead view image, and construct a historical spatial topology map with each cage as a node and the relative spatial relationship between cages as edges.

[0007] Step 2: The drone flies back to the area where the cage to be matched is located and constructs the current local topology map with the cage to be matched and a preset number of neighboring cages; the tracking ID of the cage to be matched is determined based on the topological similarity between the current local topology map and the historical spatial topology map.

[0008] Step 3: Establish a candidate set for revisiting based on the feeding time and target waiting interval of each cage, and determine the revisit order according to the duration exceeding the planned revisit time.

[0009] Step 4: After the drone arrives at the candidate net cage, it reconfirms the net cage tracking ID, collects images of residual bait on the water surface of the net cage, obtains the estimated amount of residual bait through the residual bait detection model, and determines whether to perform supplementary feeding based on the proportion of residual bait.

[0010] Step 5: Update the next target waiting interval for the corresponding net cage based on the actual return visit time, the proportion of uneaten bait, and the proportion of target uneaten bait, so that the net cage return visit frequency changes dynamically with the feeding status of the fish.

[0011] Step 6: Count the number of valid refeeding requests generated by each cage during the effective working day, and calculate the rolling feeding status index based on the number of refeeding requests of the cage group in the same period, and generate the corresponding feeding status level and abnormal prompts.

[0012] Step 7: Repeat the following steps: net cage tracking ID matching, fixed amount feeding, waiting interval timing, return visit for residual bait detection, refeeding judgment, waiting interval update, and feeding status update to form a closed loop of drone autonomous feeding and dynamic return visit.

[0013] In another aspect, the present invention also provides a UAV re-deployment decision-making system based on dynamic revisit, comprising the following modules: The historical spatial topology map module is used to acquire top-down images of offshore cage aquaculture farms, perform cage target detection, and construct a historical spatial topology map.

[0014] The cage tracking ID module is used to fly the drone to the area where the cage to be operated is located, construct the current local topology map, and determine the tracking ID of the cage to be operated based on the topological similarity between the current local topology map and the historical spatial topology map.

[0015] The follow-up module is used to establish a follow-up candidate set based on the feeding time and target waiting interval of each cage, and to determine the follow-up order according to the duration of time exceeding the planned follow-up time.

[0016] The refeeding module is used to reconfirm the net cage tracking ID after the drone arrives at the candidate net cage, collect images of residual bait on the water surface of the net cage, and determine whether to perform refeeding.

[0017] The dynamic update module is used to update the next target waiting interval of the corresponding net cage based on the actual revisit time, the proportion of uneaten bait, and the proportion of target uneaten bait, so that the net cage revisit frequency changes dynamically with the feeding status of the fish.

[0018] The rolling feeding status index module is used to count the number of valid refeeding needs generated by each cage within a valid working day, and to calculate the rolling feeding status index based on the number of refeeding needs of the cage group during the same period, generating feeding status levels and abnormal prompts.

[0019] Compared with the prior art, the present invention has the following beneficial effects: 1. By using cage target detection and spatial topology matching technology, stable cross-batch tracking of the same cage can be achieved, solving the ID mismatch problem caused by positioning deviation and cage drift.

[0020] 2. By using fixed-amount feeding and residual feed ratio thresholds, the need for supplementary feeding can be quantitatively determined, thus solving the problems of insufficient feeding and feed waste caused by fixed-period supplementary feeding.

[0021] 3. By recording the actual time of the follow-up visit and updating the next target waiting interval based on the actual observation duration, the proportion of uneaten bait, and the proportion of uneaten bait in the target, the frequency of follow-up visits can be dynamically changed according to the actual feeding status of the fish.

[0022] 4. By incorporating delays such as return refueling and battery swapping into the actual observation time, rather than treating them as independent correction factors for fish feeding status, the operational scheduling factors and fish feeding factors are distinguished, thereby improving the rationality of the waiting interval update results.

[0023] 5. By referencing the median number of effective refeeding needs and performing rolling calculations, a stable assessment of the feeding health status of fish populations can be achieved.

[0024] 6. By linking feeding, follow-up visits, and health evaluation records with the cage tracking ID, a dynamic follow-up closed loop is formed, solving the problems of independent operations and difficulty in accurately matching data. Attached Figure Description

[0025] Figure 1 A flowchart of the dynamic revisit process for a net cage; Figure 2 This is a schematic diagram of the topology ID matching process; Figure 3 This is a diagram illustrating topology ID matching. Figure 4 This is a diagram comparing fixed-cycle and dynamic revisit paths; Figure 5 A comparison chart of feed state curves between fixed-period feeding and dynamic revisit feeding; Figure 6This is a land-based test image of a drone. Figure 7 This is a picture of a drone undergoing maritime testing. Detailed Implementation

[0026] In one aspect, this invention provides a UAV re-deployment decision-making method based on dynamic revisit, such as... Figure 1 and Figure 2 As shown, the implementation process includes the following: a. Constructing a historical spatial topology map: Since the appearance and size of net cages within the same aquaculture area are usually quite similar, it is difficult to form stable identification characteristics based solely on the visual appearance of a single net cage. Furthermore, when the net cage group drifts or the drone's positioning is inaccurate, the relative distance, relative orientation, and adjacency relationship between net cages are more stable than their absolute position. Therefore, a historical spatial topology map is constructed. A drone carrying a camera flies directly below the net cages to obtain an overhead view image (Image I) of the offshore net cage aquaculture farm. The YOLO object detection model is used to detect the net cage bounding boxes in Image I, outputting the pixel coordinates of the net cage center point. The network of offshore cages in the management area is abstracted into a topological graph structure, and long-term identity association of cages is achieved through stable spatial neighborhood relationships. An undirected spatial topological graph model is constructed: The center point of the cage is considered as the set of vertices. For each independent aquaculture cage within the corresponding sea area, each graph vertex uniquely corresponds to a physical cage; edge set Used to characterize the inherent spatial constraint relationships between cages, defined For net cages With net cage The relative spatial topology relationships between them include stable spatial features such as relative distance, relative orientation, and local neighborhood arrangement structure. The historical spatial topology map can provide a relatively stable spatial reference for matching cage tracking IDs in different operation batches.

[0027] b. Cage ID Matching: The drone, equipped with a camera, is guided by GPS to fly over the cages again, acquiring an overhead image of the offshore cage aquaculture farm, using the ID of the cage as the reference. Centered on a single net cage, select a fixed number of Find the nearest neighbor boxes and construct the current local topology graph. The graphs are then sorted uniformly according to normalized distance and relative orientation; matrices are generated for the current and historical graphs using the same rules. The current local topology graph is matched with the historical topology graphs, and the topological similarity is calculated as follows: For any two adjacent net cages The pixel coordinates of the center point of the detection box are respectively Calculate the pixel distance to the center point and the relative azimuth angle: To reduce the impact of flight altitude and image scale variations, the median pixel distance between adjacent edges in the local topology map is used as the reference scale. Calculate the normalized relative distance: The normalized distance matrix of the current cage is composed of normalized relative distances. The orientation matrix of the current cage is composed of relative azimuth angles. The adjacency matrix of the current net cage is constructed based on whether there are adjacency relationships between net cages. .

[0028] The current local topology graph is matched with the local topology graph corresponding to the historical spatial topology graph m, and the topological similarity is calculated as follows: in, , and These are the normalized distance matrix, orientation matrix, and adjacency matrix corresponding to the historical spatial topology graph, respectively. This is an adjacency matching function; Weights and satisfying ; It is the Frobenius norm.

[0029] For ID matching threshold, when When the value is the maximum value in the entire historical spatial topology graph, the tracking ID of historical spatial topology graph m is assigned to the current cage. Otherwise, pause automatic feeding.

[0030] c. Initial feeding and determination of the initial target waiting interval: After the cage tracking ID is successfully matched, the drone moves to the first... A fixed number of net cages were deployed. The feed was used, and this feeding was recorded as the first feeding round, with the time of the first feeding also recorded. .

[0031] During the initial system operation phase, historical uneaten feed observation data for the target fish population was not yet available. Therefore, the initial follow-up visit time could not be directly calculated based on the actual feeding rate. Thus, before obtaining historical uneaten feed observation data for the net cage, the time was determined based on the current season. Based on breeding experience, the initial waiting interval for the season is set manually. The planned return visit time for the cage is determined by the initial feeding time and the initial target waiting interval: d. Establish a candidate set for revisits and determine the revisit order: Calculate the planned revisit time based on the feeding time of the i-th cage in the k-th round and the target waiting interval: When the current time t reaches or exceeds the planned revisit time, the net cage is added to the revisit candidate set: in, Complete the marking for the revisit of the i-th cage in the k-th round; This indicates that the revisit has not yet been completed; once the corresponding revisit is completed, [the process will proceed]. The order of follow-up visits is determined by the longest timeout period, with priority given to the cages that have exceeded the planned follow-up time for the longest time.

[0032] e. Perform Revisit, Re-baiting Judgment, and Waiting Interval Update: After the UAV reaches the candidate cage, it reconfirms the cage tracking ID through local topology matching and acquires images of residual bait on the water surface of the cage. Record the actual Revisit time. Calculate the actual observation duration for this round: The amount of residual bait is estimated using a residual bait detection model. And calculate the percentage of uneaten bait: Let the target residual bait ratio for allowing supplementary baiting be . This invention uses the same fixed feeding amount for each net cage. Using this fixed feeding amount as a normalization benchmark, the estimated amount of residual feed was converted into a dimensionless residual feed ratio. Therefore, the results of residual feed detection in different net cages and different feeding cycles can be compared on the same scale, and a uniform target residual feed ratio can be adopted. Determine if the conditions for supplementary payment are met. When the following conditions are met: At that time, it was determined that the cage required replenishment. The drone then deployed a fixed quantity of supplies to the cage. The bait was used, and this refeeding was recorded as the first time. Feeding in rotation, recording feeding times. .

[0033] Based on the actual observation time and the proportion of uneaten bait in this round, it is estimated that the proportion of uneaten bait has decreased to Time required: in, To estimate the feeding time, the target waiting interval for the i-th cage in the next round is: = Therefore, the target waiting interval can be dynamically shortened according to the actual rate of bait consumption, while remaining within the manually set seasonal return visit baseline. The next planned return visit time is then calculated. When satisfied When necessary, make up for the shortfall; If no supplementary application is performed, the waiting interval for the next Revisit will still be [time period missing]. .

[0034] f. Calculate the rolling feeding health index: During the preset daily operating period, whenever the detected residual feed ratio meets the following criteria: Record each valid re-delivery request, and use this record to document the first re-delivery request. The first net cage in the... Total number of re-delivery requests within a workday ; and then for the first The median number of times each cage requires resupply on a given workday is taken to obtain the reference resupply demand for that day. .

[0035] For the Each net cage selects the most recent preset number of valid work days to form a work day set. Calculate the rolling dietary health index: Based on the rolling feeding status index This generates corresponding feeding status levels and abnormal alerts.

[0036] Classifying feeding health levels and generating abnormal alerts: Based on the 7-day rolling feeding health index, the feeding status of the fish in the net cages is classified into the levels shown in Table 1 below: Table 1

[0037] g. Continuously update the dietary health index: At the end of each new workday, record the number of times the dietary needs are met. and reference number of supplementary investment needs Add the net cage to the rolling record and delete the corresponding record of the earliest working day, then recalculate the 7-day rolling dietary health index.

[0038] The system only needs to save: i. Forming a dynamic Revisit closed loop: By repeatedly executing cage ID matching, fixed-amount feeding, waiting interval timing, revisit residual bait detection, refeeding judgment, next round waiting interval update, and feeding health index update, a dynamic Revisit closed loop based on cage tracking ID is formed.

[0039] In another aspect, the present invention also provides a UAV re-deployment decision-making system based on dynamic revisit, used to implement the aforementioned UAV re-deployment decision-making method, including a historical spatial topology map module, a cage tracking ID module, a revisit module, a re-deployment module, a dynamic update module, and a rolling feeding status index module. The historical spatial topology map module is used to acquire top-down images of offshore cage aquaculture farms, perform cage target detection, and construct a historical spatial topology map.

[0040] The cage tracking ID module is used to fly the drone to the area where the cage to be operated is located, construct the current local topology map, and determine the tracking ID of the cage to be operated based on the topological similarity between the current local topology map and the historical spatial topology map.

[0041] The follow-up module is used to establish a follow-up candidate set based on the feeding time and target waiting interval of each cage, and to determine the follow-up order according to the duration of time exceeding the planned follow-up time.

[0042] The refeeding module is used to reconfirm the net cage tracking ID after the drone arrives at the candidate net cage, collect images of residual bait on the water surface of the net cage, and determine whether to perform refeeding.

[0043] The dynamic update module is used to update the next target waiting interval of the corresponding net cage based on the actual revisit time, the proportion of uneaten bait, and the proportion of target uneaten bait, so that the net cage revisit frequency changes dynamically with the feeding status of the fish.

[0044] The rolling feeding status index module is used to count the number of valid refeeding needs generated by each cage within a valid working day, and to calculate the rolling feeding status index based on the number of refeeding needs of the cage group during the same period, generating feeding status levels and abnormal prompts.

[0045] Because net cages on the sea surface are easily affected by wind and waves, when a GPS-guided drone reaches above the target net cage, multiple net cages may appear in the field of view, and GPS guidance alone cannot accurately locate the target net cage. Figure 2 This is a schematic diagram of the topology ID matching process. The 01 netbox represents the historical target netbox topology. When three netboxes (a, b, and c) appear within the viewpoint, taking netbox a as an example... Figure 3 This diagram illustrates the topology ID matching process. The topological similarity between the three local cage topologies and the historical target cage topology is calculated using relative distance, relative orientation, and adjacency relationships. .when At that time, the cage in question is the target cage.

[0046] Figure 4 This diagram illustrates the comparison between fixed-cycle and dynamic revisit paths. In fixed-cycle feeding, the feeding path remains unchanged, and the drone sequentially visits and feeds all fish cages after each waiting interval. In contrast, dynamic revisit feeding prioritizes which fish cage to visit based on time priority, feeding the cages that arrive at the waiting interval first. The waiting interval changes according to the feeding status of the fish.

[0047] Figure 5 Comparison of feed status curves between fixed-cycle feeding and dynamic revisit feeding: Taking a fixed-cycle feeding interval of 60 minutes as an example, and dynamic revisit feeding with an initial interval of 60 minutes and a target uneaten feed ratio of 0.25 as an example, it can be seen that with fixed-cycle feeding, fish sometimes experience hunger or feed waste; under dynamic revisit feeding, the revisit interval changes with the feeding status of the fish, greatly reducing feed waste and fish hunger.

[0048] To verify the basic feasibility of the method of the present invention, actual tests were conducted. Figure 6 It is a land-based test map, using standardized markers to replace the marine cages; drones equipped with visual inspection and quantitative feeding devices complete the entire process of take-off and landing, image acquisition, quantitative feeding, and dynamic follow-up review. Figure 7 This is a picture of a drone test at sea, showing the fish cages on the sea surface. The experiment involved multiple rounds of closed-loop operations, recording the accuracy of cage identification and matching, and the execution of restocking, thus preliminarily verifying the practicality and feasibility of the dynamic revisit drone restocking decision-making method and system in a simulated aquaculture scenario.

Claims

1. A UAV re-deployment decision-making method based on dynamic revisit, characterized in that, Includes the following steps: Step 1: Obtain an overhead view image of the offshore cage aquaculture farm, perform cage target detection, and construct a historical spatial topology map; Step 2: The drone flies to the area where the cage to be operated is located, constructs the current local topology map, and determines the tracking ID of the cage to be operated based on the topological similarity between the current local topology map and the historical spatial topology map; Step 3: Establish a candidate set for follow-up visits based on the feeding time and target waiting interval of each cage, and determine the follow-up visit order according to the duration exceeding the planned follow-up visit time; Step 4: After the drone arrives at the candidate net cage, it reconfirms the net cage tracking ID, collects images of residual bait on the water surface of the net cage, and determines whether to perform refeeding. Step 5: Update the next target waiting interval for the corresponding net cage based on the actual return visit time, the proportion of uneaten bait, and the proportion of target uneaten bait, so that the net cage return visit frequency changes dynamically with the feeding status of the fish. Step 6: Count the number of valid refeeding requests generated by each cage during the effective working day, and calculate the rolling feeding status index based on the number of refeeding requests of the cage group in the same period, and generate feeding status level and abnormal prompts.

2. The UAV re-deployment decision-making method based on dynamic revisit according to claim 1, characterized in that, In step S1, the historical spatial topology map is represented as follows: ,in A set of network cage nodes, where each network cage node uniquely corresponds to a physical network cage; This is a set of edges used to represent the relative distances, relative orientations, and local neighborhood arrangements between cages; The target detection of the cages is completed using a target detection model, which outputs the detection frame of each cage and the pixel coordinates of its center point.

3. The UAV re-deployment decision-making method based on dynamic revisit according to claim 2, characterized in that, The specific implementation process of step 2 is as follows: The drone flies to the area where the cages to be matched are located, and selects a fixed number of cages as the center. The current local topology is constructed using the nearest neighbor net boxes, and sorted uniformly according to normalized distance and relative orientation; the current local topology uses the same neighborhood selection rules and node sorting rules as the historical spatial topology. For any two adjacent net cages The pixel coordinates of the center point of the detection box are as follows: ; Calculate the pixel distance to the center point Calculate the relative azimuth angle ; The median pixel distance between adjacent edges in the current local topology graph is used as the reference scale. For the center point pixel distance Calculate normalized relative distance ; The normalized distance matrix of the current cage is composed of normalized relative distances. The orientation matrix of the current cage is composed of relative azimuth angles. The adjacency matrix of the current net cage is constructed based on whether there are adjacency relationships between net cages. Current net cage With historical candidate net cages The topological similarity between them is: in, , and These are the normalized distance matrix, orientation matrix, and adjacency matrix corresponding to the historical candidate cages, respectively; This is an adjacency matching function; As weight, Represents norm operations; For ID matching threshold, when ,and When the maximum value among all historical candidate netcities' corresponding topological similarities is reached, the historical candidate netcities are... The tracking ID is assigned to the current cage Otherwise, suspend automatic feeding of the cage.

4. The UAV re-deployment decision-making method based on dynamic revisit according to claim 2, characterized in that, The specific implementation process of step 3 is as follows: After the cage tracking ID is successfully matched, the drone moves towards the first A fixed number of net cages were deployed. The bait was provided, and the timing of the first feeding round was recorded. ; According to the current season And breeding experience to set seasonal preset waiting intervals The initial planned follow-up visit will take place at: For the The first net cage The planned return visit time for the feeding cycle is as follows: Current moment If the planned return visit time is reached or exceeded, and the corresponding return visit task has not yet been completed, the cage will be added to the return visit candidate set. Candidate cages are ranked from longest to shortest according to the timeout period of the follow-up visit, and priority is given to the cage that has exceeded the planned follow-up visit time for the longest time.

5. The UAV re-deployment decision-making method based on dynamic revisit according to claim 4, characterized in that, The specific implementation process of step 4 is as follows: After the drone arrives at the candidate cage, it records the actual time of the follow-up visit. And calculate from the first Observation time from feeding cycle to actual follow-up visit: The amount of residual bait is estimated using a residual bait detection model. and with a fixed feeding amount To standardize the calculation of residual bait ratio The residual bait detection model takes an image of residual bait on the surface of the net cage as input and an estimated amount of residual bait as output; the target residual bait ratio for allowing refeeding is set as... ;when When the system determines that the cage has a valid need for replenishment, a fixed amount of feed is released into the cage. The bait used will be recorded as the first time this bait has been added. Rotary feeding .

6. The UAV re-deployment decision-making method based on dynamic revisit according to claim 5, characterized in that, The specific implementation process of step 5 is as follows: Under the condition that the visible amount of uneaten bait changes approximately linearly with time, the time required for the proportion of uneaten bait to decrease to the target proportion is estimated based on the actual observation duration and the proportion of uneaten bait: After the supplementary payment is executed, the first The target waiting interval for the round is: The corresponding planned follow-up visit times are: When satisfied When necessary, make up for the shortfall; If no supplementary application is performed, the waiting interval for the next Revisit will still be [time period missing]. .

7. The UAV re-deployment decision-making method based on dynamic revisit according to claim 6, characterized in that, The specific implementation process of step 6 is as follows: During the effective operating period each day, whenever the detected residual bait ratio meets the requirements... Record one valid re-delivery request at a time; The first net cage in the... The number of valid refueling requests within a valid working day is recorded as follows: ; for the first The median number of replenishment requests for all valid cages within a valid workday is used to obtain the reference replenishment request number for that day. ; Select the first The most recently preset number of valid workdays for each net cage constitutes a rolling workday set. Calculate the rolling feeding state index: Based on the rolling feeding status index This generates corresponding feeding status levels and abnormal alerts.

8. A UAV re-deployment decision system based on dynamic revisit, used to implement the UAV re-deployment decision method according to any one of claims 1 to 7, characterized in that, Includes the following modules: The historical spatial topology map module is used to acquire top-down images of offshore cage aquaculture farms, perform cage target detection, and construct a historical spatial topology map. The cage tracking ID module is used to fly the drone to the area where the cage to be operated is located, construct the current local topology map, and determine the tracking ID of the cage to be operated based on the topological similarity between the current local topology map and the historical spatial topology map. The follow-up module is used to establish a follow-up candidate set based on the feeding time and target waiting interval of each cage, and to determine the follow-up order according to the duration of time exceeding the planned follow-up time; The refeeding module is used to reconfirm the net cage tracking ID after the drone arrives at the candidate net cage, collect images of residual bait on the water surface of the net cage, and determine whether to perform refeeding. The dynamic update module is used to update the next target waiting interval of the corresponding net cage based on the actual revisit time, the proportion of uneaten bait, and the proportion of target uneaten bait, so that the net cage revisit frequency changes dynamically with the feeding status of the fish. The rolling feeding status index module is used to count the number of valid refeeding needs generated by each cage within a valid working day, and to calculate the rolling feeding status index based on the number of refeeding needs of the cage group during the same period, generating feeding status levels and abnormal prompts.

Images