An intelligent monitoring system for an environment for breeding eocarcinus ivrous

Abstract

The application discloses an intelligent monitoring system for Eriocheir sinensis breeding environment, and particularly relates to the technical field of intelligent monitoring of aquaculture, which collects data through multiple climbing cage monitoring modules, calculates average weight according to the quantity and weight of image recognition and corrects the climbing cage frequency, carries out spatial analysis based on the corrected frequency to obtain a spatial aggregation global index and a local aggregation index, fuses to generate a spatial aggregation degree characteristic value, compares and calculates a deviation degree index by a deviation calculation module, and selectively starts the regulation and control equipment of different spatial regions according to the deviation degree index size by a regulation and control execution module; the climbing cage frequency is corrected through image and weight data, the single semantic problem of frequency data is solved, the spatial aggregation degree characteristic value is compared with the historical benchmark, the dynamic correlation between the environment and the behavior is realized, the deviation degree index is classified and partitioned for regulation and control, the execution lag problem is solved, and the precise intervention and energy consumption optimization are realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for aquaculture, and more specifically, to an intelligent monitoring system for the aquaculture environment of Chinese mitten crabs. Background Technology

[0002] Utilizing the climbing behavior of Chinese mitten crabs for aquaculture monitoring has become a research hotspot in this field in recent years. Current technologies employ climbing cage devices within the aquaculture area, using weight sensors to record the frequency of crab climbing, and supplementing this with image acquisition equipment for photographic recording, thus achieving non-destructive monitoring of crab activity. This technical solution, to a certain extent, replaces traditional manual pond patrols, improving monitoring efficiency.

[0003] In practical applications, existing technologies still have the following limitations: Current technology only records the number of times crabs climb onto the platform, but it cannot distinguish between "actively exploring and foraging" and "escaping due to environmental discomfort." During periods of high temperature or low dissolved oxygen, crabs will become agitated and crawl due to environmental stress. At this time, a high frequency of climbing the cage actually indicates an abnormal environment rather than good growth. However, the existing system is based on the simple logic that "the higher the frequency, the higher the activity," which may misinterpret stress signals as activity signals, leading to monitoring results that contradict the actual situation.

[0004] Current technologies analyze cage climbing frequency as an independent indicator, failing to couple it with environmental parameters such as water quality and weather over time. For example, the same cage climbing frequency of 100 times per day represents normal feeding needs at a water temperature of 28 degrees Celsius, while it represents a heat stress response at a water temperature of 35 degrees Celsius. However, current technologies lack the ability to dynamically adjust the interpretation threshold based on environmental parameters, making it difficult to accurately identify the true causes behind the behavior.

[0005] Existing technologies typically only trigger unified pond-wide operations such as switching aerators on and off, failing to intervene in advance based on qualitative changes reflected in cage-climbing behavior. By the time environmental anomalies become apparent through cage-climbing frequency data, they have often already caused actual impacts on the crabs, and the unified pond-wide control method is neither precise nor economical. Therefore, this invention proposes an intelligent monitoring system for the crab farming environment to address the aforementioned problems. Summary of the Invention

[0006] To achieve the above objectives, the present invention provides the following technical solution: A smart monitoring system for the crab farming environment includes: The climbing cage monitoring module is set up in multiple ways. Each climbing cage monitoring module has a cage body for the mitten crabs to climb, as well as a weight sensor and an image acquisition device integrated into the cage body. When the weight sensor is triggered, it generates weight data and simultaneously starts the image acquisition device to collect image data. The frequency statistics module is connected to each climbing cage monitoring module and records the frequency of crab climbing cages of each climbing cage monitoring module within each unit of time based on the trigger signal of the weight sensor. The spatial analysis module is connected to the frequency statistics module and each climbing cage monitoring module. It obtains the number of mitten crabs in each climbing cage monitoring module in each climbing cage event through image recognition, and calculates the average weight of each climbing cage event based on the weight data. Based on the average weight, the climbing cage frequency of each climbing cage monitoring module is corrected to obtain the corrected climbing cage frequency. Then, spatial analysis is performed based on the corrected climbing cage frequency to obtain the global spatial aggregation index and the local spatial aggregation index. Finally, the index is fused to generate the spatial aggregation degree feature value. The deviation calculation module, connected to the spatial analysis module, stores a sequence of spatial clustering benchmark values ​​for the same period within the historical best breeding cycle. It compares the current spatial clustering feature value with the benchmark value for the corresponding period in the spatial clustering benchmark value sequence and calculates the deviation index of the current spatial clustering feature value relative to the benchmark value. The control execution module is connected to the deviation calculation module and selectively activates control devices in different spatial areas within the aquaculture water area based on the magnitude of the deviation index.

[0007] In a preferred embodiment, multiple climbing cage monitoring modules are evenly distributed in a grid layout within the aquaculture area. The spacing between adjacent climbing cage monitoring modules is determined based on the maximum activity radius of the mitten crabs per unit time. This maximum activity radius is calculated by pre-collecting and observing mitten crab tracking data. The spacing between adjacent climbing cage monitoring modules is less than or equal to twice the maximum activity radius to ensure that a mitten crab in any location can reach at least one climbing cage monitoring module per unit time.

[0008] In a preferred embodiment, the maximum activity radius calculated by pre-collecting tracking and observation data of the mitten crabs refers to: Multiple individual crab samples were selected from the aquaculture area, and each sample was marked with an identifiable tag. The spatial position of each marked individual was recorded periodically during a preset observation period to generate position time series data. The maximum linear movement distance of each sample individual per unit time was calculated based on the position time series data. The maximum linear movement distance of all sample individuals was statistically analyzed, and the value of the predetermined percentile was taken as the maximum activity radius.

[0009] In a preferred embodiment, the method for obtaining the spatial aggregation global index includes: First, the global average is calculated based on the corrected climbing frequencies of all climbing cage monitoring modules. Then, a spatial weight matrix is ​​constructed, where each element is determined by the spatial distance between two corresponding climbing cage monitoring modules and the bio-social state vectors of the two modules through a preset bio-social coupling weight function. Next, the deviation between the corrected climbing frequency of each climbing cage monitoring module and the global average is calculated to obtain the deviation value for each module. Then, for each climbing cage monitoring module, its deviation value is multiplied by the weight elements of each module in the row of the spatial weight matrix, and all products are summed to obtain the spatial lag deviation value for that module. Then, the spatial lag deviation values ​​for all modules are accumulated to obtain the sum of spatial covariances. Then, the sum of squares of the deviation values ​​of all modules is calculated. Finally, the sum of spatial covariances is divided by the sum of squares to obtain the spatial aggregation global index.

[0010] In a preferred embodiment, the method for obtaining the local clustering index includes: For each cage-climbing monitoring module, firstly, its set of neighboring modules is determined based on a spatial distance threshold, which is related to the maximum activity radius. Then, the spatial weight between the module and each neighboring module is calculated using a biosocial coupling weighting function. Next, the neighbor-weighted average of the corrected cage-climbing frequency of each neighboring module is calculated with the spatial weight as the coefficient. Then, the first difference between the corrected cage-climbing frequency of the module and the neighbor-weighted average is calculated. Then, the second difference between the corrected cage-climbing frequency of the module and the global average is calculated. Then, the first difference and the second difference are multiplied to obtain the initial local clustering value of the module. Then, the standard deviation of the initial local clustering value of all cage-climbing monitoring modules is calculated. Finally, the initial local clustering value of the module is divided by the standard deviation to obtain the local clustering index of the module.

[0011] In a preferred embodiment, the biosocial coupling weight function uses a three-factor product form to determine the spatial weight between any two climbing cage monitoring modules: The biological social state vector is calculated based on the weight and image data of each climbing cage monitoring module. This biological social state vector consists of three dimensions: the first dimension is the average weight, which is calculated by summing the weight data of all climbing cage events of the module within a unit of time and dividing it by the total number of mitten crabs obtained from image recognition; the second dimension is the actual occurrence frequency, which is calculated by summing the actual number of mitten crabs in each climbing cage event obtained from image recognition; and the third dimension is the population density coefficient, which is calculated by the proportion of the number of events in each climbing cage event obtained from image recognition where the number of mitten crabs appearing at the same time exceeds a preset threshold to the total number of events of the module. The basic state influence factor is calculated based on the biological-social state vectors of the two climbing cage monitoring modules. This basic state influence factor is the dot product of the biological-social state vectors of the two modules divided by the product of the magnitudes of the biological-social state vectors of the two modules. The spatial distance attenuation factor is calculated based on the spatial distance between the two climbing cage monitoring modules. This spatial distance attenuation factor is determined by the spatial distance between the two climbing cage monitoring modules through a negative exponential attenuation function. The attenuation coefficient of this negative exponential attenuation function is set according to the maximum activity radius of the mitten crab, so that the factor takes the maximum value when the spatial distance is zero, decreases according to the negative exponential law when the spatial distance increases, and attenuates to zero when the spatial distance exceeds the maximum activity radius. The interaction sensing factor is calculated based on the bio-social state vectors of the two climbing cage monitoring modules. This interaction sensing factor is the magnitude of the difference between the bio-social state vectors of the two modules divided by the maximum magnitude of the bio-social state vectors of the two modules. The spatial weight between the two climbing cage monitoring modules is obtained by multiplying the basic state influence factor, spatial distance attenuation factor, and interactive perception factor.

[0012] In a preferred embodiment, the spatial clustering feature value is generated as follows: The local clustering indices of all climbing cage monitoring modules are normalized so that the value range of each local clustering index is mapped to the same numerical range as the global spatial clustering index. Then, the mean and standard deviation of the normalized local clustering indices are calculated. The sum of the mean and standard deviation is then divided by two to obtain the local comprehensive index. Finally, the global spatial clustering index is multiplied by the local comprehensive index to obtain the spatial clustering characteristic value.

[0013] In a preferred embodiment, the deviation index is calculated as follows: Calculate the absolute value of the difference between the current spatial clustering characteristic value and the corresponding unit time period benchmark value, then calculate the ratio of the absolute value of the difference to the corresponding unit time period benchmark value, and then multiply the ratio by a preset proportional coefficient to obtain the deviation index.

[0014] In a preferred embodiment, control devices for different spatial areas within the aquaculture area are selectively activated based on the magnitude of the deviation index, specifically including: When the deviation index is less than or equal to a preset first threshold, the control execution module does not activate any control equipment; when the deviation index is greater than the preset first threshold and less than or equal to a preset second threshold, the control execution module identifies the area where the cage monitoring module is located and the local aggregation index is lower than a preset low threshold, and activates the local control equipment for that area, which includes a local oxygenation device or a local odor-induced feeding device for that area; when the deviation index is greater than the preset second threshold, the control execution module activates the whole-area control equipment covering the entire aquaculture water area, which includes a whole-area oxygenation device, a whole-area circulating water device, or a whole-area emergency feeding device.

[0015] The technical effects and advantages of this invention are as follows: This invention collects weight and image data through a cage-climbing monitoring module, and a spatial analysis module obtains the number of crabs and calculates their average weight based on image recognition. The cage-climbing frequency is then corrected based on the average weight to obtain a corrected cage-climbing frequency, solving the problem of "quantity-only" reliance on cage-climbing frequency in existing technologies. The corrected cage-climbing frequency transforms the original trigger count into the actual number of occurrences, enabling the monitoring data to distinguish between single and multiple crabs entering simultaneously. This avoids underestimating the actual activity intensity due to multiple crabs being triggered at once, providing an accurate data foundation for subsequent spatial analysis.

[0016] This invention uses a spatial analysis module to obtain global and local spatial aggregation indices and fuse them to generate a spatial aggregation degree feature value. A deviation calculation module compares this feature value with a historical baseline sequence of spatial aggregation degree values ​​for the same period of the optimal breeding cycle to calculate a deviation index. This solves the problem of missing correlation between environmental and behavioral data in existing technologies. The global spatial aggregation index reflects the overall distribution pattern of mitten crabs across the entire water area, while the local aggregation index identifies hotspots and cold spots in specific areas. The fusion of these two indices and comparison with historical baseline values ​​allows the system to dynamically identify the degree of deviation between the current distribution pattern and the ideal state, avoiding the limitations of relying solely on a single frequency value.

[0017] This invention achieves precise, hierarchical, and zoned control by selectively activating control devices in different spatial regions based on the magnitude of the deviation index through a control execution module. This solves the problems of delayed and singular control execution in existing technologies. When the deviation index exceeds a first threshold but not a second threshold, local control devices are activated only in areas where the local aggregation index is below a preset low threshold, avoiding the crude approach of uniform operation across the entire pond. When the deviation index exceeds the second threshold, global control devices are activated for emergency handling, achieving a hierarchical progression from local intervention to global response. This ensures timely control while reducing energy consumption and costs. Attached Figure Description

[0018] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings; Figure 1 This is a schematic diagram of an intelligent monitoring system for the aquaculture environment of Chinese mitten crabs, as described in this invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] Reference Figure 1 The following examples were obtained: Example 1: An intelligent monitoring system for the crab farming environment, comprising: The climbing cage monitoring module consists of multiple modules. Each module has a cage for the mitten crabs to climb, along with a weight sensor and an image acquisition device integrated into the cage. When the weight sensor is triggered, it generates weight data and simultaneously activates the image acquisition device to collect image data. These multiple climbing cage monitoring modules are distributed across different locations in the aquaculture area according to a specific layout, with each module acting as an independent monitoring node. The cage structure of each module utilizes the mitten crabs' natural tendency to climb and explore to attract them. When a crab enters the cage, the weight sensor detects the weight change and is triggered, generating the current weight data. This weight data reflects the total weight of all crabs that have entered the cage during the current climbing event. Simultaneously, the weight sensor sends a synchronization start signal to the image acquisition device, which immediately captures an image of the area inside the cage, acquiring image data. The weight data and image data together constitute the raw record of a climbing event, with a one-to-one temporal correspondence, providing a foundation for subsequent data processing.

[0021] The frequency statistics module, connected to each cage monitoring module, records the frequency of crabs climbing the cages of each monitoring module within a unit of time based on the trigger signals from the weight sensors. The frequency statistics module receives trigger signals from the weight sensors of each monitoring module in real time. Each time a weight sensor of a monitoring module is triggered, the frequency statistics module increments the count for that module. The frequency statistics module summarizes and statistically analyzes the trigger counts of each monitoring module at preset unit times (e.g., every hour, every half hour) to obtain the crab climbing frequency of each monitoring module within each unit of time. This climbing frequency reflects the number of times the module is triggered by crabs within a unit of time, providing a preliminary quantitative indicator of the crab activity intensity.

[0022] The spatial analysis module is connected to the frequency statistics module and each cage monitoring module. It obtains the number of mitten crabs in each cage monitoring module during each cage climbing event through image recognition, and calculates the average weight of each cage climbing event based on the weight data. The cage climbing frequency of each cage monitoring module is corrected based on the average weight to obtain the corrected cage climbing frequency. Then, spatial analysis is performed based on the corrected cage climbing frequency to obtain the global spatial aggregation index and the local spatial aggregation index. Finally, they are fused to generate a spatial aggregation degree feature value.

[0023] The spatial analysis module receives climbing frequency data from each climbing monitoring module from the frequency statistics module, and simultaneously receives weight and image data corresponding to each climbing event from each climbing monitoring module. The spatial analysis module performs image recognition processing on the image data of each climbing event, using an image recognition algorithm to identify the number of individual crabs in the image, thus obtaining the total number of crabs in that climbing event. This image recognition overcomes the limitation of weight sensors, which can only sense the total weight but cannot distinguish the number of individuals. The spatial analysis module divides the weight data of the same climbing event by the number of crabs obtained through image recognition for that event, calculating the average weight of the crabs participating in that climbing event. This average weight reflects the average size of the crabs involved in that climbing event.

[0024] The spatial analysis module corrects the climbing frequency of each climbing cage monitoring module based on the average weight to obtain the corrected climbing frequency. This correction is necessary because the original climbing frequency only records the number of times the weight sensor is triggered. However, one trigger could correspond to one crab entering or multiple crabs entering simultaneously. For example, a climbing cage monitoring module might be triggered ten times within a unit of time. However, image recognition reveals that eight of these were single crabs entering, and two were three crabs entering simultaneously. Therefore, the actual total number of crabs appearing in that unit of time is eight times one plus two times three, equaling fourteen times. Directly using the original climbing frequency of ten for analysis would underestimate the actual activity intensity in that area. Therefore, the original climbing frequency needs to be corrected based on the actual number of crabs in each climbing event, transforming it into a corrected climbing frequency that reflects the true number of crab appearances. The revised calculation method for the frequency of crabs climbing the cage is as follows: the number of crabs climbing the cage in a unit of time is accumulated to obtain the actual frequency of occurrence of the module in that unit of time. Meanwhile, average weight data is also used for subsequent spatial analysis, because crabs of different weights have different activity levels and ecological significance; the activity of larger crabs may have a more significant environmental impact.

[0025] After obtaining the corrected climbing frequency, the spatial analysis module performs spatial analysis based on this frequency. The spatial analysis includes two aspects: first, obtaining a global spatial aggregation index, which reflects whether the corrected climbing frequencies of all climbing monitoring modules exhibit an overall aggregation trend across the entire aquaculture area, i.e., whether the mitten crabs are evenly distributed throughout the entire area or concentrated in a large region; second, obtaining a local aggregation index, which corresponds to a value for each climbing monitoring module, reflecting the degree of difference in the corrected climbing frequency of the module's local area compared to its neighboring modules, i.e., whether the module is a local hotspot. Finally, the spatial analysis module merges the obtained global spatial aggregation index with multiple local aggregation indices to generate a comprehensive spatial aggregation characteristic value. This characteristic value simultaneously includes both the overall distribution characteristics of the entire area and the distribution characteristics of local hotspots, providing a complete quantitative description of the spatial distribution of mitten crabs at the current moment.

[0026] The deviation calculation module, connected to the spatial analysis module, stores a sequence of spatial clustering benchmark values ​​for the same time period within the historical optimal breeding cycle. It compares the current spatial clustering characteristic value with the benchmark value for the corresponding time period in the spatial clustering benchmark value sequence, calculating the deviation index of the current spatial clustering characteristic value relative to the benchmark value. The deviation calculation module pre-stores a set of spatial clustering benchmark value sequences, which are derived from the spatial clustering characteristic values ​​recorded for various time periods (e.g., the same date and time as the current moment) within the historical optimal breeding cycle (e.g., the breeding cycle with the highest yield and best quality in previous years). This benchmark value sequence represents the standard spatial distribution pattern that should exist for different time periods under ideal breeding conditions.

[0027] The deviation calculation module receives the spatial clustering feature value of the current moment from the spatial analysis module and extracts the benchmark value corresponding to the time period of the current moment from the stored benchmark value sequence. Then, it compares the current feature value with the benchmark value to calculate the degree of deviation, obtaining a deviation index. The magnitude of this deviation index directly reflects the degree of difference between the current spatial distribution of the mitten crabs and the standard distribution pattern under the best historical farming conditions. The larger the deviation index, the more the current situation deviates from the ideal state, and the more likely there is an environmental anomaly.

[0028] The control execution module, connected to the deviation calculation module, selectively activates control devices in different spatial areas within the aquaculture area based on the magnitude of the deviation index. The control execution module receives the deviation index from the deviation calculation module and, based on the numerical range of the deviation index, decides and executes corresponding control operations. The control strategy of the control execution module is hierarchical and zoned: when the deviation index is low, it indicates that the current environmental condition is good and no intervention is needed; when the deviation index exceeds a certain threshold but has not yet reached a severe level, it indicates that a problem may occur in a local area. The control execution module identifies specific areas requiring intervention (e.g., the area where the local aggregation index is too low) and activates local control devices only in that area for precise intervention; when the deviation index is too high, it indicates that a serious anomaly may occur in the entire environment. The control execution module activates global control devices covering the entire aquaculture area for emergency treatment. This hierarchical and zoned control method ensures timely response to environmental changes while avoiding unnecessary energy consumption and intervention.

[0029] The multiple climbing cage monitoring modules are not placed randomly or arbitrarily, but rather evenly distributed throughout the entire aquaculture area using a grid layout. This grid layout involves dividing the aquaculture area into several regular grid units, with a climbing cage monitoring module placed at the center or a specific location in each grid unit, ensuring uniform coverage of the entire area. The straight-line distance between two adjacent climbing cage monitoring modules, i.e., the spacing, is not arbitrarily set, but determined based on the biological parameter of the mitten crab's maximum activity radius per unit time. The maximum activity radius refers to the maximum straight-line distance a mitten crab can move within a specific unit of time (e.g., one hour or half an hour), reflecting its activity level. This maximum activity radius is not an empirical estimate, but an actual measurement obtained through pre-collected mitten crab tracking and observation data and scientific calculations. The spacing between adjacent climbing cage monitoring modules is set to be less than or equal to twice this maximum activity radius. This design aims to ensure that regardless of the mitten crab's location within the aquaculture area, at least one climbing cage monitoring module is present within a circular area centered on that location and with the maximum activity radius as its radius. In other words, within a unit of time, any crab in any location can move to at least one monitoring module and be recorded, thus ensuring the comprehensiveness and representativeness of the monitoring data and avoiding monitoring blind spots.

[0030] For example, suppose that through tracking and observation, the maximum activity radius of a mitten crab within one hour is calculated to be five meters. Then, the spacing between adjacent trap monitoring modules should be set to less than or equal to ten meters. After gridding the layout according to a ten-meter spacing, a circle with any point in the water as the center and a radius of five meters will necessarily contain at least one trap monitoring module. This means that no matter where the mitten crab starts from, it can reach a trap monitoring module within one hour.

[0031] The process of collecting tracking and observation data of mitten crabs in advance and calculating the maximum activity radius includes the following steps: First, select multiple sample individuals of mitten crabs in the aquaculture water area. The selected sample individuals should be representative and cover mitten crabs of different sizes, sexes and vitality states to ensure that the maximum activity radius calculated later can reflect the general activity capacity of the entire group.

[0032] The second step is to apply identifiable markings to each individual sample. Marking methods can be external (such as attaching colored numbered labels to the carapace or binding colored marker rings to the legs) or internal (such as implanting electronic tags). The aim is to accurately distinguish and identify each individual sample in subsequent observations and avoid confusion.

[0033] The third step involves periodically recording the spatial location of each tagged individual within a pre-defined observation period, generating temporal location data. This pre-defined observation period can be continuous for 24 hours or longer, covering the peak and trough periods of crab activity. Within the observation period, the location of each tagged individual is measured and recorded at fixed time intervals (e.g., every ten minutes or every half hour). The records include the individual's ID, the recording time, and the coordinates of its location. After multiple recordings, each sample individual generates a temporal location data set consisting of time and spatial coordinates, reflecting the individual's movement trajectory within the observation period.

[0034] The fourth step is to calculate the maximum linear distance traveled by each sample individual per unit time based on the location time series data. For each sample individual, find the maximum linear distance between the starting and ending positions within any unit time interval (e.g., one hour) from the individual's location time series data. This maximum value is the maximum linear distance traveled by that sample individual per unit time. Only the linear distance is considered in the calculation; the tortuousness of the actual movement path is not taken into account, because the maximum activity radius focuses on displacement capacity rather than the actual movement distance.

[0035] The fifth step involves statistically analyzing the maximum straight-line movement distance of all sample individuals and taking the value at a predetermined percentile as the maximum activity radius. The maximum straight-line movement distances of all sample individuals constitute a dataset. This dataset is statistically analyzed, sorted from smallest to largest, and the value at a predetermined percentile is taken as the final maximum activity radius. The predetermined percentile is usually a relatively high value, such as 90% or 95%, to ensure that the obtained maximum activity radius covers the activity capabilities of the vast majority of crabs and avoids using extreme maximum values ​​that would result in an overly sparse distribution.

[0036] For example, select fifty Chinese mitten crabs as sample individuals, and attach a colored numbered label to each crab's carapace. Observe continuously for twelve hours from 8:00 AM to 8:00 PM, recording the location coordinates of each crab every half hour. Analyze the temporal data of each crab's location to find the farthest straight-line distance the crab moves within one hour (e.g., from 10:00 AM to 11:00 AM). Assume that fifty maximum straight-line distance data are obtained, namely 3.2 meters, 3.5 meters, 4.1 meters...6.8 meters. Sort these data from smallest to largest, and take the data at the 90th percentile, let's say 5.5 meters. Then 5.5 meters is the maximum activity radius of the Chinese mitten crabs in this aquaculture area. This means that the maximum straight-line distance of 90% of the Chinese mitten crabs within one hour does not exceed 5.5 meters. Based on this, the spacing of the monitoring modules can be set to ensure coverage of the activity range of the vast majority of Chinese mitten crabs.

[0037] Methods for obtaining the global index of spatial aggregation include: The process involves summarizing and statistically analyzing the corrected crawling frequencies of all crawling cage monitoring modules within a unit of time, and calculating their arithmetic mean, i.e., the global average. The global average represents the average level of crab activity intensity at each monitoring point across the entire aquaculture area. This average serves as the benchmark for all subsequent calculations, measuring the degree to which each specific monitoring point deviates from the overall average level. For example: Suppose there are ten crawling cage monitoring modules in the aquaculture area. In a certain hour, the corrected crawling frequencies of each module are 14, 16, 10, 22, 18, 15, 12, 21, 17, and 15 times respectively. Adding these ten values ​​gives a total of 160 times, which, divided by 10, yields a global average of 16 times. This 16 times represents the average intensity of crab activity across the entire aquaculture area for that hour.

[0038] After obtaining the global average, a spatial weight matrix needs to be constructed to quantify the degree of mutual influence between the various monitoring modules. This matrix is ​​a square matrix with the number of monitoring modules as its dimension. Each element in the matrix represents the spatial association strength between two modules in the corresponding row and column. The value of this element is not simply determined by distance, but is calculated through a pre-defined biosocial coupling weight function. This function takes into account the spatial distance between the two modules and the biosocial state vectors of each module. The biosocial state vector is a multi-dimensional vector calculated based on the weight data and image data of each module, containing information reflecting the individual size and group characteristics of the mitten crabs in the area where the module is located, such as average weight, actual occurrence frequency, and population density coefficient. The spatial weights calculated in this way not only consider the geographical distance, but also comprehensively consider the influence of the size and population density of the mitten crabs in the area where the module is located on the mutual attraction or repulsion, making the weight matrix more consistent with the actual situation of biosocial interaction. For example: Suppose there are three monitoring modules A, B, and C, with spatial distances of 5 meters between A and B, 10 meters between A and C, and 8 meters between B and C, respectively. The biosocial state vector of module A reflects that the region is dominated by large-sized crabs with a high density, module B by medium-sized crabs with a medium density, and module C by small-sized crabs with a low density. After calculation using the biosocial coupling weight function, the resulting spatial weight matrix shows that the weight between A and B might be 1.2, the weight between A and C might be 0.3, and the weight between B and C might be 0.6. This result indicates that although A and C are geographically distant, due to the high-sized and high-density characteristics of A, the correlation weight between them is still higher than that calculated by simple distance calculation.

[0039] After obtaining the global average, for each crab-climbing monitoring module, its corrected climbing frequency is subtracted from the global average, and the difference is calculated. This difference is the deviation value for that module. The deviation value reflects the direction and degree of deviation of the crab activity intensity in the area where the module is located from the average level of the entire water area. A positive deviation value indicates that the activity intensity in the area is higher than the average level; a negative deviation value indicates that the activity intensity in the area is lower than the average level; and a deviation value of zero indicates that it is at the average level. This series of deviation values ​​is the basic data for subsequent spatial autocorrelation calculations. For example: continuing the previous example, the global average is sixteen times. The corrected climbing frequencies of the ten modules are fourteen, sixteen, ten, twenty-two, eighteen, fifteen, twelve, twenty-one, seventeen, and fifteen, respectively. The corresponding deviation values ​​are -2, zero, -6, +6, +2, -1, -4, +5, +1, and -1, respectively.

[0040] For each climbing cage monitoring module, its spatial lag bias value needs to be calculated. The specific operation is as follows: Extract all weight elements from the row containing the module in the spatial weight matrix. These weight elements represent the spatial correlation strength between the module and all other modules, including itself. Ignore the module's own bias value temporarily, and multiply the bias value of each other module by its corresponding weight element, obtaining a series of products. Then, sum all these products; the sum is the spatial lag bias value of the module. The spatial lag bias value represents the combined impact of the bias values ​​of the module's neighboring modules on the module after considering spatial weights. If the bias values ​​of the neighboring modules are mostly positive and have large weights, the module's spatial lag bias value is a large positive value, indicating that the module is positively influenced by the surrounding active areas; otherwise, it is a negative value. For example: Suppose there are three modules A, B, and C. The weight elements in the row containing A in the spatial weight matrix are: A with itself has a weight of 0, A with B has a weight of 1.2, and A with C has a weight of 0.3. The bias values ​​of A, B, and C are +2, -1, and +6, respectively. The spatial lag bias of A is equal to the bias of B (negative one) multiplied by the weight 1.2, plus the bias of C (positive six) multiplied by the weight 0.3, which equals -1.2 plus 1.8, equaling positive 0.6. This positive 0.6 indicates that the combined influence of A's surrounding neighbors is positive, meaning the surrounding area is relatively active.

[0041] The spatial lag bias values ​​of all climbing cage monitoring modules are summed to obtain the total spatial covariance. This sum reflects the overall level of the spatial lag bias values ​​of all modules. In spatial statistics, this sum is correlated with the product of the bias values ​​of each module and is used to measure the degree of spatial clustering of the entire area. If the spatial lag bias values ​​of all modules are mostly positive and large, the sum is large, indicating positive spatial autocorrelation, i.e., high values ​​cluster with high values ​​and low values ​​cluster with low values; if they are mostly negative, the sum is negative, indicating negative spatial autocorrelation, i.e., high values ​​are adjacent to low values. For example: Suppose there are ten modules with spatial lag bias values ​​of 0.6, 0.2, -0.3, 1.1, 0.8, -0.1, -0.5, 0.9, 0.4, and -0.2. Summing these values ​​together, let's say the sum is 2.9, which is the total spatial covariance.

[0042] Simultaneously, it is necessary to calculate the sum of squares of the deviation values ​​of all climbing cage monitoring modules. This involves first squarening the deviation value of each module, and then summing all the squared values. The sum of squares reflects the total variation in the activity intensity of each module relative to the global average, i.e., the magnitude of the overall fluctuation. This value is used as the denominator to normalize the sum of spatial covariances. For example, if the deviation values ​​of ten modules are -2, 0, -6, +6, +2, -1, -4, +5, +1, and -1, first calculate the square of each deviation value: 4, 0, 36, 36, 4, 1, 16, 25, 1, 1. Then sum these squared values ​​to obtain a sum of squares of 124.

[0043] Dividing the sum of spatial covariances calculated above by the sum of squared deviations yields the global spatial clustering index. This index is a dimensionless value, typically ranging within a certain range (e.g., between -1 and +1). A positive index with a larger absolute value indicates a stronger positive clustering trend in the spatial distribution of mitten crabs within the aquaculture area, meaning high-activity areas are close to each other, and low-activity areas are close to each other. A negative index with a larger absolute value indicates a stronger negative clustering trend, meaning high-activity areas are spaced apart from low-activity areas. An index close to zero indicates a tendency towards random spatial distribution, with no obvious clustering or dispersion trend. This global index quantifies the overall spatial distribution pattern of mitten crabs across the entire water area. For example, dividing the sum of spatial covariances calculated above (2.9) by the sum of squared deviations (124) yields 0.023. This value is positive and relatively small, indicating that the current spatial distribution of mitten crabs shows a slight positive clustering trend, but the degree of clustering is not high, approaching a random distribution.

[0044] Methods for obtaining local clustering indices include: For each cage monitoring module, it's necessary to identify which other modules can be considered its "neighbors"—modules that are spatially close enough to potentially interact with it. To this end, a spatial distance threshold is set, correlated with the previously determined maximum activity radius of the mitten crab, typically equal to or slightly larger than the maximum activity radius. A circle is drawn with this spatial distance threshold as the radius, centered on each target module. Other cage monitoring modules falling within this circular area constitute the neighboring module set for that target module. This neighboring module set forms the basis for subsequent calculations of local spatial relationships; only modules that are sufficiently close are considered to have direct local interactions. For example, assuming the maximum activity radius is five meters and the spatial distance threshold is five meters, for cage monitoring module A, measuring the straight-line distances to other modules reveals that module B is three meters away, module C is four and a half meters away, and module D is six meters away. Modules B and C, being less than or equal to five meters away, are included in A's neighboring module set; module D, being more than five meters away, is not considered a neighboring module.

[0045] After determining the set of neighboring modules, it is necessary to quantify the association strength between the target module and each neighboring module, i.e., the spatial weight. This weight is not simply determined by distance, but is calculated using a biosocial coupling weight function. This function considers both the spatial distance between the two modules and their respective biosocial state vectors (including information such as average weight, actual occurrence frequency, and population density coefficient), allowing the weight to reflect the impact of the similarity or difference in biosocial characteristics on spatial interaction. The calculated spatial weights are used for subsequent weighted averaging. For example, for module A and its neighboring module B, the distance between A and B is known to be three meters. A's biosocial state vector reflects a high density of large-sized crabs in the area, while B represents a medium density of medium-sized crabs. The spatial weight between A and B, calculated using the biosocial coupling weight function, is 1.5. For A and C, the distance is 4.5 meters, and C represents a low density of small-sized crabs, resulting in a weight of 0.8. These weight values ​​represent the degree of local influence of B and C on A.

[0046] For target module A, the corrected climbing frequency of all its neighboring modules is multiplied by their corresponding spatial weights. These products are then summed and divided by the sum of the spatial weights of all neighboring modules. The result is the neighbor-weighted average of A. This weighted average reflects the overall level of crab activity intensity in the area surrounding A after considering the spatial weights. The larger the weight, the greater the impact of the corrected climbing frequency of the corresponding neighboring module on the average. For example, suppose A's neighboring modules are B and C. B has a corrected climbing frequency of 18 times with a weight of 1.5; C has a corrected climbing frequency of 10 times with a weight of 0.8. Then the neighbor-weighted average is equal to (18 multiplied by 1.5 plus 10 multiplied by 0.8) divided by (1.5 plus 0.8), which equals (27 plus 8) divided by 2.3, which equals 35 divided by 2.3, approximately equal to 15.22.

[0047] The first difference is obtained by subtracting the previously calculated neighbor-weighted average from the corrected climbing frequency of target module A. This first difference reflects the difference between the activity intensity of the area where A is located and the average activity intensity of its neighboring areas. If the first difference is positive, it indicates that the activity intensity of A is higher than the surrounding average; if it is negative, it indicates that it is lower than the surrounding average; if it is close to zero, it indicates that it is on par with the surrounding areas. For example, assuming the corrected climbing frequency of module A is 22, and the neighbor-weighted average is 15.2, then the first difference equals 22 minus 15.2, which equals 6.8. This positive number indicates that the activity intensity of A is significantly higher than that of its surrounding areas.

[0048] The second difference is obtained by subtracting the previously calculated global average (the average of the corrected cage-climbing frequencies of all modules) from the corrected cage-climbing frequency of target module A. This second difference reflects the difference between the activity intensity of area A and the overall average level of the entire aquaculture area. A positive second difference indicates that A's activity intensity is higher than the overall average; a negative second difference indicates that A's activity intensity is lower than the overall average. For example, assuming the global average is sixteen times and module A's corrected cage-climbing frequency is twenty-two times, the second difference equals twenty-two minus sixteen times, which equals six times. This positive number indicates that A's activity intensity is higher than the overall average level of the entire aquaculture area.

[0049] Multiplying the first difference by the second difference yields the initial local clustering value for that module. This product combines two dimensions: "local difference" and "global difference." The first difference measures deviation relative to the surrounding area, while the second difference measures deviation relative to the overall area. The product amplifies hotspots that are higher than both the surrounding area and the overall area (positive times positive equals positive), and coldspots that are lower than both the surrounding area and the overall area (negative times negative equals positive). For areas where the deviation direction is inconsistent with the overall direction (e.g., higher than the surrounding area but lower than the overall area, or lower than the surrounding area but higher than the overall area), the product may be negative. Therefore, the sign and magnitude of the initial local clustering value can initially identify true local hotspots or coldspots. For example, module A has a first difference of 6.8 times and a second difference of 6 times; multiplying them yields 40.8 (squared). This large positive value indicates that A is a clear local hotspot, meaning its activity intensity is significantly higher than both the surrounding area and the global average.

[0050] To standardize the initial local cluster values ​​of each module, it is necessary to first calculate the standard deviation of all initial local cluster values. The standard deviation reflects the dispersion of the initial local cluster values ​​of all modules, i.e., the magnitude of the differences between them. A larger standard deviation indicates a more dispersed distribution of local cluster values ​​among the modules; a smaller standard deviation indicates a more concentrated distribution. This standard deviation will be used as the denominator in subsequent standardization steps to eliminate differences in dimensions and scale. For example, suppose there are ten climbing cage monitoring modules with initial local cluster values ​​of 40.8, -20.3, 15.6, -5.2, 30.1, -12.7, 8.9, -30.5, 25.4, and -8.6. Calculate the standard deviation of these values, assuming it is 20.5.

[0051] For each cage-climbing monitoring module, its initial local clustering value is divided by the standard deviation of the initial local clustering values ​​of all modules. The quotient is the local clustering index of that module. This standardization process makes the local clustering index a dimensionless relative value, facilitating comparisons between different modules and different time periods. The absolute value of the local clustering index indicates the intensity of the module as a local hotspot or coldspot. A positive value indicates a hotspot (activity intensity higher than the surrounding area and higher than the overall area), and a negative value indicates a coldspot (activity intensity lower than the surrounding area and lower than the overall area). The larger the absolute value, the more significant the feature. Each module has its own corresponding local clustering index, which can identify specific hotspot and coldspot areas within the aquaculture area, providing a locational basis for subsequent precise control. For example, for module A, its initial local clustering value is 40.8, and its standard deviation is 20.5. Then, the local clustering index is 40.8 divided by 20.5, approximately equal to 1.99. This value, greater than 1, indicates that module A is a significant local hotspot area. For another module, assuming its initial local clustering value is -30.5, dividing by the standard deviation of 20.5 yields -1.49, indicating that this module is a significant local coldspot area.

[0052] The biosocial coupling weight function uses a three-factor product to determine the spatial weight between any two monitoring modules. The first step of the biosocial coupling weight function is to construct a multi-dimensional vector, namely the biosocial state vector, for each monitoring module that can comprehensively describe the biological and social characteristics of the mitten crabs in the area where the module is located. This vector contains three dimensions, each calculated based on the weight data and image data collected by the module itself.

[0053] The first dimension is average weight. For each cage-climbing monitoring module, all cage-climbing events that occur within a unit of time are counted, and the weight data of each event is summed to obtain the total weight of the module within that unit of time. Simultaneously, the actual number of crabs appearing in each cage-climbing event is obtained through image recognition, and the numbers from all events are summed to obtain the total number of crabs appearing in that module within that unit of time. Dividing the total weight by the total number yields the average weight of the module. Average weight reflects the average size of crabs in the region; larger crabs typically indicate good fattening conditions or older age.

[0054] The second dimension is the actual occurrence frequency. Image recognition is used to obtain the actual number of crabs appearing in each cage-climbing event. These numbers are then summed across all events within a unit of time to obtain the module's actual occurrence frequency within that unit of time. This frequency differs from the original cage-climbing frequency; it corrects for situations where one event might correspond to multiple crabs, accurately reflecting how many crabs visited the module. A higher actual occurrence frequency indicates a stronger attraction of the area for crabs.

[0055] The third dimension is the population density coefficient. Image recognition is used to determine whether the number of crabs appearing simultaneously in each trap-climbing event exceeds a preset threshold, such as three. The number of events exceeding this threshold is counted and divided by the total number of events per unit time to obtain the population density coefficient. This coefficient reflects the degree to which crabs in the area tend to engage in group or individual activity. A higher coefficient indicates more frequent group activity, potentially signifying abundant food or a suitable environment for crab aggregation.

[0056] For example: A crab-climbing cage monitoring module recorded ten climbing events within one hour. Image recognition results showed that five events involved a single crab entering, three events involved two crabs entering simultaneously, and two events involved four crabs entering simultaneously. The weight data were as follows: the five single-crab events each weighed 150g, 160g, 155g, 165g, and 158g; the three two-crab events each weighed 310g, 320g, and 305g; and the two four-crab events each weighed 630g and 640g. First, the total number is calculated as 5 x 1 + 3 x 2 + 2 x 4 = 5 + 6 + 8 = 19 crabs. The total weight is the sum of the five single-crab weights (788g), the sum of the three two-crab weights (935g), and the sum of the two four-crab weights (1270g), equaling 2993g. The average weight is 2993g divided by 19 crabs, approximately 157.5g. The actual frequency of occurrence was 19 times. The preset threshold is three animals. There are two instances where four animals enter at the same time, which exceed the threshold. The total number of events is ten, and the population density coefficient is equal to two divided by ten, which equals 0.2.

[0057] After obtaining the biosocial state vector for each module, the basic state influence factor between any two modules needs to be calculated. This factor is calculated by performing a dot product operation on the biosocial state vectors of the two modules; that is, multiplying the corresponding dimension values, summing the results, and then dividing the dot product by the product of the magnitudes of the two vectors. The product of the magnitudes of the two vectors is the product of the square roots of the sum of the squares of the values ​​of each dimension of each vector. This result is actually the cosine of the angle between the two vectors, ranging from zero to one. A larger value indicates that the two vectors are more aligned in direction, meaning the biosocial characteristics of the two modules are more similar; a smaller value indicates a greater difference in direction, meaning the biosocial characteristics of the two modules are more different. The basic state influence factor reflects the fundamental impact of the similarity in size, activity intensity, and group behavior between two regions of mitten crabs on their spatial interaction.

[0058] For example: Assume module A's biological social state vector is an average weight of 157.5 grams, an actual occurrence frequency of 19 times, and a population density coefficient of 0.2; module B's vector is an average weight of 142 grams, an actual occurrence frequency of 14 times, and a population density coefficient of 0.3. First, calculating the dot product of A requires combining it with B, but the actual calculation requires two vectors. Assume B's values ​​are as shown above. The dot product of A and B is 157.5 multiplied by 142, plus 19 multiplied by 14, plus 0.2 multiplied by 0.3, resulting in approximately 22,365 plus 266 plus 0.06, which equals 22,631.06. The modulus of A is the square root of 157.5 squared plus 19 squared plus 0.2 squared, approximately equal to the square root of 24,806.25 plus 361 plus 0.04, which equals the square root of 25,167.29, approximately equal to 158.6. The modulus of B is calculated as 142 squared + 14 squared + 0.3 squared, which is approximately equal to 20,164 squared + 196 squared + 0.09 squared, or 20,360.09 squared, or approximately 142.7. The product of the modulus lengths is 158.6 multiplied by 142.7, which is approximately 22,638. The basic state influence factor is 22,631.06 divided by 22,638, which is approximately 0.997, indicating that the biological and social characteristics of the two modules are highly similar.

[0059] For any two climbing cage monitoring modules, the spatial distance between them is a crucial factor influencing spatial interaction. A spatial distance attenuation factor is used to quantify the impact of distance on interaction intensity. This factor is calculated using a negative exponential attenuation function. The input to the function is the spatial distance between the two modules, and the output is a value between zero and one. The attenuation coefficient is set based on the maximum activity radius of the mitten crab. Specifically, when the spatial distance is zero, the attenuation factor reaches its maximum value of one; as the spatial distance gradually increases, the attenuation factor gradually decreases according to a negative exponential law; when the spatial distance exceeds the maximum activity radius, the attenuation factor decays to zero, indicating that beyond this distance, there is no longer any direct spatial interaction between the two modules. This design conforms to the basic laws of biological activity: the closer the distance, the stronger the mutual influence; when the distance exceeds the activity range, the mutual influence becomes negligible.

[0060] For example: Assume the maximum activity radius is five meters. The negative exponential decay function is set so that the factor is one when the distance is zero, 0.13 when the distance is five meters, and zero when the distance exceeds five meters. For modules A and B, the spatial distance is three meters, and the calculated spatial distance decay factor is 0.37. For modules A and C, the spatial distance is four and a half meters, and the calculated factor is 0.15. For modules A and D, the spatial distance is six meters, exceeding the maximum activity radius, and the factor is zero.

[0061] Besides similarity and distance, the differences between two regions can also generate interaction dynamics. For example, a region with large crabs may attract smaller crabs to forage or explore. The interaction perception factor is used to quantify the intensity of this difference-driven interaction. This factor is calculated as follows: First, calculate the difference vector between the biosocial state vectors of the two modules, i.e., subtract the corresponding dimension values; then calculate the magnitude of this difference vector, i.e., the square root of the sum of the squares of the differences in each dimension; next, calculate the larger of the magnitudes of the biosocial state vectors of the two modules; finally, divide the difference magnitude by the larger magnitude to obtain a value between zero and one. A larger factor indicates a greater difference in the biosocial states of the two modules, potentially leading to stronger interaction perception; a smaller factor indicates a closer similarity in the states of the two modules, resulting in weaker difference-driven interaction.

[0062] For example, let's continue using the vector data from modules A and B. A's vector is 157.5, 19, 0.2, and B's vector is 142, 14, 0.3. The difference vector is 15.5, 5, -0.1. The difference magnitude is the square root of 15.5² + 5² + 0.1², which equals the square root of 240.25 + 25 + 0.01, which equals the square root of 265.26, approximately 16.3. A's magnitude is 158.6, and B's magnitude is 142.7, with the larger value being 158.6. The interaction perception factor is 16.3 divided by 158.6, approximately 0.103. This smaller value indicates that the difference between the two modules is not significant, resulting in weak difference-driven interaction.

[0063] Multiplying the three factors above yields the spatial weight between the two monitoring modules. This product ensures the spatial weight is influenced by three factors simultaneously: the baseline state factor represents the contribution of biosocial similarity, the spatial distance attenuation factor represents the contribution of geographical proximity, and the interaction perception factor represents the contribution of biosocial differences. Multiplying these three factors means that a small value in any one factor will lower the final spatial weight; only when all three factors are large will the spatial weight be significantly large. The spatial weight calculated in this way is neither a purely geometric weight based on distance nor a purely feature-based similarity weight, but a composite weight that comprehensively considers distance, similarity, and differences—three biosocial spatial factors—more realistically reflecting the actual interaction intensity between crab populations in different areas.

[0064] For example: For modules A and B, the basic state influence factor is 0.997, the spatial distance attenuation factor is 0.37, and the interaction perception factor is 0.103. Multiplying these three together gives 0.997 multiplied by 0.37 multiplied by 0.103, which is approximately 0.038. For modules A and C, assuming the basic state influence factor is 0.85, the spatial distance attenuation factor is 0.15, and the interaction perception factor is 0.25, then the spatial weight is 0.85 multiplied by 0.15 multiplied by 0.25, which is approximately 0.032. For modules A and D, the spatial distance attenuation factor is zero, and regardless of other factors, the spatial weight is zero. These final spatial weight values ​​will be filled into the corresponding positions in the spatial weight matrix for subsequent calculations of the global and local spatial aggregation indices.

[0065] The method for generating spatial clustering feature values ​​is as follows: Before fusion, it's crucial to address the inconsistency in dimensions and value ranges between the global and local clustering indices. The global clustering index typically has a fixed numerical range, such as between -1 and +1. While the local clustering index, after standardization, is a dimensionless relative value, its actual value range may vary depending on data distribution and may not perfectly align with the global index's range. To meaningfully fuse them, the local clustering indices of all crawler monitoring modules need to be normalized. Normalization aims to find the minimum and maximum values ​​among all local clustering indices, then uses a linear transformation to map each local clustering index to a new numerical range identical to the global clustering index's range. After normalization, the minimum value of the local clustering index becomes equal to the minimum value of the global index, the maximum value becomes equal to the maximum value of the global index, and intermediate values ​​correspond proportionally. This places the global index and the normalized local clustering index on the same comparable scale, laying the foundation for subsequent fusion calculations. For example: Suppose the global spatial aggregation index ranges from -1 to +1. Currently, there are ten cage-climbing monitoring modules with local aggregation indices of 1.99, -1.49, 0.76, -0.25, 1.47, -0.62, 0.43, -1.50, 1.24, and -0.42. The minimum value is -1.50, and the maximum value is 1.99. During normalization, -1.50 is mapped to -1, 1.99 to +1, and intermediate values ​​are mapped proportionally. For example, a local aggregation index of 0.76 might become 0.15 after normalization, ensuring that all normalized local aggregation indices fall within the range of -1 to +1.

[0066] After normalization, the normalized local clustering indices of all cage monitoring modules are statistically analyzed, and their mean and standard deviation are calculated. The mean reflects the average level of local clustering in various local areas across the entire aquaculture area, i.e., whether the overall distribution is biased towards hot spots, cold spots, or roughly balanced. The standard deviation reflects the magnitude of the differences in local clustering between different local areas, i.e., whether the distribution of hot spots and cold spots is disparate. The mean and standard deviation describe the overall distribution characteristics of the local clustering indices from different perspectives: the mean indicates the central tendency, and the standard deviation indicates the degree of dispersion. For example, suppose the normalized local clustering indices of ten modules are 0.35, -0.62, 0.15, -0.10, 0.30, -0.25, 0.05, -0.65, 0.25, and -0.18. Calculating the mean of these values, let's say it's -0.07, indicating a slight bias towards cold spots. Calculating the standard deviation of these values, let's say it's 0.35, indicating a moderate degree of difference between the modules.

[0067] The calculated mean and standard deviation are added together, and then the sum is divided by two to obtain the local composite index. This step aims to combine the average level and dispersion of the local clustering index into a single value to characterize the local spatial distribution characteristics of the entire aquaculture area. The sum of the mean and standard deviation includes both the overall tendency and the magnitude of the differences. Dividing by two adjusts the composite value to a scale range similar to the global index, avoiding an excessively large value due to addition. The resulting local composite index is a value within the possible range of the normalized local clustering index, comprehensively reflecting the overall intensity of local hotspots and colds, as well as the degree of contrast between them. For example, if the mean is -0.07 and the standard deviation is 0.35, the sum is 0.28, which, divided by two, yields 0.14. This 0.14 is the local composite index, ranging from -1 to +1, a positive value that is not large, indicating that the local area is slightly hotter overall, but to a moderate degree.

[0068] Multiplying the previously calculated global spatial aggregation index with the newly obtained local comprehensive index yields the final spatial aggregation characteristic value. This multiplication operation ensures that the characteristic value simultaneously incorporates information about both the global spatial distribution pattern and local spatial distribution characteristics. The global index reflects the overall trend of whether the mitten crabs are clustered or dispersed across the entire water area, while the local comprehensive index reflects the overall intensity and contrast of local hotspots and cold spots. The result of multiplying the two indices provides a more comprehensive description of the spatial distribution of mitten crabs at the current moment: if both the global index and the local comprehensive index are large positive values, the characteristic value is also large positive, indicating overall clustering with prominent local hotspots; if the global index is positive and the local comprehensive index is negative, the characteristic value is negative, indicating that although there is a clustering trend across the entire area, some local areas are relatively cold; if both are zero or close to zero, the characteristic value is close to zero, indicating a uniform distribution. This characteristic value will be used as input for subsequent deviation calculations, comparing it with the baseline value of the historical best cycle. For example: Suppose the global spatial aggregation index is 1.2 (note that the global index itself may be scaled; this is just an example value), and the local comprehensive index is 0.14. Multiplying the two gives 1.2 multiplied by 0.14, which equals 0.168. This is the spatial aggregation characteristic value at the current moment. If the baseline value for the same period in the historical best cycle is 0.20, then the current characteristic value of 0.168 is slightly lower than the baseline value, indicating that the current spatial distribution deviates somewhat from the ideal state.

[0069] The first step in calculating the deviation index is to compare the spatial clustering characteristic value calculated at the current moment with the corresponding unit-period benchmark value extracted from the historical optimal breeding cycle's benchmark value sequence. The difference between the two is calculated, and the absolute value of this difference is taken. The spatial clustering characteristic value is a numerical value that comprehensively reflects the current spatial distribution of mitten crabs, while the benchmark value represents the standard distribution state that should exist under the historical optimal breeding conditions at the same time. The absolute value of the difference is calculated because whether the characteristic value is higher or lower than the benchmark value, it means a deviation from the ideal state. Different directions of deviation may represent different types of anomalies, but the degree of deviation needs to be quantified into a non-negative value. After taking the absolute value, the resulting difference is always a positive number; the larger the value, the greater the degree of deviation. For example: Suppose the spatial clustering characteristic value at the current moment is 0.168, and the corresponding unit-period benchmark value extracted from the benchmark value sequence is 0.20. The difference is calculated as 0.168 minus 0.20, which equals -0.032. The absolute value is 0.032. If the current feature value is 0.25 and the baseline value is still 0.20, then the difference is 0.05, and the absolute value is also 0.05. Regardless of whether it is positive or negative, the degree of deviation is quantified as 0.032 or 0.05.

[0070] After obtaining the absolute value of the difference, divide it by the baseline value for the corresponding time period to calculate the ratio. This step converts the absolute deviation into a relative degree of deviation. Since the baseline value may differ across time periods—for example, the baseline value differs significantly between the peak and stagnant periods of crab growth—the same absolute deviation may indicate a severe deviation in a period with a smaller baseline value, while it may only represent a minor fluctuation in a period with a larger baseline value. Calculating the ratio eliminates the influence of the absolute size of the baseline value, allowing the deviation to reflect the proportion of deviation relative to the standard state, making it more scientific and comparable. This ratio is a dimensionless, non-negative number representing the relative magnitude of the current characteristic value's deviation from the baseline value. For example, if the absolute value of the difference is 0.032 and the baseline value is 0.20, the ratio is 0.032 divided by 0.20, which equals 0.16, or 16%. This means the current characteristic value deviates from the baseline value by 16%. If the absolute value of the difference in another time period is 0.05 and the baseline value is 0.35, the ratio is 0.14, or 14%. Although the absolute difference is larger, the relative deviation is smaller.

[0071] The calculated ratio is multiplied by a preset proportionality coefficient, and the product is the final deviation index. The proportionality coefficient amplifies or adjusts the ratio to a numerical range that is easy to observe and compare. For example, if the deviation index is to be presented as a percentage, the proportionality coefficient can be set to one hundred. Thus, multiplying the ratio 0.16 by one hundred yields sixteen, representing a deviation of sixteen percent, which is intuitive and easy to understand. If the deviation index is to be between zero and one, the proportionality coefficient can be set to a value less than or equal to one. The specific value of the proportionality coefficient can be predetermined according to the actual application scenario and threshold setting requirements, as long as it remains consistent throughout the system. The deviation index obtained after multiplication is a dimensionless value, and its magnitude directly reflects the degree to which the current spatial aggregation deviates from the historical optimal state. A larger value indicates a more severe deviation, and a smaller value indicates a closer relationship to the ideal state. This index will serve as input to the subsequent control execution module, used to decide whether to activate the control equipment and what level of control to activate.

[0072] For example: If the preset ratio coefficient is 100, the ratio value is 0.16, and the deviation index is 0.16 multiplied by 100, which equals 16. If the ratio value is 0.32, the deviation index is 32. If the ratio value is 0.05, the deviation index is 5. The resulting deviation index typically ranges from 0 to 100, providing a clear numerical value and facilitating the setting of thresholds for tiered control. For instance, if the first threshold is set to 30 and the second threshold to 60, a deviation index of 16 (less than 30) indicates no intervention is needed; 32 (between 30 and 60) indicates the need for localized control; and an index exceeding 60 requires nationwide emergency control.

[0073] The deviation index directly reflects the degree of deviation between the current spatial distribution of mitten crabs and the historical optimal farming conditions. The preset first threshold is a pre-defined numerical limit used to distinguish between normal environmental conditions and conditions requiring attention. When the deviation index is less than or equal to this first threshold, it indicates that the current spatial aggregation characteristic value differs very little from the historical optimal benchmark value. The spatial distribution of mitten crabs within the entire farming area is within the normal fluctuation range, the environmental conditions are good, and no abnormalities requiring artificial intervention have occurred. Under these circumstances, the control execution module does not activate any control equipment, maintaining the natural state of the farming environment. This avoids unnecessary energy consumption and equipment wear and tear, and also reduces interference with the normal activities of the mitten crabs. For example: The preset first threshold is thirty. The currently calculated deviation index is sixteen, less than or equal to thirty. At this time, the control execution module judges the environmental conditions to be normal and does not activate any aeration devices, feeding devices, or circulating water devices; the farming activities remain unchanged.

[0074] When the deviation index exceeds the first threshold but not the second threshold, it indicates that the current spatial distribution has deviated from the normal range, requiring attention and intervention, but the deviation has not yet reached a severe, area-wide anomaly. At this point, precise intervention is needed, rather than blindly activating all pond-wide equipment. The control execution module first retrieves the local aggregation index of all climbing cage monitoring modules and compares it with a preset low threshold. Areas where the local aggregation index is below the low threshold are identified as target areas requiring intervention. These areas are typically cold spots where crab activity is significantly reduced, possibly due to insufficient dissolved oxygen, lack of food, or other environmental factors causing crabs to avoid them. For these identified areas, the control execution module activates corresponding local control equipment. Local aeration devices increase the dissolved oxygen concentration in the area, improving water quality; local odor-induced feeding devices attract crabs from surrounding areas by releasing wet fermented feed with a special odor, using the diffusion of the odor. This localized and precise control method addresses environmental deficiencies in problem areas while avoiding unnecessary intervention across the entire pond, embodying the concepts of energy efficiency and precise management. For example: The preset first threshold is 30, the second threshold is 60, and the preset low threshold is 0.5. The current deviation index is 32, which is greater than 30 and less than or equal to 60. The control execution module checks the local aggregation index of the ten climbing cage monitoring modules and finds that the local aggregation index of module D is -1.49 and the local aggregation index of module H is -1.50, both below the low threshold of 0.5. Therefore, the two areas where modules D and H are located are identified as target areas requiring intervention. The local oxygenation device and the local odor-induced feeding device in these two areas are activated respectively to enhance the local environmental attractiveness.

[0075] When the deviation index further increases and exceeds the preset second threshold, it indicates that the current spatial distribution has deviated significantly from the historical optimal state, potentially signifying a global environmental anomaly in the entire aquaculture area, such as pond-wide hypoxia, excessively high water temperature, or water quality deterioration. At this point, localized control is insufficient to resolve the problem, and it is necessary to activate comprehensive control equipment covering the entire aquaculture area for emergency treatment. The comprehensive aeration device simultaneously activates all aeration equipment to rapidly increase the dissolved oxygen level throughout the entire area; the comprehensive circulating water device activates high-powered water pumps to promote water flow and exchange throughout the area, improving overall water quality; and the comprehensive emergency feeding device evenly distributes feed throughout the entire area to stabilize the crabs' feeding activities and reduce abnormal behavior caused by environmental stress. This comprehensive emergency control aims to quickly curb the trend of environmental deterioration, ensuring the survival and growth safety of the crabs. Normal management will be gradually restored based on changes in the deviation index after the environment recovers. For example: the preset first threshold is 30, and the second threshold is 60. The current deviation index is 65, which is greater than 60. If the control and execution module determines that a global environmental anomaly has occurred, it will immediately activate the global oxygenation device covering the entire aquaculture area, and at the same time start the global circulating water device. If necessary, it will activate the global emergency feeding device to carry out unified emergency control of the entire water area until the deviation index falls back to within the safe range.

[0076] The specific values ​​of the thresholds and coefficients involved in this invention are comprehensively set based on the biological characteristics of mitten crabs, breeding experience, and statistical laws. The predetermined percentile of the maximum activity radius is usually set at 90% or 95%, ensuring coverage of the vast majority of individuals based on the distribution characteristics of mitten crab activity capacity. The preset threshold for the population density coefficient is set to three individuals based on observation data of mitten crab population behavior; this value serves as the statistical distinction point between natural aggregation and accidental co-occurrence of mitten crabs. The attenuation coefficient of the spatial distance attenuation function is calculated inversely based on the maximum activity radius, so that the attenuation factor drops to approximately 0.13 when the distance equals the maximum activity radius, conforming to the general law of biological interaction attenuation with distance. The division by two operation in the normalization processing of the local aggregation index and the calculation of the local comprehensive index maintains dimensional consistency based on mathematical transformation. The proportional coefficient of the deviation index is set to one hundred, aiming to convert the ratio into an intuitive percentage form. The first threshold of thirty, the second threshold of sixty, and the low threshold of the local aggregation index of 0.5 are adjusted based on the statistical distribution of normal fluctuation range and abnormal deviation degree in historical breeding data, reflecting the distinction between slight and severe abnormalities. The specific values ​​mentioned above can be adjusted by system users based on actual aquaculture water conditions, the characteristics of the mitten crab species, and the accumulation of historical operating data.

[0077] The above-mentioned models or function formulas are all dimensionless and numerical calculations. The models or function formulas are obtained by software simulation based on a large amount of collected data to obtain the most recent real situation. The preset parameters in the models or function formulas are set by those skilled in the art according to the actual situation.

[0078] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0079] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0080] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0081] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A smart monitoring system for the crab farming environment, characterized in that, include: The climbing cage monitoring module is set up in multiple ways. Each climbing cage monitoring module has a cage body for the mitten crabs to climb, as well as a weight sensor and an image acquisition device integrated into the cage body. When the weight sensor is triggered, it generates weight data and simultaneously starts the image acquisition device to collect image data. The frequency statistics module is connected to each climbing cage monitoring module and records the frequency of crab climbing cages of each climbing cage monitoring module within each unit of time based on the trigger signal of the weight sensor. The spatial analysis module is connected to the frequency statistics module and each climbing cage monitoring module. It obtains the number of mitten crabs in each climbing cage monitoring module in each climbing cage event through image recognition, and calculates the average weight of each climbing cage event based on the weight data. Based on the average weight, the climbing cage frequency of each climbing cage monitoring module is corrected to obtain the corrected climbing cage frequency. Then, spatial analysis is performed based on the corrected climbing cage frequency to obtain the global spatial aggregation index and the local spatial aggregation index. Finally, the index is fused to generate the spatial aggregation degree feature value. The deviation calculation module, connected to the spatial analysis module, stores a sequence of spatial clustering benchmark values ​​for the same period within the historical best breeding cycle. It compares the current spatial clustering feature value with the benchmark value for the corresponding period in the spatial clustering benchmark value sequence and calculates the deviation index of the current spatial clustering feature value relative to the benchmark value. The control execution module is connected to the deviation calculation module and selectively activates control devices in different spatial areas within the aquaculture water area based on the magnitude of the deviation index.

2. The intelligent monitoring system for the crab farming environment according to claim 1, characterized in that, Multiple climbing cage monitoring modules are evenly distributed in a grid layout within the aquaculture area. The spacing between adjacent climbing cage monitoring modules is determined based on the maximum activity radius of the mitten crabs per unit time. This maximum activity radius is calculated by pre-collecting and observing the tracking data of the mitten crabs. The spacing between adjacent climbing cage monitoring modules is less than or equal to twice this maximum activity radius to ensure that a mitten crab in any position can reach at least one climbing cage monitoring module per unit time.

3. The intelligent monitoring system for the crab farming environment according to claim 2, characterized in that, The maximum activity radius obtained by collecting and tracking data of mitten crabs in advance refers to: Multiple samples of Chinese mitten crabs were selected from the aquaculture area, and each sample was marked with an identifiable tag. The spatial location of each marked individual is recorded periodically during the preset observation period to generate location time series data; the maximum linear distance of each sample individual is calculated per unit time based on the location time series data. Statistical analysis was performed on the maximum straight-line movement distance of all sample individuals, and the value of the predetermined percentile was taken as the maximum activity radius.

4. The intelligent monitoring system for the crab farming environment according to claim 3, characterized in that, Methods for obtaining the global index of spatial aggregation include: First, the global average is calculated based on the corrected climbing frequencies of all climbing cage monitoring modules. Then, a spatial weight matrix is ​​constructed, where each element is determined by the spatial distance between two corresponding climbing cage monitoring modules and the bio-social state vectors of the two modules through a preset bio-social coupling weight function. Next, the deviation between the corrected climbing frequency of each climbing cage monitoring module and the global average is calculated to obtain the deviation value for each module. Then, for each climbing cage monitoring module, its deviation value is multiplied by the weight elements of each module in the row of the spatial weight matrix, and all products are summed to obtain the spatial lag deviation value for that module. Then, the spatial lag deviation values ​​for all modules are accumulated to obtain the sum of spatial covariances. Then, the sum of squares of the deviation values ​​of all modules is calculated. Finally, the sum of spatial covariances is divided by the sum of squares to obtain the spatial aggregation global index.

5. The intelligent monitoring system for the crab farming environment according to claim 4, characterized in that, Methods for obtaining local clustering indices include: For each cage-climbing monitoring module, firstly, its set of neighboring modules is determined based on a spatial distance threshold, which is related to the maximum activity radius. Then, the spatial weight between the module and each neighboring module is calculated using a biosocial coupling weighting function. Next, the neighbor-weighted average of the corrected cage-climbing frequency of each neighboring module is calculated with the spatial weight as the coefficient. Then, the first difference between the corrected cage-climbing frequency of the module and the neighbor-weighted average is calculated. Then, the second difference between the corrected cage-climbing frequency of the module and the global average is calculated. Then, the first difference and the second difference are multiplied to obtain the initial local clustering value of the module. Then, the standard deviation of the initial local clustering value of all cage-climbing monitoring modules is calculated. Finally, the initial local clustering value of the module is divided by the standard deviation to obtain the local clustering index of the module.

6. The intelligent monitoring system for the crab farming environment according to claim 5, characterized in that, The biosocial coupling weight function uses a three-factor product to determine the spatial weight between any two climbing cage monitoring modules: The biological social state vector is calculated based on the weight and image data of each climbing cage monitoring module. This biological social state vector consists of three dimensions: the first dimension is the average weight, which is calculated by summing the weight data of all climbing cage events of the module within a unit of time and dividing it by the total number of mitten crabs obtained from image recognition; the second dimension is the actual occurrence frequency, which is calculated by summing the actual number of mitten crabs in each climbing cage event obtained from image recognition; and the third dimension is the population density coefficient, which is calculated by the proportion of the number of events in each climbing cage event obtained from image recognition where the number of mitten crabs appearing at the same time exceeds a preset threshold to the total number of events of the module. The basic state influence factor is calculated based on the biological-social state vectors of the two climbing cage monitoring modules. This basic state influence factor is the dot product of the biological-social state vectors of the two modules divided by the product of the magnitudes of the biological-social state vectors of the two modules. The spatial distance attenuation factor is calculated based on the spatial distance between the two climbing cage monitoring modules. This spatial distance attenuation factor is determined by the spatial distance between the two climbing cage monitoring modules through a negative exponential attenuation function. The attenuation coefficient of this negative exponential attenuation function is set according to the maximum activity radius of the mitten crab. The interaction sensing factor is calculated based on the bio-social state vectors of the two climbing cage monitoring modules. This interaction sensing factor is the magnitude of the difference between the bio-social state vectors of the two modules divided by the maximum magnitude of the bio-social state vectors of the two modules. The spatial weight between the two climbing cage monitoring modules is obtained by multiplying the basic state influence factor, spatial distance attenuation factor, and interactive perception factor.

7. The intelligent monitoring system for the crab farming environment according to claim 6, characterized in that, The method for generating spatial clustering feature values ​​is as follows: The local clustering indices of all climbing cage monitoring modules are normalized so that the value range of each local clustering index is mapped to the same numerical range as the global spatial clustering index. Then, the mean and standard deviation of the normalized local clustering indices are calculated. The sum of the mean and standard deviation is then divided by two to obtain the local comprehensive index. Finally, the global spatial clustering index is multiplied by the local comprehensive index to obtain the spatial clustering characteristic value.

8. The intelligent monitoring system for the crab farming environment according to claim 7, characterized in that, The deviation index is calculated as follows: Calculate the absolute value of the difference between the current spatial clustering characteristic value and the corresponding unit time period benchmark value, then calculate the ratio of the absolute value of the difference to the corresponding unit time period benchmark value, and then multiply the ratio by a preset proportional coefficient to obtain the deviation index.

9. The intelligent monitoring system for the crab farming environment according to claim 8, characterized in that, Based on the magnitude of the deviation index, control devices in different spatial areas within the aquaculture water area are selectively activated, specifically including: When the deviation index is less than or equal to the preset first threshold, the control execution module does not activate any control equipment; When the deviation index is greater than the preset first threshold and less than or equal to the preset second threshold, the control execution module identifies the area where the climbing cage monitoring module is located and the local aggregation index is lower than the preset low threshold, and starts the local control device for that area. When the deviation index is greater than the preset second threshold, the control execution module activates the full-area control equipment covering the entire aquaculture water area.

Images