Crocodile stress intelligent evaluation system based on behavior characteristics

Abstract

The invention relates to the technical field of behavior recognition, in particular to a crocodile stress intelligent evaluation system based on behavior characteristics, which comprises a time sequence behavior collection module, an interval dynamic analysis module, a feeding fluctuation collaboration module, a trend grading judgment module and a stress alarm linkage module. According to the method, through multi-dimensional structured recombination of behavior characteristic data, deep cooperation of a feeding response time sequence, environment change and a behavior link is achieved, unified archiving of the total feeding amount, interval change and trend classification is achieved, the accuracy of stress state judgment is improved, links such as periodic behavior comparison, fluctuation trend classification and label output penetrate through the whole process, and the accuracy of stress state judgment is improved. Whole-process tracing and fine anomaly analysis of a crocodile stress link are supported, coherence and consistency of evaluation output are guaranteed through automatically generated trend classification and grade labels, isolated judgment limitation is eliminated through data intercommunication, integrated triggering of grading, alarming and displaying is achieved, and the intelligent stress evaluation requirement for continuous period and accurate quantification under dynamic breeding management is met.

Description

Technical Field

[0001] This invention relates to the field of behavior recognition technology, and in particular to a crocodile stress intelligent assessment system based on behavioral characteristics. Background Technology

[0002] The field of behavior recognition involves the collection, analysis, and identification of behavioral characteristics such as actions, postures, and trajectories of individuals in specific environments. This includes acquiring behavioral characteristic data, modeling and classifying behavioral patterns, and identifying and assessing behavioral states. It is widely applied in security monitoring, animal behavior research, human-computer interaction, and physiological state monitoring. Among these, the traditional crocodile stress intelligent assessment system refers to a system that identifies crocodile behavior and judges stress states based on manual observation or simple monitoring equipment. Typically, it involves manual recording by keepers of the crocodile's movement frequency, body posture, and activity range, or using fixed video equipment to capture video of the crocodile's behavior, combining visual analysis or playback to determine whether the crocodile has a stress response.

[0003] Existing technologies rely on manual recording and fixed-point video observation. The scope of data collection on crocodile feeding behavior is limited by the monitoring angle and human intervention. It is impossible to summarize the behavioral data over multiple periods in a time series. Environmental monitoring and behavioral recording are independent of each other, making it impossible to achieve precise linkage between data. There is a lack of systematic judgment on continuous feeding changes and response delays. When faced with environmental fluctuations, it is impossible to track the overall trend of behavioral changes. It has technical shortcomings such as insufficient stress level discrimination, strong subjectivity of grading standards, and difficulty in timely identification of subtle anomalies within a period. Stress assessment is prone to omissions or misjudgments and is difficult to adapt to the actual needs of continuous, precise, and data-driven identification in dynamic aquaculture scenarios. Summary of the Invention

[0004] To address the technical problems existing in the prior art, this invention provides a crocodile stress intelligent assessment system based on behavioral characteristics. The technical solution is as follows: On the one hand, a crocodile stress intelligent assessment system based on behavioral characteristics is provided, including: The temporal behavior aggregation module extracts the time of the first feeding action collected by the underwater infrared camera based on the data recorded by the feeding device, numbers and pairs the two sets of time signals in sequence, filters the environmental monitoring data of the temperature and humidity sensor, and aggregates various parameters to obtain the behavior temporal association set. Based on the behavioral time-series association set, the interval dynamic analysis module compares the number sequence, extracts the differences in behavioral response intervals between adjacent weeks, organizes the interval differences in order, and obtains the change in response interval. The feeding fluctuation coordination module calculates the amount of food released in the current cycle based on the change in the response interval, combines it with the remaining amount to obtain the actual feeding amount, compares it with the total feeding amount in the previous cycle, analyzes the fluctuation amplitude, and obtains the feeding behavior offset. The trend classification judgment module determines the trend category change based on the feeding behavior offset and the segmented recognition baseline, compares the trend category with the baseline order, and issues a state change signal when a change occurs to obtain the stress trend classification group. The stress alarm linkage module analyzes the content of the graded tags based on the stress trend classification group, compares the tag order, and emits an audio-visual linkage signal when the tags match. The graded tag information is synchronized to the remote terminal to obtain the stress linkage trigger result.

[0005] On the other hand, the behavioral temporal association set includes a behavioral time structure set, an action trigger index group, and an environmental association mapping cluster; the response interval change includes a periodic rhythm feature group, a response gradient sequence, and a behavioral fluctuation trajectory slice; the feeding behavior offset includes a feeding trend quadrant, a behavioral difference factor group, and a dynamic offset surface; the stress trend classification group includes a trend level spectrum, a state evolution chain, and segmented classification data; and the stress linkage triggering result includes an alarm trigger slice, a linkage control node, and a level output box.

[0006] On the other hand, the time-series behavior collection module includes: The data time stamp numbering submodule analyzes the feeding start time and the first feeding action time obtained by the underwater infrared camera based on the data recorded by the feeding device, compares the order of the two types of time data, calculates the number index corresponding to the time, and obtains the time sequence behavior mark set. The behavior triggering pairing submodule filters the pairing information corresponding to the number sequence based on the time sequence behavior tag set, judges the continuity of the number to exclude discontinuous items, analyzes the environmental data obtained by the temperature and humidity sensor, and matches according to the pairing time number to obtain the structure corresponding to the number behavior. The environment mapping and fusion submodule calculates the correspondence between each behavior event and the environmental data based on the numbered behavior correspondence structure, summarizes the time points and action sequences associated with feeding behaviors, optimizes the synchronization method between environmental data and behavior numbers, and obtains a behavior temporal association set.

[0007] On the other hand, the interval dynamic analysis module includes: The behavior cycle segmentation submodule, based on the behavior time sequence association set, determines the continuity between each number, compares the relative changes between the feeding start time and the feeding action time, groups and classifies the behavior cycle data in sequence, optimizes the number mapping relationship of the cycle data, and obtains the behavior response cycle set. The response time difference calculation submodule calculates the time interval between feeding initiation and feeding action within each set of behavior response cycles based on the set of behavior response cycles, compares the change range of time intervals in adjacent cycles, determines the increasing or decreasing trend of time intervals in the cycle sequence, summarizes the change content of intervals between each cycle, and obtains the cycle interval change group. The periodic difference processing submodule analyzes the time interval changes of consecutive numbered periods based on the periodic interval change group, filters representative interval changes in the periodic sequence, integrates the number and interval change data, and obtains the response interval change amount.

[0008] On the other hand, the feeding fluctuation coordination module includes: The bait release monitoring submodule determines whether the bait release data recorded by the intelligent feeding device is consistent with the cycle number based on the change in the response interval. It compares the bait falling segment with the device record to determine the validity of the release action. It calls the camera image to identify the appearance area of ​​the remaining bait and filters out environmental interference. It calculates the difference between the released quantity and the remaining quantity to obtain the cycle feeding quantity. The feeding difference calculation submodule calculates the difference between the current feeding quantity and the previous feeding quantity based on the feeding quantity of the cycle, compares the directional changes of the difference between adjacent cycles, analyzes the correspondence between the changes in feeding quantity and the cycle relationship in the numbering sequence, judges the trend of increase or decrease in feeding quantity in consecutive cycles, and obtains the feeding fluctuation sequence. The behavior state discrimination submodule determines the continuity of the feeding change trend in each cycle based on the feeding fluctuation sequence, analyzes the relationship between the change trend and the cycle number, compares the consistency between the current cycle change trend and the previous cycle trend, adjusts the behavior category label of the corresponding number, and eliminates the conflict between the previous and previous cycles to obtain the feeding behavior offset.

[0009] On the other hand, the trend classification judgment module includes: The segmented baseline comparison submodule determines the corresponding order between each cycle offset and the segmented baseline based on the feeding behavior offset, compares the matching of offset and baseline under the cycle sequence, identifies cycle segments where the baseline order and offset are inconsistent, and obtains trend category comparison information. The trend category determination submodule determines whether the trend category has changed in each period based on the trend category comparison information, analyzes the evolution of the trend category label in the period sequence, compares the switching of category labels in adjacent periods, marks the category change nodes, and obtains the trend category change sequence. The trend classification and organization submodule analyzes the category change pattern in continuous cycles based on the trend category change sequence, determines the consistency interval of trend categories within the cycle sequence, optimizes the mapping method between category distribution and cycle sequence, and obtains stress trend classification groups.

[0010] On the other hand, the stress alarm linkage module includes: The hierarchical label discrimination submodule, based on the stress trend classification group, determines the order relationship between each periodic hierarchical label and the set label, compares the matching status of the periodic label content with the warning standard, identifies periodic segments with consistent order and content, and obtains label order matching information. Based on the tag sequence matching information, the linkage signal output submodule analyzes the relationship between the graded tags and the linkage control standard, compares whether the tag content meets the requirements for sound and light linkage signal output, identifies the cycle corresponding to the tag matching state, adjusts the linkage signal output command, and obtains the sound and light alarm linkage sequence. The tag remote synchronization submodule, based on the sound and light alarm linkage sequence, determines the synchronization requirements of the tag between the local and remote terminals, compares the synchronization status of the tag content, optimizes the information transmission and display of the hierarchical tags between terminals, outputs remote synchronization feedback, and obtains the stress linkage triggering result.

[0011] On the other hand, the first feeding action time refers to the time point when the crocodile first attempts to feed is detected in the monitoring video stream, which is obtained by underwater infrared camera equipment combined with action recognition. The time signal number refers to assigning a number to each group of feeding start and feeding action times in sequence.

[0012] On the other hand, the number sequence refers to the number of each group of feeding initiation to feeding action arranged in chronological order, the behavior response interval refers to the actual time difference between feeding initiation and the first feeding action within each cycle, and the interval difference refers to the change value of response time between adjacent cycles.

[0013] On the other hand, the amount of feed released refers to the actual weight or quantity of feed released by the intelligent feeding device in the current cycle, and the remaining quantity refers to the weight or quantity of uneaten feed detected by the camera or sensor device after the end of this feeding cycle.

[0014] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: By reorganizing behavioral characteristic data in a multidimensional structure, deep collaboration between feeding response timing, environmental changes, and behavioral links is achieved. Unified archiving of total food intake, interval changes, and trend classification improves the accuracy of stress state identification. Periodic behavior comparison, fluctuation trend classification, and tag output are integrated throughout the entire process, supporting full-process tracing of crocodile stress links and fine-grained anomaly analysis. Automatically generated trend classifications and level tags ensure the consistency of assessment output. Data interoperability eliminates the limitations of isolated judgments. Classification, alarms, and display are triggered in an integrated manner, meeting the needs of continuous and quantitatively accurate intelligent stress assessment under dynamic aquaculture management. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the system of the present invention; Figure 2 This is a schematic diagram of the system framework of the present invention; Figure 3 This is a flowchart of the timing behavior collection module of the present invention; Figure 4 This is a flowchart of the interval dynamic analysis module of the present invention; Figure 5 This is a flowchart of the feeding fluctuation coordination module of the present invention; Figure 6 This is a flowchart of the trend classification judgment module of the present invention; Figure 7 This is a flowchart of the stress alarm linkage module of the present invention. Detailed Implementation

[0017] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0018] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0019] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0020] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0021] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0022] This invention provides a crocodile stress intelligent assessment system based on behavioral characteristics, such as... Figure 1 As shown, the system includes: The temporal behavior collection module analyzes the feeding start time based on the data recorded by the feeding device. By reading the time of the first feeding action generated by the underwater infrared camera, the two sets of time signals are numbered and paired sequentially according to the time sequence. The environmental data continuously monitored by the temperature and humidity sensor is filtered, and all time tags are normalized and matched. The behavioral parameters are associated according to the arrangement order of each set of data to obtain the behavioral temporal association set. The interval dynamic analysis module is based on the behavioral time series association set. It compares the number sequence in the behavioral response cycle, extracts the difference between the behavioral response intervals corresponding to the second group and the first group of numbers, and then processes the third group and the second group of numbers in the same way. The extracted interval differences are sorted by period to obtain the change in response interval. The feeding fluctuation coordination module calculates the amount of food released in the current cycle based on the change in response interval, combines the remaining amount identified by video monitoring, processes the difference between the two, and judges the fluctuation amplitude by referring to the actual total amount of food consumed in the previous cycle, adjusts the behavior status category, and obtains the feeding behavior offset. The trend classification judgment module is based on the feeding behavior offset, according to the segmented identification baseline, analyzes the trend category changes, compares the trend category with the baseline order, and issues a stress state change signal when the trend category changes. When the same trend category appears continuously, the fluctuation amplitude and the difference from the baseline order are sorted out to obtain the stress trend classification group. The stress alarm linkage module analyzes the associated graded tags based on stress trend classification groups, compares the order of the two groups of tags with reference to the content of the graded tags, and issues an audible and visual linkage signal when the tag content is consistent. Then, the graded tag information is synchronized to the remote terminal for display to obtain the stress linkage trigger result.

[0023] The behavioral temporal association set includes a behavioral time structure set, an action trigger index group, and an environmental association mapping cluster. The response interval change includes a periodic rhythm feature group, a response gradient sequence, and a behavioral fluctuation trajectory slice. The feeding behavior offset includes a feeding trend quadrant, a behavioral difference factor group, and a dynamic offset surface. The stress trend classification group includes a trend level spectrum, a state evolution chain, and segmented classification data. The stress linkage triggering results include alarm trigger slices, linkage control nodes, and level output boxes.

[0024] In the temporal behavior collection module, the feeding start time refers to the precise moment when the intelligent feeding device begins to release feed each time, derived from the timestamp of the feeding control device; the first feeding action time refers to the time point when the crocodile first attempts to feed is detected in the monitoring video stream, obtained by underwater infrared camera equipment combined with action recognition algorithm; the time signal number refers to assigning a number to each set of feeding start and feeding action times in sequence, so that they can correspond and be paired, facilitating subsequent time series analysis; the continuously monitored environmental data refers to the raw data stream of pond environmental temperature and humidity continuously recorded by temperature and humidity sensors, used for synchronous analysis of behavior and environment; the time tag refers to the standardized time mark assigned to all behavior and environmental data records, ensuring that the content collected from different data sources can be strictly aligned; the sorting order refers to sorting the above types of data in chronological order to achieve one-to-one correspondence, preparing for subsequent periodic behavior analysis and parameter extraction; the behavior parameters refer to various quantitative information representing the behavioral characteristics of crocodiles, such as the time and duration of the feeding action.

[0025] In the interval dynamic analysis module, the behavioral response cycle refers to the time period of a complete behavioral process formed between the feeding and ingestion actions, serving as the basic unit of analysis; the number sequence refers to the numbering of each group of "feeding initiation-feeding action" arranged in chronological order, facilitating one-to-one correspondence and comparison between different periods; the behavioral response interval refers to the actual time difference between the feeding initiation and the first feeding action within each period, used to characterize the response speed; difference extraction refers to analyzing interval changes by comparing the behavioral response intervals of adjacent periods to identify any abnormalities in behavioral responses; identical step processing refers to processing all adjacent numbers sequentially using the same difference extraction method to ensure data analysis consistency; interval difference refers to the change in response time between adjacent periods, reflecting the fluctuation trend of crocodile behavior; and period consolidation refers to integrating all difference results in periodic order for subsequent rate and trend analysis.

[0026] In the feeding fluctuation coordination module, the feed release quantity refers to the actual weight or quantity of feed released by the intelligent feeding device in the current cycle, which is automatically measured and recorded by the feeding equipment; the remaining quantity refers to the weight or quantity of uneaten feed detected by the camera or sensor device after the end of this feeding cycle; the difference between the two refers to subtracting the remaining quantity from the feed release quantity to obtain the total amount of feed actually consumed by the crocodiles, which is used as the feed intake input for subsequent judgment; the fluctuation amplitude refers to the magnitude of the change between the actual feed intake in the current cycle and the previous cycle, used to judge the stability or abnormality of the feeding state; the behavior state category refers to classifying the behavior results of each cycle according to the changes in feed intake and response interval, such as "normal", "decreased", "recovered", etc., used for trend judgment.

[0027] In the trend classification judgment module, the segmentation identification baseline refers to the reference line or standard set for dividing each behavioral trend, such as separating behavioral change intervals at different stress levels; trend category change refers to the change in category label after continuous analysis of behavioral data, such as changing from "stable" to "declining" or "rising"; baseline order refers to the order in which behavioral trend category changes and identification baselines are arranged, used to determine whether the category has crossed the boundary; trend category change refers to the detection of an event in which the behavioral state category changes from one trend to another, such as a continuous decrease in food intake; stress state change signal refers to the control signal automatically generated by the system to mark the change in stress level when the behavioral trend category change reaches a certain criterion; continuous occurrence of the same trend category means that the behavioral trend category is consistent in multiple periods, indicating that the behavioral state has remained unchanged, used to help determine whether the level needs to be maintained or updated.

[0028] In the stress alarm linkage module, the graded label content refers to the unique mark or identifier assigned to different stress levels, serving as the basis for control and alarm judgment; the order of the two sets of labels refers to the sequence or correspondence between the current stress level label and the system warning setting label, determining whether to trigger an alarm; label content consistency refers to judging whether the current status label and the alarm setting label are the same, if they are the same, it means that the preset stress warning threshold has been reached; the audible and visual linkage signal refers to the start signal sent by the control system to the audible and visual alarm device, which executes the alarm action when a stress level warning occurs; the graded label information refers to the identifier or data related to the stress level, which is displayed synchronously in linkage and remote monitoring; the remote terminal refers to a visualization platform or mobile device outside the main control center, which receives and displays alarm information and stress level status in real time.

[0029] like Figure 2 and Figure 3 As shown, the time-series behavior aggregation module includes: The data time stamp numbering submodule analyzes the feeding start time and the first feeding action time obtained by the underwater infrared camera based on the data recorded by the feeding device, compares the order of the two types of time data, calculates the number index corresponding to the time, and obtains the time sequence behavior mark set. The timestamps of each feeding cycle are analyzed to extract the start time information of each bait release. For example, if a feeding is performed at 8:30:05 AM, after obtaining this information, the raw video data collected by the underwater infrared camera is retrieved. Action recognition is performed on each frame to identify whether there are any features of a crocodile opening its mouth to approach the bait. By comparing the continuity between frames, the specific time point of the first feeding behavior is found. For example, if the feeding action is identified as occurring at 8:30:47 AM, then the sequence of these two time points is manually compared to clarify the timing. Feeding behavior occurs after feeding. Each pair of "feeding time - ingestion time" is then packaged and labeled with a number in chronological order, for example, starting from T01 and incrementing sequentially. The numbers are arranged in chronological order to establish a unique identifier for tracking subsequent behavioral events. Then, the number and time group are combined to form a standardized behavior record structure. Each data point in the structure includes the number, feeding time, ingestion time, and the time difference between the two. The above process is performed once for each feeding. After accumulation, a set of behavior sequences arranged in chronological order is generated, constructing a complete set of temporal behavior tags.

[0030] The behavior-triggered pairing submodule filters pairing information corresponding to the number sequence based on the time-series behavior tag set, judges the continuity of the number to exclude discontinuous items, analyzes the environmental data obtained by the temperature and humidity sensor, and matches according to the pairing time number to obtain the structure corresponding to the number behavior. Extract two consecutive sets of numbers and compare their differences. If the difference between two adjacent numbers is 1, it indicates that the numbers are consecutive, and the pairing information is retained. If the difference is not 1, such as a skipping number (e.g., jumping directly from T02 to T04), it indicates that the data for that round is missing or the recording failed, and it is not included in the valid behavior sequence. Subsequent processing only performs analysis on consecutive numbered behaviors. Then, for the retained pairing numbers, extract the time period between feeding and ingestion as the behavior time range. Read the raw temperature and humidity data stream from the environmental monitoring equipment and filter all temperature and humidity sampling points within this time period. For example, if there are 4 environmental sampling points between feeding and ingestion corresponding to a certain number, and the time corresponding to the points is within the behavior time period, the data is extracted and bound to the behavior number to form a new structure. Each record is composed of a number identifier and all environmental parameters within the time period of that number, completing the synchronous association between the behavior number and environmental data. This structure facilitates subsequent joint analysis of behavior and environment.

[0031] The environment mapping and fusion submodule calculates the correspondence between each behavioral event and environmental data based on the numbered behavior correspondence structure, summarizes the time points and action sequences associated with feeding behaviors, optimizes the synchronization method between environmental data and behavior numbers, and obtains a behavioral temporal association set. The behavior time period and environmental data records corresponding to each number are extracted sequentially. First, the specific time point of each behavior event is located, such as the start time of the feeding behavior and the end time of the feeding action. Based on this, the temperature and humidity records collected during the behavior period are matched, and the environmental data with the closest time to the behavior event is extracted as the environmental value corresponding to the behavior. Then, the temperature and humidity data points are extracted one by one according to the number, and each action is paired with its environmental value. For example, in number T01, the nearest environmental record point corresponding to the feeding action time point is 8:30:40 am, with a temperature of 29.1 degrees and a humidity of 81%. This value is then bound to the feeding behavior. Subsequently, the temperature and humidity record sequence covered by this number is extracted to observe whether the temperature and humidity have continuously increased, decreased, or fluctuated. By comparing the changes in the values ​​before and after, the trend of change is judged. Then, all environmental fluctuation trends and behavior data are integrated according to the number and uniformly compiled into a structured record based on the number. This makes each behavior event not only associated with the action time, but also simultaneously mapped to the environmental change state during the behavior period, forming a behavior time sequence association set with multi-dimensional information fusion of time, behavior, and environment.

[0032] like Figure 2 and Figure 4 As shown, the interval dynamic analysis module includes: The behavior cycle segmentation submodule determines the continuity between each number based on the behavior time sequence association set, compares the relative changes between the feeding start time and the feeding action time, groups and classifies the behavior cycle data in sequence, optimizes the number mapping relationship of the cycle data, and obtains the behavior response cycle set. The system retrieves the behavioral data corresponding to each group of numbers from the time-series association set, extracts the feeding start time and the first feeding action time, and then arranges all numbers in chronological order. It compares the time interval differences between adjacent numbers to determine if the numbers are continuous in the behavioral cycle division logic. When the feeding time of a certain number is greater than the feeding time of the previous number by a set time difference threshold, it is marked as a cycle interval breakpoint. For example, if the threshold is set to 600 seconds, and the feeding time of number T03 is 610 seconds away from the feeding time of number T02, it is considered the boundary between two independent cycles. Then, the continuous number segments are periodically packaged, forming structures like T01 to T02, T03 to T04, etc. Grouping is performed to form preliminary periodic division groups. Then, the time difference between feeding and ingestion in each group's period is compared to see if it shows an increasing, decreasing, or constant pattern. By comparing the time difference between two behavioral events in each group, for example, if the time difference for T01 is 40 seconds and for T02 is 45 seconds, then it shows an increasing trend. The trend of each period is marked. When further optimizing the numbering mapping relationship, the period number is set independently of the behavior number. For example, the first behavior period number is set as C01 and the second period is set as C02. The period number is used to track changes in periodic characteristics. After arranging the above-divided period numbers, corresponding behavior numbers, and time difference trends one by one, the output is a structured period set, which yields the behavioral response period set.

[0033] The response time difference calculation submodule is based on the behavioral response cycle set. It calculates the time interval between feeding initiation and feeding action in each cycle, compares the change of time interval in adjacent cycles, judges the increasing or decreasing trend of time interval in the cycle sequence, summarizes the interval change content between each cycle, and obtains the cycle interval change group. The feeding time and ingestion time within each cycle are paired one by one to extract the time point information from the structure. Then, the time interval value is obtained by subtraction. For example, in cycle C01, feeding occurs at 08:00:00 and ingestion occurs at 08:00:40, so the time interval is 40 seconds. This interval is recorded as the response time interval of C01. The corresponding time points under all cycle numbers are processed in sequence to generate a set of response time value sequences corresponding to cycle numbers. Then, the difference between the response times of two adjacent cycles is calculated to extract the change amplitude. For example, if the response time of C02 is 50 seconds, then the change in response time of C02 relative to C01 is 10 seconds. The absolute value difference is used to judge the change amplitude. When judging the increase or decrease trend, the sign of the difference is used to identify it. For example, if the change value is positive, it is judged that the response time has increased. If it is negative, it is judged that the response time has decreased. The number and amplitude of increase and decrease changes in the entire cycle sequence are further counted to prepare data for subsequent trend classification. The response interval difference between every two cycles is bound to the trend label according to the cycle number sequence and output to form a cycle interval change group.

[0034] The periodic difference sorting submodule analyzes the time interval changes of consecutive numbered periods based on the periodic interval change group, filters representative interval changes in the period sequence, integrates the number and interval change data, and obtains the response interval change amount. The interval change data between every two consecutive periods is read sequentially according to the numbering order. The change value in each numbered sequence and its corresponding period position are extracted. When analyzing the change trend, the fluctuation range of the time interval change is judged first. When the absolute value of a certain change value exceeds the preset representative difference threshold, the change is marked as a representative interval change. The threshold is set with reference to the average change range and standard deviation of the past period series. For example, when the change range is greater than 20 seconds, it is considered to be a representative change. Then, all period numbers marked as representative changes are filtered and retained. Their period numbers and time difference change values ​​are paired to form records. For example, the change range of period C03 is 25 seconds, and the record is [C03, 25]. All such records form a group of representative interval change items. During the data integration process, they are arranged in ascending order of period number to ensure that the subsequent trend tracking order is not confused and to obtain the response interval change amount.

[0035] like Figure 2 and Figure 5 As shown, the feeding fluctuation coordination module includes: The bait release monitoring submodule determines whether the bait release data recorded by the intelligent feeding device is consistent with the cycle number based on the change in response interval. It compares the bait falling segment with the device record to determine the validity of the release action. It calls up the camera image to identify the appearance area of ​​the remaining bait and filters out environmental interference. It calculates the difference between the released quantity and the remaining quantity to obtain the cycle feeding quantity. The system retrieves the index data group that links the cycle number to the feeding behavior, extracts the actual feeding time point and the feeding response time difference associated with each number, and then reads the timestamp, quantity record field, and number identifier for each release on that day from the bait release data recorded by the intelligent feeding device. It compares the device number with the behavior cycle number to see if they match. If the number field is missing or the correspondence is incorrect, it is recorded as a mismatch and the cycle is removed. Next, it compares the difference between the feeding time marked in the behavior cycle number and the time field recorded by the device release to see if it is within a set reasonable offset range. The allowable offset is set to no more than 5 seconds. If the difference between the actual release time and the time corresponding to the number exceeds this threshold, the release record is deemed invalid. Finally, it retrieves frame segments related to the feeding cycle from the infrared camera image to identify the falling bait particles. The trajectory analysis performs continuous frame analysis on the falling area to identify effective bait release actions. It also excludes images of background water flow changes and bait similar to the bait within the camera area. The combined conditions of color, shape, and motion trajectory are set as filtering conditions to eliminate false identification items such as floating objects and reflections. Then, it identifies the image area of ​​uneaten bait residue in still frames. The number of remaining baits is calculated based on the unit size of the bait and the pixel ratio of the area in the image. For example, if the identified residual bait occupies a pixel area of ​​200 square pixels, and each 10 square pixels represents one bait unit, the number of remaining baits in the current cycle is 20. The number of releases recorded by the device is 35. The difference between the two is the number of baits ingested in the current cycle, which is 15. This value is bound to the cycle number as the cycle feeding count.

[0036] The feeding difference calculation submodule calculates the difference between the current feeding quantity and the previous feeding quantity based on the periodic feeding quantity, compares the directional changes of the quantity difference between adjacent periods, analyzes the correspondence between the changes in feeding quantity and the periodic relationship in the numbering sequence, judges the trend of increase or decrease in feeding quantity in continuous periods, and obtains the feeding fluctuation sequence. The process involves sequentially calling the data structure that binds the period number to the corresponding food intake quantity. The food intake quantity values ​​of two adjacent period numbers are extracted and subtracted to obtain the difference in food intake over two weeks. For example, if the quantity corresponding to period number C05 is 18 and C06 is 15, the current difference is -3. This difference is recorded, indicating a decreasing trend in food intake. The same operation is then performed on the next period number C07, and the difference is compared with C06 to obtain +4, indicating an increasing trend in food intake. When comparing the direction of the difference between the two periods, the signs of the differences are compared. If the signs of the differences are opposite, they are marked as... If the trend reverses, and the same trend continues, it is marked as a continuation of the trend. A trend status record table is then established in the numbering order, with fields including the current number, feeding quantity, difference from the previous period, direction of change, and trend status label. It is then determined whether the trend status has appeared more than twice consecutively. When the same trend status lasts for three consecutive periods, it is marked as a stable trend. For example, if C05 to C07 are all decreasing trends, the output trend status is "continuous decline". At the same time, the trend direction and period range are recorded. The above items are combined in the order of numbering into a trend sequence structure to obtain the feeding fluctuation sequence.

[0037] The behavior state discrimination submodule is based on the feeding fluctuation sequence to determine the continuity of the feeding change trend in each cycle, analyze the relationship between the change trend and the cycle number, compare the consistency between the current cycle change trend and the previous cycle trend, adjust the behavior category label of the corresponding number, and eliminate the conflict between the previous and previous cycles to obtain the feeding behavior offset. To compare the consistency between the current cycle trend and the previous cycle trend, the following formula is used: ; Calculate the trend consistency deviation value and eliminate conflicts between previous and subsequent cycles to obtain the feeding behavior offset; in, Representing the Cyclical trend consistency deviation value Representing the The first in the cycle Parameters on the trend of food intake changes Representing the The first in the cycle Parameters on the trend of food intake changes Representing the The sum of parameters related to the periodic feeding trend. Representing the The sum of parameters related to the periodic feeding trend. The number of trend parameters; Trend consistency deviation is a numerical indicator that measures the degree of overall consistency or difference between the current cycle and the previous cycle in terms of feeding trend. It reflects the overall magnitude and difference of the feeding trend parameters in two consecutive cycles and is used to intuitively indicate whether the crocodile's behavioral trend is continuous, fluctuating, or has changed significantly. It represents the sum of absolute changes in each feeding trend parameter (such as changes in food intake per unit time, changes in feeding response time, etc.) between the current period and the previous period. The greater the change, the larger the numerator. The sum of the absolute values ​​of all trend parameters in the previous period (serving as a benchmark for the strength of the behavioral trend) is compared with the sum of the overall trends over the two periods, and is used to normalize and amplify actual fluctuations. The trend consistency deviation value is a quantitative indicator used to indicate whether there is "similarity" or "deviation" between the crocodile feeding behavior trends of two consecutive periods, and is directly used for subsequent behavior category labeling and the discrimination and adjustment of conflicts between preceding and following periods.

[0038] Retrieve cycle in sequence With period Each trend parameter value and Three parameters were selected for each cycle: total feed intake (kg), the response interval from feeding initiation to the first feeding (s), and the number of feeding actions detected within the cycle (times). These three data points were extracted using an automatic feed meter and an infrared camera monitoring and behavior recognition system, respectively. All trend parameters were standardized using an extremum normalization method. The maximum and minimum values ​​for each parameter were extracted from monitoring data over 20 consecutive cycles as the normalization benchmark. The extreme range of feed intake was set as follows: The response interval is The number of feedings is Taking the crocodile numbered G012 in the rearing tank as an example, its cycle The amount of food consumed is The response interval is The number of feedings is Next, in the cycle The amount of food consumed is The response interval is The number of feedings is After normalization, the standard values ​​of the three trend parameters are as follows: , ; , ; , ; According to the formula, where The absolute values ​​of the differences between the three sets of trend parameters in the numerator are as follows: ; ; ; The sum of the three items is The sum of the absolute values ​​of the three periodic parameters in the denominator is: ; The trend sum difference is: ; Therefore, the complete denominator is: ; Substituting into the formula, we get: ; Trend consistency deviation value value space It is divided into three judgment intervals, each corresponding to a specific trend state category, used to characterize the degree of change in behavioral trend. The intervals are defined as follows: when When the behavior trend changes very little, it is judged as a "stable segment". In this state, the behavior data of the current cycle is basically the same as that of the previous cycle and does not constitute a key node in the behavior evolution chain. when When the time indicates a slight fluctuation in the behavioral trend, it is judged as a "transition zone". The current cycle shows initial signs of behavioral change, but does not yet have significant offset characteristics. An offset node is formed only when the zone appears twice or more consecutively. when When this occurs, it indicates a significant change in behavioral trend, which is identified as a "mutation segment." There is a significant inconsistency between the behavioral response in the current cycle and that in the previous cycle, which constitutes the basis for the establishment of a behavioral deviation event.

[0039] The calculated deviation value is This value is located at Within the "mutation zone," it indicates that the behavioral trend of the current cycle has significantly differed from the previous cycle, belonging to a mutation state, and thus necessitates inclusion in the trend evolution chain. This result represents a dramatic change in the co-directional direction of the feeding behavior in the current cycle across three dimensions (feeding amount, response interval, and feeding frequency). When this value is detected falling into the mutation zone, conflict elimination logic is immediately executed to determine the consistency of trend direction. If the three trend changes are consistent, it is considered a "conflict-free mutation." The corresponding current period is marked as a valid behavior offset node.

[0040] like Figure 2 and Figure 6 As shown, the trend classification judgment module includes: The segmented baseline comparison submodule determines the corresponding order between the offset of each cycle and the segmented baseline based on the feeding behavior offset, compares the matching of the offset and the baseline under the cycle sequence, identifies cycle segments where the baseline order and the offset are inconsistent, and obtains trend category comparison information. The calculated feeding offset value for each behavioral cycle is extracted and matched sequentially with preset segmented baseline values. The segmented baselines are pre-defined into three intervals: a stable interval with an absolute offset value between 0 and 2, a fluctuating interval with an absolute offset value between 2 and 5, and a stress interval with an absolute offset value greater than 5. When the offset for a given cycle is 3.6, it is classified into the fluctuating interval. The cycle number is then paired with the segmented interval number and recorded. Subsequently, the position order of the segmented baseline corresponding to each number in the cycle sequence is listed and compared with the actual position order of the offset changes. For example, if the offset order is 1.2, 3... 6 and 6.3 should be in the stable, fluctuating, and stress intervals, respectively. However, if the second offset is classified into the stable interval during actual detection, it is considered a classification error. When performing this type of comparison, each period number is used as an index to compare the matching status between its segment classification result and the theoretical segment range where its actual offset value is located. If they match, they are marked as "matched"; if they do not match, they are marked as "mismatched". Then, the period numbers of all "mismatched" items are summarized to form a trend deviation segment set. The output fields include five items: period number, actual offset value, theoretical segment, current segment, and matching status. This set is used to construct trend category comparison information.

[0041] The trend category determination submodule determines whether the trend category has changed in each period based on the trend category comparison information, analyzes the evolution of the trend category label in the period sequence, compares the switching of category labels in adjacent periods, marks the category change nodes, and obtains the trend category change sequence. Read the offset classification label corresponding to each period sequentially, extract and label the trend category to which the current period belongs. For example, if a certain offset number is classified as a stress interval, the corresponding trend category is "rising". If it is classified as stable, it is "maintaining". Establish a trend category sequence table accordingly. Compare the changes in trend labels between two consecutive periods within the category sequence, based on the increasing period number. For example, if number C05 is "maintaining" and C06 is "rising", this is identified as a change in trend category from "maintaining" to "rising", marked as a switch. If C07 continues to be "rising", it is recorded as a trend continuation. Then, determine whether the category is... When a change occurs, the previous period's category label is compared with the current period's label. If the strings match, it is marked as "no change"; otherwise, it is marked as "change". The period number of each change is recorded as the change node, and the position of the node and the trend category before and after it are marked. For example, if C06 was "maintained" before and "rising" after, it is recorded as [C06, maintain → rise]. Then, a full traversal is performed on the entire period sequence to identify all trend switching points and arrange them in chronological order to generate a trend category change sequence. The recorded fields include three pieces of information: change period number, previous category, next category, and change type.

[0042] The trend classification and organization submodule analyzes the category change pattern in continuous cycles based on the trend category change sequence, determines the consistency interval of trend categories within the cycle sequence, optimizes the mapping method between category distribution and cycle sequence, and obtains stress trend classification groups. The algorithm sequentially extracts the position of each change node and the corresponding trend category label before and after the change. Then, it segments the trend category labels according to the cycle number order and finds segments with the same trend category in consecutive cycles. For example, if C08 to C11 are all "decline", this interval can be regarded as a consistent trend segment. The boundary condition for identifying consistent intervals is that the number of consecutive repetitions of the trend category is greater than or equal to 3. If two consecutive categories are of the same type but change in subsequent cycles, the identification is terminated. Then, each identified consistent interval is grouped and its starting cycle number, ending cycle number, and trend category label are recorded. On this basis, the original cycle number sequence is mapped one by one to the trend category interval. That is, the trend category segment to which each cycle belongs is marked in the cycle list. If a cycle number C10 is located in the "decline" segment C08 to C11, the trend category field corresponding to this number is "decline". Finally, the above mapping structure is sorted in ascending order of number and the trend category classification corresponding to each cycle number is output. The data is organized into structured fields, including cycle number, trend category, whether it is the starting point of change, and the number of the consistent interval to which it belongs. The output is the stress trend classification group.

[0043] like Figure 2 and Figure 7 As shown, the stress alarm linkage module includes: The hierarchical label discrimination sub-module determines the sequential relationship between the hierarchical labels of each period and the set labels based on the stress trend classification group, compares the matching status of the period label content with the warning standard, identifies the period segments where both the sequence and content are consistent, and obtains the label sequence matching information; Extract the corresponding label identification fields from the trend classification results of each period. For example, in the classification result, the hierarchical label corresponding to the "decrease" trend is L2, and the label corresponding to the "sharp decrease" is L3. Arrange the label sequences corresponding to all periods in numerical order, establish a label time series index. Then, extract the warning sequence of each level label from the set standard warning label library. For example, the standard sequence is set as L1 < L2 < L3 < L4. Compare whether the position of each period label in the trend classification group is consistent with the position of the same-level label in the set standard sequence to determine whether there is a label sequence misalignment. The judgment method is to read the period labels one by one starting from the number C01 and perform a string comparison with the content at the same numbered position in the standard label sequence. If the label content is exactly the same, it is considered that the label sequence is consistent; otherwise, record the number as inconsistent. Subsequently, summarize all the consistent result items and their period numbers to form a label matching index set, and then check whether there are three or more consecutive periods in the matching items where both the label sequence and content are exactly the same, which is used as the basis for judging the subsequent linkage signal control, and output the label sequence matching information.

[0044] The linkage signal output sub-module analyzes the relationship between the hierarchical label and the linkage control standard based on the label sequence matching information, compares whether the label content meets the requirements for the output of the sound and light linkage signal, identifies the period corresponding to the label matching status, adjusts the linkage signal output instruction, and obtains the sound and light alarm linkage sequence; Call the sound and light alarm control parameter configuration item to extract the set linkage trigger label level threshold, which is set as L3. When the label of a certain period is L3 or a higher-level label L4, it meets the linkage signal trigger condition. Then, read the label information with the matching status of "completely consistent" in each period, filter out the period numbers where the current period label is greater than or equal to the set level threshold, and mark them as "trigger candidate" under these period numbers. Then, analyze whether the linkage output control status in the same period is in the closed state. If it is in the closed state, prepare to activate the sound and light output. If it is already in the activated state, skip the generation of duplicate signals. Write the linkage output instruction identifier to each candidate period number, set it as "1" in the signal output field, and set the corresponding non-trigger period as "0" to generate a binary linkage control sequence with a length equal to the total number of periods. Each bit corresponds to a period, and the value of 1 represents that the sound and light alarm linkage signal needs to be triggered in this period, and the value of 0 represents no action. Record the label value and matching status of each action trigger source during the construction of this sequence, and output it as the sound and light alarm linkage sequence.

[0045] The tag remote synchronization submodule is based on the sound and light alarm linkage sequence. It determines the synchronization requirements of tags on local and remote terminals, compares the synchronization status of tag content, optimizes the information transmission and display of hierarchical tags between terminals, outputs remote synchronization feedback, and obtains the stress linkage triggering result. The system calls the current cached status information of the remote terminal communication module to check whether the periodic tags have been synchronized with the local tags. It reads the tag sequence stored in the remote terminal and compares the tag values ​​of the corresponding period in the remote and local terminals to determine whether the synchronization status is "synchronized" or "asynchronous". For periods with a status of "asynchronous", the tag value is written to the data synchronization buffer and prepared for transmission through the communication interface. After executing the transmission command, the status of the corresponding period in the remote terminal is updated to "synchronized". During the comparison process, it identifies intervals where multiple consecutive periodic tags are not synchronized and prioritizes batch synchronization operations. It uses whole-segment data transmission to reduce communication load. For periods where the tag value changes, the tag identifier and its corresponding period number need to be re-rendered and displayed on the remote terminal interface. The local tag change record is written to the remote terminal status log to ensure data recording continuity. The system generates a synchronization feedback result for each period, which includes five items: period number, current tag value, status before synchronization, status after synchronization, and whether remote display update is triggered. The output structure is the trigger result of the response linkage.

[0046] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A crocodile stress intelligent evaluation system based on behavioral characteristics, characterized in that, The system comprises: The time sequence behavior collection module extracts the first feeding action time collected by the underwater infrared camera equipment based on the feeding device record data, numbers the two groups of time signals and sequentially pairs them, screens the environmental monitoring data of the temperature and humidity sensor, collects various parameters, and obtains a behavior time sequence association set; The interval dynamic analysis module compares the numbering sequence based on the behavior time sequence association set, extracts the difference of the behavior response interval between adjacent periods, sequentially arranges the interval difference, and obtains a response interval change amount; The feeding fluctuation coordination module calculates the number of released bait in the current period based on the response interval change amount, combines the remaining number, obtains the actual feeding amount, compares the total feeding amount of the previous period, analyzes the fluctuation amplitude, and obtains a feeding behavior deviation amount; The trend classification judgment module judges the trend category change based on the feeding behavior deviation amount, combines the segmented recognition baseline, compares the trend category and the baseline sequence, sends a state change signal when a change occurs, and obtains a stress trend classification group; The stress alarm linkage module analyzes the classification label content based on the stress trend classification group, compares the label sequence, sends a sound and light linkage signal when the labels are consistent, synchronizes the classification label information to a remote terminal, and obtains a stress linkage triggering result.

2. The crocodile stress intelligent evaluation system based on behavioral characteristics according to claim 1, characterized in that, The behavior time sequence association set comprises a behavior time structure set, an action trigger index group and an environmental association mapping cluster, the response interval change amount comprises a period rhythm feature group, a reaction gradient sequence and a behavior fluctuation track piece, the feeding behavior deviation amount comprises a feeding trend quadrant, a behavior difference factor group and a dynamic deviation facet, the stress trend classification group comprises a trend level spectrum, a state evolution chain and segmented classification data, and the stress linkage triggering result comprises an alarm triggering piece, a linkage control node and a level output box.

3. The crocodile stress intelligent evaluation system based on behavioral characteristics according to claim 1, characterized in that, The time sequence behavior collection module comprises: The data time tag numbering submodule analyzes the feeding start time and the first feeding action time obtained by the underwater infrared camera equipment based on the feeding device record data, compares the order of the two types of time data, calculates the numbering index corresponding to the time, and obtains a time sequence behavior marker set; The behavior trigger pairing submodule screens the pairing information corresponding to the numbering sequence based on the time sequence behavior marker set, judges the numbering continuity to exclude discontinuous items, analyzes the environmental data obtained by the temperature and humidity sensor, matches according to the pairing time numbering, and obtains a numbered behavior corresponding structure; The environmental mapping fusion submodule calculates the corresponding relationship between each behavior event and environmental data based on the numbered behavior corresponding structure, induces the time point and action sequence of the feeding behavior association, optimizes the synchronization mode of the environmental data and behavior numbering, and obtains a behavior time sequence association set.

4. The behavior feature-based crocodile stress intelligent evaluation system according to claim 1, wherein The interval dynamic analysis module comprises: The behavior period division submodule judges the continuity between each numbering based on the behavior time sequence association set, compares the relative change of the feeding start time and the feeding action time, groups and classifies the behavior period data in order, optimizes the numbering mapping relationship of the period data, and obtains a behavior response period set; The response time difference calculation submodule calculates the time interval between the feeding initiation and the feeding action in each group of periods based on the set of behavior response periods, compares the change range of the time interval in adjacent periods, judges the increasing or decreasing trend of the time interval in the period sequence, summarizes the interval change content between periods, and obtains a period interval change group; The period difference arrangement submodule analyzes the time interval change of the numbered continuous periods based on the period interval change group, screens the representative interval change situation in the period sequence, integrates the number and interval change data, and obtains a response interval change amount.

5. The behavior feature-based crocodile stress intelligent evaluation system according to claim 1, wherein The feeding fluctuation coordination module includes: The bait release monitoring submodule judges whether the bait release data recorded by the intelligent feeding device is consistent with the period number based on the response interval change amount, compares the bait falling segment and the device record, determines the effectiveness of the release action, calls the camera screen to identify the appearance area of the remaining bait and filter environmental interference, calculates the difference between the release amount and the remaining amount, and obtains the period feeding amount; The feeding difference calculation submodule calculates the amount difference between the current period feeding amount and the last period feeding amount based on the period feeding amount, compares the direction change of the amount difference in adjacent periods, analyzes the correspondence between the feeding amount change and the period relationship in the number sequence, judges the change trend of the feeding amount increase or decrease in the continuous period, and obtains a feeding fluctuation sequence; The behavior state discrimination submodule judges the continuity of the feeding change trend in each period based on the feeding fluctuation sequence, analyzes the relationship between the change trend and the period number, compares the consistency of the current period change trend and the previous period trend, adjusts the behavior category label of the corresponding number, and eliminates the conflict between the previous and next periods, and obtains a feeding behavior offset amount.

6. The behavior feature-based crocodile stress intelligent evaluation system according to claim 1, wherein The trend grading judgment module includes: The segmented baseline comparison submodule judges the corresponding sequence between the offset amount of each period and the segmented baseline based on the feeding behavior offset amount, compares the coordination of the offset amount and the baseline in the period sequence, identifies the period segment where the baseline sequence and the offset amount are inconsistent, and obtains trend category comparison information; The trend category determination submodule judges whether the trend category of each period changes based on the trend category comparison information, analyzes the change process of the trend category label in the period sequence, compares the switching situation of the adjacent period category labels, marks the category change node, and obtains a trend category change sequence; The trend classification arrangement submodule analyzes the category change rule of the continuous period based on the trend category change sequence, judges the consistency interval of the trend category in the period sequence, optimizes the mapping mode of the category distribution and the period sequence, and obtains a stress trend classification group.

7. The behavior feature-based crocodile stress intelligent evaluation system according to claim 1, characterized in that, The stress alarm linkage module includes: The hierarchical label discrimination submodule judges the sequential relationship between the hierarchical label of each period and the set label based on the stress trend classification group, compares the matching state of the period label content and the warning standard, identifies the period segment with consistent sequence and content, and obtains label sequence matching information; The linkage signal output sub-module analyzes the relationship between the hierarchical label and the linkage control standard based on the label sequence matching information, compares whether the label content meets the sound-light linkage signal output requirement, identifies the period corresponding to the label matching state, adjusts the linkage signal output instruction, and obtains a sound-light alarm linkage sequence. The label remote synchronization sub-module judges the synchronization demand of the label between the local and remote terminals based on the sound-light alarm linkage sequence, compares the synchronization state of the label content, optimizes the information transmission and display of the hierarchical label between the terminals, outputs a remote synchronization feedback, and obtains a stress linkage trigger result.

8. The behavior feature-based crocodile stress intelligent evaluation system according to claim 1, wherein The first feeding action time refers to the time point when the crocodile first attempts to feed in the monitored video stream, which is obtained by underwater infrared camera equipment combined with action recognition. The time signal number refers to assigning a number to each group of feeding start and feeding action in sequence.

9. The behavior feature-based crocodile stress intelligent evaluation system according to claim 1, wherein The number sequence refers to the number of each group of feeding start to feeding action arranged in time sequence. The behavior response interval refers to the actual time difference between feeding start and the first feeding action in each period. The interval difference refers to the response time change value between adjacent periods.

10. The behavior feature based crocodile stress intelligent evaluation system according to claim 1, wherein, The bait release quantity refers to the actual weight or quantity of bait released by the intelligent feeding device in the current period. The remaining quantity refers to the weight or quantity of remaining bait in the pool that is not fed by the camera or sensor device after the current feeding round ends.

Images