Intelligent bait casting method and system for seawater fry of scylla paramamosain

Abstract

The invention relates to the technical field of intelligent bait casting, and discloses an intelligent bait casting method and system for scylla paramamosain seawater fries, and the method comprises the steps: extracting a motion track and distribution characteristics according to a real-time motion image of the scylla paramamosain fries in a culture pond, and obtaining behavior characteristic parameters of the scylla paramamosain fries; when the activity degree of the motion trail in the behavior characteristic parameters exceeds a historical data threshold value, it is judged that the scylla paramamosain seeds are in a real-time starvation state; according to the linear relation between the historical bait casting amount and the historical hunger state, the real-time bait casting amount is determined; controlling a bait casting device to scatter bait according to the real-time bait casting amount; after the bait distribution is completed, acquiring the feeding image of the offspring seeds again, and analyzing the feeding image of the offspring seeds, so as to obtain feedback data of the distributed bait; based on the feedback data, the linear relation is dynamically adjusted, and the real-time bait casting amount of the bait casting device is updated according to the dynamically-adjusted linear relation; the intelligent bait casting efficiency can be improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent baiting, and in particular to an intelligent baiting method and system for mud crab sea fry. BACKGROUND

[0002] In the process of mud crab sea fry culture, the traditional baiting method highly depends on manual experience and lacks accurate perception and scientific analysis of the real-time behavior state of the fry. Cultivation personnel usually determine the baiting time and baiting amount according to fixed time nodes or subjective observation, which cannot effectively capture the core behavior characteristic parameters such as the movement trajectory activity and distribution density of the fry, and thus cannot accurately identify the real-time hunger state of the fry. This baiting mode is prone to cause the baiting amount to deviate from the actual demand of the fry, and when the baiting amount is excessive, the excess bait will be deposited at the bottom of the culture pond, causing water pollution and destroying the ecological balance of the culture; when the baiting amount is insufficient, the fry will not be able to feed adequately, delaying the growth and development process and reducing the survival rate and overall culture quality of the fry.

[0003] Although some existing semi-intelligent baiting schemes attempt to introduce parameterized control, they do not establish a dynamic adaptation mechanism between historical baiting data and the hunger state of the fry, and ignore the influence of the culture environment variables and the growth stage of the fry on the baiting demand. Such schemes mostly determine the baiting amount based on a preset fixed linear relationship, and cannot calibrate and optimize the linear relationship in real time according to the actual feedback after the fry feeds. Under long-term operation, the baiting accuracy will gradually decrease as the culture period progresses, which not only cannot match the dynamic changes in the behavior law of the fry, but also causes waste of bait resources and increases the culture cost, and at the same time, cannot guarantee the stability of the growth of the fry, seriously restricting the intelligent upgrading and efficient development of mud crab sea fry culture. SUMMARY

[0004] The present application provides an intelligent baiting method and system for mud crab sea fry to solve the problems raised in the background.

[0005] To achieve the above-mentioned purpose, the present application provides an intelligent baiting method for mud crab sea fry, comprising:

[0006] S1, extracting the movement trajectory and distribution characteristics of the mud crab fry from the real-time activity image of the mud crab fry in the culture pond to obtain the behavior characteristic parameters of the mud crab fry;

[0007] S2, determining that the mud crab fry is in a real-time hunger state when the activity of the movement trajectory in the behavior characteristic parameters exceeds the historical data threshold;

[0008] S3, determining the real-time baiting amount corresponding to the real-time hunger state according to the linear relationship between the historical baiting amount and the historical hunger state;

[0009] S4. The bait scattering device according to the real-time baiting amount control scatters bait;

[0010] S5. After the bait scattering is completed, a seedling feeding image is acquired again, and a feeding behavior of the seedling feeding image is analyzed to obtain feedback data of the scattered bait;

[0011] S6. Based on the feedback data, the linear relationship is dynamically adjusted, and the real-time baiting amount of the baiting device is updated according to the dynamically adjusted linear relationship.

[0012] In a preferred embodiment, the behavior characteristic parameter of the Scylla paramamosain seedling is obtained by extracting a motion trajectory and a distribution characteristic of the Scylla paramamosain seedling from a real-time activity image of the Scylla paramamosain seedling in a breeding pond, and the behavior characteristic parameter of the Scylla paramamosain seedling includes:

[0013] The real-time activity image of the Scylla paramamosain seedling is subjected to brightness correction and noise filtering processing to obtain an enhanced image of the Scylla paramamosain seedling;

[0014] Based on the enhanced image, a contour boundary of the Scylla paramamosain seedling is recognized, and a centroid position of the seedling is labeled to obtain a position coordinate set of the Scylla paramamosain seedling;

[0015] According to the position coordinate set in a continuous time sequence, a displacement vector of the Scylla paramamosain seedling is determined to obtain a motion trajectory sequence of the Scylla paramamosain seedling;

[0016] The aggregation degree of the position coordinate set in a preset area of the breeding pond is analyzed to obtain a distribution density characteristic of the Scylla paramamosain seedling;

[0017] The motion trajectory sequence and the distribution density characteristic are fused to generate the behavior characteristic parameter of the Scylla paramamosain seedling.

[0018] In a preferred embodiment, when the activity degree of the motion trajectory in the behavior characteristic parameter exceeds a historical data threshold value, the Scylla paramamosain seedling is determined to be in a real-time hunger state, and the method includes:

[0019] Based on typical behavior characteristic parameters of the Scylla paramamosain seedling in a normal feeding state and a hunger state obtained in a historical breeding period, a historical behavior database of the Scylla paramamosain seedling is constructed;

[0020] Statistical analysis is performed on the motion trajectory activity degree data in the historical behavior database to obtain an upper limit threshold value of the activity degree and an activity degree change trend characteristic of the Scylla paramamosain seedling in the historical breeding period;

[0021] Based on a cumulative amount of displacement distance of the Scylla paramamosain seedling per unit time and a change frequency of the Scylla paramamosain seedling, a real-time activity degree index of the Scylla paramamosain seedling in a current monitoring period is determined.

[0022] comparing the real-time activity index with the upper limit threshold of activity, when the real-time activity index continuously exceeds the upper limit threshold of activity, generating a primary starvation signal of the Scylla paramamosain fry;

[0023] analyzing the deviation degree of the real-time activity index from the activity change trend feature, when the deviation degree reaches a preset level, generating a confirmation starvation signal of the Scylla paramamosain fry;

[0024] comprehensively determining that the Scylla paramamosain fry is in a real-time starvation state according to the primary starvation signal and the confirmation starvation signal.

[0025] In a preferred embodiment, the historical behavior database of the Scylla paramamosain fry is constructed based on the typical behavior characteristic parameters of the Scylla paramamosain fry in normal feeding state and starvation state obtained in the historical breeding cycle, including:

[0026] segmenting the motion trajectory activity data in the historical behavior database according to the breeding time window to obtain time series data segments of the Scylla paramamosain fry obtained in the historical breeding cycle;

[0027] extracting the activity peak value and the activity fluctuation range of each data segment in the time series data segment;

[0028] identifying the periodic feature of the activity peak value changing with the breeding cycle between the continuous time series data segments;

[0029] determining the upper limit threshold of activity representing the normal behavior boundary based on the distribution statistical result of the activity peak value;

[0030] establishing the activity change trend feature reflecting the behavior regularity change of the Scylla paramamosain fry according to the periodic feature and the activity fluctuation range;

[0031] storing the upper limit threshold of activity and the activity change trend feature in association to form a reference standard for real-time starvation state determination.

[0032] In a preferred embodiment, the real-time feeding amount corresponding to the real-time starvation state is determined according to the linear relationship between the historical feeding amount and the historical starvation state, including:

[0033] extracting the actual feeding amount data corresponding to different starvation state levels from the historical breeding records to obtain a historical data set of historical feeding amount and historical starvation state;

[0034] performing correlation analysis on the historical data set to obtain a corresponding relationship model of real-time historical feeding amount and real-time historical starvation state;

[0035] According to the intensity level of the real-time hunger state, a corresponding reference feeding amount is matched in the corresponding relationship model;

[0036] The reference feeding amount is adaptively adjusted in combination with the temperature parameter of the current aquaculture water body and the growth stage of the fry;

[0037] Based on the adaptive adjustment result, a real-time feeding amount corresponding to the real-time hunger state is determined.

[0038] In a preferred embodiment, the correlation analysis on the historical data set to obtain the corresponding relationship model of the real-time historical feeding amount and the real-time historical hunger state comprises:

[0039] The historical hunger state data and the historical feeding amount data in the historical data set are standardized to generate a regularized data set of the historical feeding amount and the real-time historical hunger state;

[0040] The correlation degree between different hunger state levels and the corresponding feeding amount in the regularized data set is analyzed to generate a correlation matrix of the historical feeding amount and the real-time historical hunger state;

[0041] Based on the correlation matrix, a core matching relationship between the hunger state level and the feeding amount is extracted to form an initial relationship model of the historical feeding amount and the real-time historical hunger state;

[0042] The stability of the core matching relationship in the initial relationship model is evaluated using a verification data set;

[0043] According to the verification result, the parameter weight of the core matching relationship is adjusted to generate an optimized initial relationship model;

[0044] The optimized initial relationship model is associated and calibrated with the feeding response effect of the fry to obtain a corresponding relationship model of the real-time historical feeding amount and the real-time historical hunger state.

[0045] In a preferred embodiment, the calculation formula of the relationship function in the initial relationship model is as follows:

[0046] ;

[0047] In the formula, is a matching intensity value, is a weight coefficient of the correlation relationship between the th hunger state and the th feeding amount, is a characteristic factor of the th hunger state, is a characteristic factor of the th feeding amount, and

[0048] In a preferred embodiment, after the completion of bait scattering, the fry feeding image is obtained again, and the fry feeding image is analyzed to obtain feedback data of the scattered bait, including:

[0049] After a preset time interval, the fry feeding image and image sequence of the breeding pond are collected;

[0050] The contrast between the fry individuals and the bait in the fry feeding image is improved to obtain an enhanced feeding image;

[0051] Based on the enhanced feeding image, the aggregation density of the Scylla paramamosain fry in the bait area is identified to obtain an aggregation distribution map of the Scylla paramamosain fry;

[0052] The feeding action frequency of the Scylla paramamosain fry in the image sequence is analyzed to obtain a feeding activity parameter of the Scylla paramamosain fry;

[0053] The residual amount of bait in the bait residual area in the enhanced feeding image is evaluated to obtain a bait consumption index of the bait residual area;

[0054] The aggregation distribution map, the feeding activity parameter, and the bait consumption index are integrated to generate the feedback data of the scattered bait.

[0055] In a preferred embodiment, based on the feedback data, the linear relationship is dynamically adjusted, and the real-time bait amount of the bait feeding device is updated according to the dynamically adjusted linear relationship, including:

[0056] The fry aggregation distribution map, the feeding activity parameter, and the bait consumption index in the feedback data are analyzed to obtain a bait feeding effect evaluation result of the scattered bait;

[0057] The deviation degree between the bait feeding effect evaluation result and the expected feeding effect in the bait amount and the actual demand is identified;

[0058] The linear relationship adjustment instruction is generated according to the deviation degree;

[0059] The linear relationship between the historical bait amount and the historical hunger state is updated based on the adjustment instruction to generate a dynamically adjusted linear relationship;

[0060] Based on the dynamically adjusted linear relationship, the real-time bait amount control parameter of the bait feeding device is updated.

[0061] In order to solve the above problems, the present application also provides an intelligent bait feeding system for Scylla paramamosain fry in seawater, which comprises:

[0062] An image extraction module is configured to extract a motion trail and distribution characteristics of the Scylla paramamosain fry according to a real-time activity image of the Scylla paramamosain fry in a breeding pond, and obtain a behavior characteristic parameter of the Scylla paramamosain fry.

[0063] A hunger state judgment module is configured to determine that the Scylla paramamosain fry is in a real-time hunger state when an activity degree of the motion trail in the behavior characteristic parameter exceeds a historical data threshold.

[0064] A real-time feeding amount generation module is configured to determine a real-time feeding amount corresponding to the real-time hunger state according to a linear relationship between a historical feeding amount and a historical hunger state.

[0065] A bait scattering module is configured to control a bait scattering device to scatter bait according to the real-time feeding amount.

[0066] A bait feeding feedback module is configured to obtain a fry feeding image again after the bait scattering is completed, analyze a feeding behavior of the fry feeding image, and obtain feedback data of the scattered bait.

[0067] A bait feeding update module is configured to dynamically adjust the linear relationship based on the feedback data, and update a real-time feeding amount of the bait scattering device according to the dynamically adjusted linear relationship.

[0068] Compared with the prior art, the present application has the following beneficial effects:

[0069] 1. The present application can accurately extract the motion trail sequence and distribution density characteristics of the fry by performing brightness correction and noise filtering on the real-time activity image of the fry, generate the behavior characteristic parameter, and accurately determine the real-time hunger state of the fry by combining the activity degree threshold and the change trend characteristics in the historical behavior database, so as to ensure that the feeding opportunity is highly consistent with the actual physiological needs of the fry. When determining the real-time feeding amount, the present application can effectively improve the adaptability of the feeding amount by adjusting the linear relationship between the historical feeding amount and the historical hunger state, the current breeding water temperature parameter and the growth stage of the fry, so as to ensure that the fry can feed sufficiently and reasonably, promote the stability of the growth and development of the fry, and reduce the invalid consumption of bait resources.

[0070] 2. The application can continuously adapt to the dynamic changes of the behavior rules of the fry by forming a closed-loop optimization mechanism of the feeding strategy after the bait scattering is completed, which can generate feedback data by acquiring fry feeding images again, analyzing the fry aggregation density, feeding activity parameters and bait consumption indicators, and dynamically adjusting the linear relationship between the historical bait amount and the hunger state based on the feedback data, and synchronously updating the real-time bait amount of the bait feeding device, thereby maintaining a high feeding precision for a long time. In addition, the entire feeding process relies on the intelligent operation of the image extraction, hunger state judgment, feeding feedback and other modules to greatly improve the efficiency of intelligent feeding and reduce the dependence on manual intervention. At the same time, through accurate feeding, the deposition of excess bait is reduced, the pollution of the breeding water body is avoided, the stability of the breeding ecological environment is ensured, and the breeding quality and survival rate of the Scylla paramamosain marine fry are further improved. BRIEF DESCRIPTION OF DRAWINGS

[0071] Figure 1 A flowchart of an intelligent bait feeding method for Scylla paramamosain marine fry according to an embodiment of the application is shown in the figure.

[0072] Figure 2 A functional module diagram of an intelligent bait feeding system for Scylla paramamosain marine fry according to an embodiment of the application is shown in the figure.

[0073] The implementation, functional features and advantages of the application will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0074] It should be understood that the specific embodiments described herein are only used to explain the application and do not limit the application.

[0075] The embodiment of the application provides an intelligent bait feeding method for Scylla paramamosain marine fry. The execution subject of the intelligent bait feeding method for Scylla paramamosain marine fry includes but is not limited to at least one of the electronic devices which can be configured to execute the method provided by the embodiment of the application, such as a server, a terminal and the like. In other words, the intelligent bait feeding method for Scylla paramamosain marine fry can be executed by software or hardware installed in a terminal device or a server device. The server includes but is not limited to a single server, a server cluster, a cloud server or a cloud server cluster, etc. The server can be a stand-alone server, or a cloud server providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content distribution networks (CDN), and big data and artificial intelligence platforms, etc. Basic cloud computing services.

[0076] Reference Figure 1As shown, a flowchart of an intelligent feeding method for Scylla paramamosain seawater fry provided by an embodiment of the present application is shown. In this embodiment, the intelligent feeding method for Scylla paramamosain seawater fry comprises:

[0077] S1, according to the real-time activity image of Scylla paramamosain fry in the breeding pond, the motion trajectory and distribution characteristics of the Scylla paramamosain fry are extracted, and the behavior characteristic parameters of the Scylla paramamosain fry are obtained;

[0078] In the embodiment of the present application, the behavior characteristic parameters of the Scylla paramamosain fry are obtained by extracting the motion trajectory and distribution characteristics of the Scylla paramamosain fry according to the real-time activity image of the Scylla paramamosain fry in the breeding pond, comprising:

[0079] The real-time activity image of the Scylla paramamosain fry is subjected to brightness correction and noise filtering processing to obtain an enhanced image of the Scylla paramamosain fry;

[0080] Based on the enhanced image, the contour boundary of the Scylla paramamosain fry is recognized, and the center of mass position of the fry is labeled to obtain a position coordinate set of the Scylla paramamosain fry;

[0081] According to the position coordinate set of the continuous time sequence, the displacement vector of the Scylla paramamosain fry is determined to obtain a motion trajectory sequence of the Scylla paramamosain fry;

[0082] The aggregation degree of the position coordinate set in the preset area of the breeding pond is analyzed to obtain the distribution density characteristics of the Scylla paramamosain fry;

[0083] The motion trajectory sequence and the distribution density characteristics are fused to generate the behavior characteristic parameters of the Scylla paramamosain fry.

[0084] Specifically, first, the real-time activity image of the Scylla paramamosain fry is subjected to brightness correction, and the image is converted to a gray image after gray histogram equalization, the number of pixels of each gray level is counted, the cumulative distribution function is calculated and the gray value of each pixel is adjusted according to the function, so that the image brightness is uniform, then noise filtering is performed, mean filtering is used, the average value of the gray values of the adjacent pixels in a fixed range around each pixel is calculated, and the average value is used to replace the original pixel gray value to remove noise, and finally the enhanced image of the Scylla paramamosain fry is obtained.

[0085] Further, based on the obtained enhanced image, the contour boundary of the Scylla paramamosain larvae is identified, and a Sobel operator method is used to calculate the gray value difference between each pixel in the enhanced image and its adjacent pixels in the horizontal and vertical directions, respectively. According to the difference, it is determined whether it is an edge pixel. All edge pixels are connected to form a contour boundary. Then, the centroid position of the larvae is labeled, and the coordinates of all pixels in the region surrounded by the contour boundary are determined. The average values of the horizontal and vertical coordinates of these pixels are calculated, and the coordinates composed of the two average values are taken as the centroid position. Thus, the position coordinate set of the Scylla paramamosain larvae is obtained.

[0086] Further, the displacement vector of the Scylla paramamosain larvae is determined according to the position coordinate set in the continuous time sequence. The position coordinates of two adjacent time points are taken from the position coordinate set, and the position coordinates of the latter time point are subtracted from the position coordinates of the former time point. The result is the displacement vector of the time period. All adjacent time period displacement vectors are arranged in chronological order to obtain the motion trajectory sequence of the Scylla paramamosain larvae.

[0087] Further, the distribution density feature is obtained by analyzing the aggregation degree of the position coordinate set in the preset area of the culture pond. The preset area of the culture pond is divided into several sub-areas of the same size, the number of Scylla paramamosain larvae in each sub-area is counted, and the ratio of the number of larvae in each sub-area to the area of the sub-area is calculated. The ratio is the distribution density of the corresponding sub-area. The aggregation degree of the larvae in the preset area is determined according to the distribution density of all sub-areas. The distribution density feature of the Scylla paramamosain larvae is obtained by comprehensively considering the distribution density of all sub-areas.

[0088] Further, the behavior characteristic parameter of the Scylla paramamosain larvae is generated by fusing the motion trajectory sequence and the distribution density feature. The key information such as the displacement direction and distance in the motion trajectory sequence and the aggregation area and average distribution density in the distribution density feature is determined. These key information is integrated to form comprehensive data by combining the displacement direction and distance of each time period with the distribution density of the sub-area where the larvae are located in the corresponding time period. The comprehensive data is sorted to generate the behavior characteristic parameter reflecting the activity rule of the Scylla paramamosain larvae.

[0089] In summary, the brightness correction and noise filtering are performed on the real-time activity image of the Scylla paramamosain larvae to effectively improve the image clarity, provide a high-quality image basis for accurately identifying the contour boundary of the larvae and labeling the centroid position, avoid feature extraction errors caused by image quality problems, and ensure the accuracy of the position coordinate set of the larvae.

[0090] Overall, based on the accurate position coordinate set, both the movement trajectory sequence of the fry can be determined through the continuous time sequence to intuitively reflect the activity state of the fry, and the aggregation degree of the fry in the preset area of the breeding pond can be analyzed to obtain the distribution density characteristics, and the behavior characteristic parameters generated by the fusion of the two can multi-dimensionally and comprehensively depict the real-time behavior state of the fry, thereby providing reliable data support for the subsequent accurate determination of the real-time hunger state of the fry, ensuring the scientificity and rationality of the subsequent feeding decision, improving the perception accuracy of the intelligent feeding system for the behavior of the fry from the source, and laying a key foundation for realizing accurate feeding.

[0091] S2, when the activity degree of the movement trajectory in the behavior characteristic parameter exceeds the historical data threshold, determining that the Scylla paramamosain fry is in a real-time hunger state;

[0092] In the embodiment of the present application, when the activity degree of the movement trajectory in the behavior characteristic parameter exceeds the historical data threshold, the Scylla paramamosain fry is determined to be in a real-time hunger state, comprising:

[0093] Based on the typical behavior characteristic parameters of the Scylla paramamosain fry in the normal feeding state and the hunger state obtained in the historical breeding period, a historical behavior database of the Scylla paramamosain fry is constructed;

[0094] Statistical analysis is performed on the movement trajectory activity data in the historical behavior database to obtain the upper limit threshold of the activity degree and the activity degree change trend characteristics of the Scylla paramamosain fry in the historical breeding period;

[0095] Based on the cumulative amount of displacement distance of the Scylla paramamosain fry per unit time and the change frequency of the Scylla paramamosain fry, a real-time activity index of the Scylla paramamosain fry in the current monitoring period is determined;

[0096] The real-time activity index is compared with the upper limit threshold of the activity degree, and when the real-time activity index continuously exceeds the upper limit threshold of the activity degree, a primary hunger signal of the Scylla paramamosain fry is generated;

[0097] The deviation degree of the real-time activity index relative to the activity degree change trend characteristics is analyzed, and when the deviation degree reaches a preset level, a confirmation hunger signal of the Scylla paramamosain fry is generated;

[0098] The primary hunger signal and the confirmation hunger signal are comprehensively judged to determine that the Scylla paramamosain fry is in a real-time hunger state.

[0099] In the embodiment of the present application, based on the typical behavior characteristic parameters of the Scylla paramamosain fry in the normal feeding state and the hunger state obtained in the historical breeding period, the historical behavior database of the Scylla paramamosain fry is constructed, comprising:

[0100] segment the motion trajectory activity data in the historical behavior database according to a culture time window to obtain time series data segments of the time sequence data of the Trapa incisa fry in the historical culture period;

[0101] extract the activity peak value and the activity fluctuation range of each data segment in the time series data segment;

[0102] identify the periodicity characteristic of the activity peak value changing with the culture period between the continuous time series data segments;

[0103] determine the upper limit threshold of the activity value representing the normal behavior boundary based on the distribution statistical result of the activity peak value;

[0104] establish the activity change trend characteristic reflecting the behavior rule change of the Trapa incisa fry according to the periodicity characteristic and the activity fluctuation range;

[0105] store the upper limit threshold of the activity value and the activity change trend characteristic in association to form a reference standard for real-time starvation state determination.

[0106] Specifically, from the historical culture period, the typical behavior characteristic parameters of the Trapa incisa fry in the normal feeding state and the typical behavior characteristic parameters of the Trapa incisa fry in the starvation state are collected respectively, these parameters are classified and arranged according to the culture period and the fry state, and are uniformly stored in a designated storage carrier to construct a historical behavior database of the Trapa incisa fry.

[0107] Further, the motion trajectory activity data of the Trapa incisa fry in all historical culture periods is extracted from the historical behavior database, these data are arranged in sequence according to time sequence, the maximum value data among them is found as the upper limit threshold of the activity value of the Trapa incisa fry in the historical culture period, and the activity change trend characteristic is summarized by observing the rising and falling rule of the motion trajectory activity data in each time segment in different historical culture periods.

[0108] Further, in the current monitoring period, the distance of each displacement of the Trapa incisa fry in unit time is recorded, the cumulative amount of the displacement distance of the Trapa incisa fry in unit time is obtained by adding all the displacement distances, and the number of times of obvious increase and decrease of the displacement distance of the Trapa incisa fry in unit time is counted as the change frequency of the Trapa incisa fry, the cumulative amount of the displacement distance in unit time and the change frequency are combined to form the real-time activity index of the Trapa incisa fry in the current monitoring period.

[0109] Further, a continuous monitoring time period is set as a judgment period, and in the judgment period, each time the real-time activity index of the Scylla paramamosain fry is obtained, it is compared with the upper limit threshold of the activity, and if the comparison result is that the real-time activity index exceeds the upper limit threshold of the activity in the whole judgment period, it is determined that the real-time activity index continuously exceeds the upper limit threshold of the activity, and at this time, the primary hunger signal of the Scylla paramamosain fry is generated.

[0110] Further, according to the activity change trend feature, the normal fluctuation range of the activity of the Scylla paramamosain fry in different time periods is determined, the difference between the real-time activity index of each time point in the current monitoring period and the normal activity value of the corresponding time point in the activity change trend feature is calculated, and if the difference exceeds the set normal fluctuation range, it is determined that the deviation reaches the preset level, and the confirmation hunger signal of the Scylla paramamosain fry is generated.

[0111] Further, the obtained signal is checked, and when the primary hunger signal of the Scylla paramamosain fry and the confirmation hunger signal of the Scylla paramamosain fry exist at the same time, it is directly determined that the Scylla paramamosain fry is in a real-time hunger state.

[0112] Specifically, the motion trajectory activity data of all Scylla paramamosain fry is extracted from the historical behavior database, a fixed length of cultivation time window is set, for example, 2 hours as a time window, and then the motion trajectory activity data is sequentially classified into corresponding time windows according to the order of cultivation time, all data in each time window form a data segment, and finally the time sequence data segment of the Scylla paramamosain fry in the historical cultivation period is obtained.

[0113] Further, for each time sequence data segment, the values of all motion trajectory activity data in the data segment are viewed one by one, the data with the maximum value in the data segment is found, which is determined as the activity peak value of the data segment, and the activity peak value of the data segment is subtracted from the activity data with the minimum value, and the result is the activity fluctuation range of the data segment, so that the activity peak value and the activity fluctuation range of each data segment are extracted.

[0114] Further, all time sequence data segments are arranged according to the time sequence of the historical cultivation period, the activity peak values of each continuous data segment are sequentially recorded, and the value changes of these activity peak values at the same time nodes in different cultivation periods are observed, such as the activity peak values of the data segment of the same time window on the first day of each week, and if it is found that these peak values will appear similar high-low change rule every fixed cultivation period, for example, the peak value will rise and then fall every 7 days, then this cyclic change rule is determined as the periodic feature of the activity peak value between the continuous time sequence data segments with the change of the cultivation period.

[0115] Further, the peak values of the activity of all time series data segments are collected, these peak values are arranged in descending order of numerical value, the number of occurrences of each peak value is counted, and the peak values with fewer occurrences and in a higher numerical range are found, for example, among all peak values, the peak values ranked in the top 5% and with only 1-2 occurrences are determined as the upper limit threshold of the activity representing the normal behavior boundary, ensuring that the peak value of the activity of the mud crab fry in the normal behavior state will not exceed the threshold.

[0116] Further, according to the identified periodic characteristics, the change direction and approximate range of the activity peak value of the mud crab fry corresponding to different breeding cycle stages are determined, for example, in the early stage of breeding, the activity peak value shows a gradually rising trend, in the middle stage of breeding, it remains stable, and in the later stage of breeding, it gradually decreases; combined with the activity fluctuation range of each time series data segment, the normal fluctuation interval of the activity in each breeding cycle stage is determined, the peak value change rule of different breeding cycle stages is integrated with the corresponding normal fluctuation interval, and the activity change trend characteristics reflecting the change of the behavior rule of the mud crab fry are established.

[0117] Further, the determined upper limit threshold of the activity is associated with the established activity change trend characteristics, for example, the upper limit threshold of the activity corresponding to each breeding cycle stage is marked in the activity change trend characteristics, and then the associated upper limit threshold of the activity and the activity change trend characteristics are stored in a designated database or file together to form a reference standard for real-time starvation state determination.

[0118] In summary, the starvation state determination method is based on the typical behavior parameters of normal feeding and starvation state of fry in the historical breeding cycle, constructs a historical behavior database and extracts the upper limit threshold of the activity and the activity change trend characteristics, provides a scientific and behavior rule fitting reference for real-time determination, and avoids the determination deviation caused by subjective or single standard without historical data support.

[0119] In summary, the real-time activity index combined with the displacement distance accumulation and change frequency calculation in unit time can accurately capture the real activity state of the fry, ensure the effective identification of the activity change related to starvation, and through the double signal comprehensive determination logic of "real-time index exceeding threshold generating primary signal + deviating from trend characteristics reaching preset level generating confirmation signal", the false judgment probability can be greatly reduced, the starvation state is avoided to be misjudged due to temporary activity fluctuation, and it is ensured that the subsequent feeding is triggered only when the fry is in the starvation state, which lays a key premise for the subsequent accurate matching of feeding amount and ensures that the feeding decision is highly consistent with the actual demand of the fry.

[0120] S3, according to the linear relationship between the historical feeding amount and the historical starvation state, determining the real-time feeding amount corresponding to the real-time starvation state;

[0121] In the embodiment of the present application, the determination of the real-time feeding amount corresponding to the real-time hunger state according to the linear relationship between the historical feeding amount and the historical hunger state comprises:

[0122] extracting the actual feeding amount data corresponding to different hunger state levels from the historical breeding records to obtain a historical data set of the historical feeding amount and the historical hunger state;

[0123] performing correlation analysis on the historical data set to obtain a corresponding relationship model of the real-time historical feeding amount and the real-time historical hunger state;

[0124] matching the corresponding reference feeding amount in the corresponding relationship model according to the intensity level of the real-time hunger state;

[0125] adapting the reference feeding amount in combination with the temperature parameter of the current breeding water body and the growth stage of the fry;

[0126] determining the real-time feeding amount corresponding to the real-time hunger state based on the adaptive adjustment result.

[0127] In the embodiment of the present application, the correlation analysis on the historical data set to obtain the corresponding relationship model of the real-time historical feeding amount and the real-time historical hunger state comprises:

[0128] performing standardization processing on the historical hunger state data and the historical feeding amount data in the historical data set to generate a normalized data set of the historical feeding amount and the real-time historical hunger state;

[0129] analyzing the correlation degree between different hunger state levels and the corresponding feeding amount in the normalized data set to generate a correlation matrix of the historical feeding amount and the real-time historical hunger state;

[0130] extracting the core matching relationship between the hunger state level and the feeding amount based on the correlation matrix to form an initial relationship model of the historical feeding amount and the real-time historical hunger state;

[0131] evaluating the stability of the core matching relationship in the initial relationship model using a verification data set;

[0132] adjusting the parameter weight of the core matching relationship according to the verification result to generate an optimized initial relationship model;

[0133] associating and calibrating the optimized initial relationship model with the feeding response effect of the fry to obtain the corresponding relationship model of the real-time historical feeding amount and the real-time historical hunger state.

[0134] In the embodiment of the present application, the calculation formula of the relationship function in the initial relationship model is as follows:

[0135] ;

[0136] In the formula, is a matching intensity value, is a weight coefficient of the association relationship between the first hunger state and the first feeding amount, is a characteristic factor of the first hunger state, is a characteristic factor of the first feeding amount, and

[0137] Specifically, all records containing the hunger state grade of the Scylla paramamosain fry and the corresponding actual feeding amount are screened from the historical breeding records. The hunger state is first divided into three fixed grades of slight hunger, moderate hunger, and severe hunger. Then the actual feeding weight data in each breeding process under each grade is extracted one by one. All the historical feeding amount data and the corresponding historical hunger state grade data are sorted in chronological order to obtain a historical data set of historical feeding amount and historical hunger state.

[0138] Further, the historical hunger state grade and the corresponding historical feeding amount in the historical data set are paired and grouped according to the hunger state grade. The values of all historical feeding amounts under each grade are counted respectively. The concentration range of the historical feeding amount under the same hunger state grade is observed to determine the corresponding relationship between each hunger state grade and the concentration range. The association of all grades and the corresponding feeding amount range is sorted into a structure to obtain a corresponding relationship model of real-time historical feeding amount and real-time historical hunger state.

[0139] Further, the intensity grade of the real-time hunger state of the Scylla paramamosain fry is first determined, such as moderate hunger. Then the feeding amount range associated with the intensity grade in the corresponding relationship model is checked. The middle value of the range is taken as the corresponding reference feeding amount to complete the matching operation in the corresponding relationship model.

[0140] Further, the temperature parameter of the current breeding water body is obtained. If the temperature is higher than the average temperature of the fry growth stage in the historical breeding records, it means that the metabolism speed of the fry is accelerated, and a certain proportion of feeding amount needs to be added to the reference feeding amount. If the temperature is lower than the average temperature, it means that the metabolism speed is slowed down, and a certain proportion of feeding amount needs to be reduced from the reference feeding amount. At the same time, if the fry is in a rapid growth stage, a small amount of feeding is added to the feeding amount after temperature adjustment. If the fry is in a slow growth stage, the feeding amount after temperature adjustment is maintained to complete the adaptive adjustment of the reference feeding amount.

[0141] Further, the feeding amount value adjusted by the current breeding water temperature parameter and the growth stage adaptability of the fry is directly determined as the real-time feeding amount corresponding to the real-time hunger state of the current Scylla paramamosain fry.

[0142] Specifically, the historical hunger state data in the historical data set is first divided into fixed levels, with slight hunger recorded as 1, moderate hunger recorded as 2, and severe hunger recorded as 3. Then, the difference between the maximum value and the minimum value of the historical feeding amount data is calculated. Each historical feeding amount data is subtracted by the minimum value and then divided by the difference, so that all feeding amount data are in the interval of 0-1. The processed historical hunger state data and the historical feeding amount data are arranged according to the original corresponding relationship to generate a normalized data set of historical feeding amount and real-time historical hunger state.

[0143] Further, for each hunger state level in the normalized data set, the number of occurrences of each feeding amount under the level is counted. The number of occurrences of each feeding amount is divided by the total number of feeding amounts under the level to obtain the proportion of each feeding amount under the corresponding level. The higher the proportion, the stronger the correlation. Taking the hunger state level as the row and the feeding amount as the column, the proportion values are filled into the corresponding cells to generate a correlation matrix of historical feeding amount and real-time historical hunger state.

[0144] Further, by checking the values in each row of the correlation matrix, the feeding amount corresponding to the cell with the maximum value in each row is selected as the matching feeding amount for the hunger state level. The corresponding relationship between all hunger state levels and the corresponding matching feeding amount is arranged in table form to form an initial relationship model of historical feeding amount and real-time historical hunger state.

[0145] Further, part of the data in the historical data set that does not participate in the construction of the initial relationship model is selected as a verification data set. Each historical hunger state level in the verification data set is input into the initial relationship model to obtain the feeding amount output by the model. The output feeding amount is compared with the actual historical feeding amount in the verification data set. If the difference between the two is less than 10% of the total feeding amount range in more than 80% of the comparison results, it is determined that the core matching relationship is stable, otherwise it is not stable.

[0146] Further, the parameter weight is adjusted according to the verification result. If the matching relationship between a hunger state level and a feeding amount has a small difference in verification, the influence of the matching relationship in the model is increased, that is, the matching relationship is preferentially referred to in subsequent use. If the difference is large, the influence is reduced. An optimized initial relationship model is generated after adjustment.

[0147] Further, the feeding response effect of the mud crab fry under different feeding amounts is collected, and the feeding response effect is good within 1 hour after feeding if the bait remaining amount is less than 10%, the feeding amount corresponding to each hunger state level in the optimized initial relationship model is associated with the feeding response effect, if the response corresponding to a certain feeding amount is poor, the feeding amount is adjusted to the value of good response, and the corresponding relationship model of the real-time historical feeding amount and the real-time historical hunger state is obtained after calibration.

[0148] Specifically, the calculation formula of the relationship function in the initial relationship model is used to calculate the matching strength value , and in the calculation, first, for each th hunger state and th feeding amount, the weight coefficient corresponding to the two is multiplied by the characteristic factor of the th hunger state and the characteristic factor of the th feeding amount, to obtain the product result of each group, and then all the product results of all groups are added, and the final cumulative result is the matching strength value , which can reflect the closeness of the matching between the th hunger state and the th feeding amount.

[0149] In general, based on the actual feeding amount data corresponding to different hunger state levels in the historical breeding records, the corresponding relationship model is constructed through correlation analysis of the historical data set, so that the determination of the real-time feeding amount has solid historical data support, avoids subjective setting without data support, and guarantees the scientificity and reliability of the feeding amount calculation, providing a foundation for precise feeding from the data level.

[0150] In general, after the matching of the reference feeding amount, adaptive adjustment is made in combination with the current water temperature parameter and the fry growth stage, so that the real-time feeding amount not only fits the real-time hunger state of the fry, but also adapts to the differences in environmental conditions and growth needs, further improving the precision and adaptability of the feeding amount. Ultimately, the real-time feeding amount can be highly consistent with the actual feeding needs of the fry, meeting the nutritional needs of the fry growth, and avoiding waste or insufficient bait, providing a key guarantee for subsequent efficient feeding and stable growth of the fry.

[0151] S4, controlling the feeding device to scatter bait according to the real-time feeding amount;

[0152] In the embodiment of the application, the real-time feeding amount data is extracted from the storage carrier of the real-time feeding amount, and the data is sent to the control module of the feeding device through wired transmission mode, so as to ensure that the control module accurately receives the real-time feeding amount information.

[0153] Furthermore, after receiving real-time feeding data, the control module of the feeding device analyzes the data, identifies the feeding amount value, and then converts the value into specific control commands, such as commands to control the motor rotation duration and valve opening range, according to the feeding device's feeding distribution mechanism.

[0154] Furthermore, the execution component of the feeding device receives the control command sent by the control module, first checks whether the bait in the bait storage bin inside the device meets the real-time feeding requirements. If the bait is sufficient, it starts the internal preparation program to ensure that the bait conveying channel is unobstructed and enters the ready-to-distribute state.

[0155] Furthermore, the actuator of the feeding device drives the motor to rotate the feed conveying screw according to the control command. The screw gradually conveys the feed in the storage bin to the distributing plate according to the real-time feeding amount. The distributing plate rotates at a fixed speed, evenly scattering the delivered feed into the breeding pond, thus completing the feed distribution.

[0156] Furthermore, the bait balance sensor on the feeding device detects the amount of bait reduction in the storage bin. When the reduction reaches the real-time feeding amount, the sensor sends a distribution completion signal to the control module. After receiving the signal, the control module records the completion time and actual feeding amount of this feeding operation, forming a feeding record.

[0157] In summary, using scientifically calculated real-time feeding amounts as the core control basis, the feeding device can accurately execute the feeding action, ensuring that the feeding amount completely corresponds to the real-time feeding needs of the seedlings. This avoids problems such as feeding amount deviation and uneven distribution range caused by experience errors in manual operation, ensuring the accuracy of feeding from the execution level, and allowing each unit of feed to act efficiently in the seedling feeding process, providing key support for the seedlings to fully obtain nutrition.

[0158] In summary, the modular distribution of feed based on real-time feeding effectively avoids resource waste caused by excessive feed accumulation and prevents insufficient feeding of fry. This reduces aquaculture costs while minimizing the risk of excess feed polluting the aquaculture water and maintaining a stable aquaculture ecosystem. Furthermore, modular automated control reduces the uncertainty of human intervention, improves the stability and efficiency of the feeding process, and lays a reliable foundation for subsequent dynamic optimization of feeding strategies based on feeding feedback, thus ensuring the continuous and efficient operation of the intelligent feeding system.

[0159] S5. After the bait is distributed, the feeding images of the seedlings are acquired again, and the feeding behavior of the seedlings in the feeding images is analyzed to obtain the feedback data of the distributed bait.

[0160] In the embodiment of the present application, after the bait scattering is completed, the fry feeding image is acquired again, and the feeding behavior of the fry is analyzed to obtain the feedback data of the scattered bait, including:

[0161] The fry feeding image and the image sequence of the breeding pond are collected after a preset time interval;

[0162] The contrast between the fry individuals and the bait in the fry feeding image is improved to obtain an enhanced feeding image;

[0163] Based on the enhanced feeding image, the aggregation density of the Scylla paramamosain fry in the bait area is identified to obtain the aggregation distribution map of the Scylla paramamosain fry;

[0164] The feeding action frequency of the Scylla paramamosain fry in the image sequence is analyzed to obtain the feeding activity parameter of the Scylla paramamosain fry;

[0165] The residual amount of bait in the bait residual area in the enhanced feeding image is evaluated to obtain the bait consumption index of the bait residual area;

[0166] The aggregation distribution map, the feeding activity parameter, and the bait consumption index are integrated to generate the feedback data of the scattered bait.

[0167] Specifically, when the set fixed preset time interval arrives, the image acquisition devices installed at different positions above the breeding pond are started, each device simultaneously photographs different areas in the breeding pond, a single photograph acquires a single fry feeding image, continuous photographing acquires multiple images arranged in chronological order, forming a fry feeding image sequence, and finally the fry feeding image and the image sequence of the breeding pond are collected.

[0168] Further, the collected fry feeding image is converted into a gray-scale image, the number of pixels contained in each gray-scale level in the gray-scale image is counted, the cumulative distribution of each gray-scale level is calculated according to the number of pixels, the gray-scale value of each pixel in the image is adjusted according to the cumulative distribution, the difference between the gray-scale value corresponding to the fry individuals and the gray-scale value corresponding to the bait is increased, thereby improving the contrast between the fry individuals and the bait, and obtaining the enhanced feeding image.

[0169] Further, in the enhanced feeding image, the bait area where the bait is located is determined by identifying the color and shape features of the bait, the bait area is evenly divided into multiple sub-areas of the same size, the number of Scylla paramamosain fry in each sub-area is counted one by one, the color identifier is set according to the number of fry in each sub-area, all sub-areas with color identifiers are combined to obtain the aggregation distribution map of the Scylla paramamosain fry.

[0170] Further, the adjacent two images in the larva feeding image sequence are extracted, the opening and closing state of the chelae and the body movement amplitude of the Scylla paramamosain larvae in the two images are compared, it is judged whether feeding action exists, all adjacent images are compared in time sequence, the total number of feeding action of the Scylla paramamosain larvae in unit time is counted, and the total number is the feeding activity parameter of the Scylla paramamosain larvae.

[0171] Further, in the enhanced feeding image, the bait residual area is determined, the bait residual area is divided into a plurality of standard size grids, the area occupied by the bait in each grid is calculated, the total area of the bait remaining in all grids is added to obtain the total area of the bait remaining, the total area of the bait at the initial baiting is subtracted from the total area of the bait remaining to obtain the bait consumption area, and the ratio of the bait consumption area to the total area of the bait at the initial baiting is the bait consumption index of the bait residual area.

[0172] Further, the larva aggregation in different regions in the Scylla paramamosain larva aggregation distribution map, the feeding activity parameter value of the Scylla paramamosain larvae and the bait consumption index value of the bait residual area are arranged in the same record document, the time of data collection and the corresponding breeding pond number are marked, and the feedback data of the scattered bait is formed.

[0173] In summary, after the bait is scattered, the larva feeding image is collected, the image quality is optimized by improving the contrast between the larvae and the bait in the image, the image interference is cleared for accurately identifying the larva aggregation density, the feeding action frequency and the bait remaining amount, the analysis deviation caused by image blur is effectively avoided, and it is ensured that the feedback data can truly reflect the actual feeding condition, and reliable data basis is provided for subsequent evaluation of the baiting effect.

[0174] In summary, the aggregation distribution map, the feeding activity parameter and the bait consumption index obtained by analyzing the image can comprehensively present the matching state of the real-time baiting amount and the larva demand from three core dimensions of larva behavior response, feeding intensity and bait consumption, break the limitation of single data dimension, and make the baiting effect evaluation more stereoscopic and comprehensive. The feedback data can directly provide clear basis for subsequent dynamic adjustment of the linear relationship, ensure that the subsequent baiting strategy optimization has accurate data support, help the intelligent baiting system to continuously adapt to the demand of the larvae, and further improve the scientificity and effectiveness of the baiting.

[0175] S6, based on the feedback data, dynamically adjusting the linear relationship, and updating the real-time baiting amount of the baiting device according to the dynamically adjusted linear relationship.

[0176] In the embodiment of the application, the real-time baiting amount of the baiting device is updated according to the dynamically adjusted linear relationship based on the feedback data, which comprises:

[0177] analyzing seedling aggregation distribution, feeding activity parameter and bait consumption index in the feedback data to obtain the feeding effect evaluation result of the scattered bait;

[0178] identifying deviation degree between the feeding effect evaluation result and expected feeding effect in feeding amount and actual demand;

[0179] generating linear relationship adjustment instruction according to the deviation degree;

[0180] updating linear relationship between historical feeding amount and historical hunger state based on the adjustment instruction to generate dynamically adjusted linear relationship;

[0181] updating real-time feeding amount control parameter of the feeding device based on the dynamically adjusted linear relationship.

[0182] Specifically, seedling aggregation distribution is extracted from feedback data to observe whether Scylla paramamosain seedlings are mainly concentrated in bait area in the graph. If concentrated, the aggregation effect is good, and if dispersed, the aggregation effect is poor. Feeding activity parameter is extracted and compared with normal feeding activity range in corresponding hunger state in historical breeding. If within the range, the activity is normal, and if above or below, it is abnormal. Bait consumption index is extracted. If the index value is high, bait consumption is high, and if the value is low, bait consumption is low. The aggregation effect, activity and bait consumption are combined to determine the feeding effect evaluation result of the scattered bait.

[0183] Further, the expected feeding effect standard is set, including that seedling aggregation rate in bait area needs to reach more than 80%, feeding activity parameter is in the middle value of historical normal range, and bait consumption index needs to reach more than 70%. Then, the aggregation rate, feeding activity parameter and bait consumption index in the feeding effect evaluation result are compared with the expected standard respectively, the difference value between actual value and expected value is calculated, and the deviation degree between feeding amount and actual demand is determined according to the difference value. For example, if the actual bait consumption index is 50% and the expected value is 70%, the difference value is 20%, and the deviation degree is 20%.

[0184] Further, the adjustment direction and amplitude are determined according to the deviation degree. If the deviation degree is that feeding amount is too much, the linear relationship adjustment instruction of “reducing historical feeding amount corresponding to current hunger state, and the reduction amplitude matches the deviation degree” is generated. If the deviation degree is that feeding amount is too little, the linear relationship adjustment instruction of “increasing historical feeding amount corresponding to current hunger state, and the increase amplitude matches the deviation degree” is generated.

[0185] Furthermore, from the linear relationship between historical feeding amount and historical hunger state, find the hunger state level corresponding to the current deviation and the historical feeding amount under that level. Modify the historical feeding amount according to the magnitude in the linear relationship adjustment instruction. For example, if the original feeding amount corresponding to mild hunger was 100g, and the instruction requires a 20% reduction, then modify it to 80g. Replace the corresponding value in the original linear relationship with the modified feeding amount to generate a dynamically adjusted linear relationship.

[0186] Furthermore, examine the feeding amount corresponding to each hunger state level in the dynamically adjusted linear relationship, and convert each feeding amount into the control parameters of the feeding device. For example, a feeding amount of 80g corresponds to the motor of the feeding device rotating for 8 seconds and the valve opening at 30 degrees. Replace the original control parameters in the feeding device, such as the motor rotation time and valve opening degree corresponding to each hunger state, with the new parameters to complete the update of the real-time feeding amount control parameters of the feeding device.

[0187] In summary, by analyzing the seedling aggregation and distribution map, feeding activity parameters, and feed consumption indicators in the feedback data, the degree of deviation between the feeding amount and the actual needs of the seedlings can be accurately identified. Based on the linear relationship adjustment instructions generated by the deviation, the linear relationship between historical feeding amount and historical hunger state can be dynamically adapted to the changes in seedling behavior patterns, avoiding the decrease in adaptability caused by the fixation of the linear relationship. This ensures that the linear relationship always fits the real-time physiological needs of the seedlings and the characteristics of the breeding environment, providing accurate model support for subsequent feeding amount calculations.

[0188] In summary, updating the real-time feeding control parameters of the feeding device based on the dynamically adjusted linear relationship allows subsequent feeding actions to directly match the adjusted precise strategy, effectively eliminating potential overfeeding or underfeeding in the early stages and ensuring that the amount of feed given each time is highly consistent with the current needs of the seedlings. This closed-loop mechanism of "feedback-adjustment-update" also allows the feeding system to continuously optimize its feeding logic, constantly improving feeding accuracy as the breeding cycle progresses. This reduces feed resource waste, controls breeding costs, and provides stable and suitable feeding conditions for the seedlings, helping to ensure the stability of seedling growth and the quality of breeding.

[0189] like Figure 2 The diagram shown is a functional block diagram of an intelligent feeding system for marine seedlings of mud crabs provided in an embodiment of the present invention.

[0190] The intelligent bait feeding system 100 for the Scylla paramamosain seawater fry can be installed in an electronic device. According to the functions to be realized, the intelligent bait feeding system 100 for the Scylla paramamosain seawater fry can include an image extraction module 101, a hunger state judgment module 102, a real-time bait amount generation module 103, a bait scattering module 104, a bait feeding feedback module 105, and a bait feeding updating module 106. The modules in the present application can also be referred to as units, which refer to a series of computer program segments capable of being executed by an electronic device processor and capable of completing fixed functions, and are stored in the memory of the electronic device.

[0191] In the present embodiment, the functions of each module / unit are as follows:

[0192] The image extraction module 101 is configured to extract the motion trajectory and distribution characteristics of the Scylla paramamosain fry according to the real-time activity image of the Scylla paramamosain fry in the culture pond, so as to obtain the behavior characteristic parameters of the Scylla paramamosain fry.

[0193] The hunger state judgment module 102 is configured to determine that the Scylla paramamosain fry is in a real-time hunger state when the activity level of the motion trajectory in the behavior characteristic parameters exceeds a historical data threshold.

[0194] The real-time bait amount generation module 103 is configured to determine the real-time bait amount corresponding to the real-time hunger state according to the linear relationship between the historical bait amount and the historical hunger state.

[0195] The bait scattering module 104 is configured to control the bait feeding device to scatter bait according to the real-time bait amount.

[0196] The bait feeding feedback module 105 is configured to obtain the fry feeding image again after the bait scattering is completed, analyze the feeding behavior of the fry feeding image, and obtain the feedback data of the scattered bait.

[0197] The bait feeding updating module 106 is configured to dynamically adjust the linear relationship based on the feedback data, and update the real-time bait amount of the bait feeding device according to the dynamically adjusted linear relationship.

[0198] In several embodiments provided in the present application, it should be understood that the disclosed method and system can be implemented in other ways. For example, the system embodiments described above are only illustrative, for example, the division of the modules is only a logical function division, and another division mode can be used in actual implementation.

[0199] The modules described as separate components may or may not be physically separate, and the components displayed as modules may or may not be physical units, that is, may be located in one place, or may be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiment scheme.

[0200] In addition, each functional module in various embodiments of the application can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of hardware plus software functional modules.

[0201] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application.

[0202] The embodiments of the present application can acquire and process related data based on artificial intelligence technology. Among them, artificial intelligence is to use digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, obtain knowledge and use knowledge to obtain the best results.

[0203] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application and are not limiting. Although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solutions of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present application.

Claims

1. An intelligent feeding method for Scylla paramamosain seawater fry, characterized in that, The method comprises: S1, extracting the motion trajectory and distribution characteristics of the Scylla paramamosain fry according to the real-time activity image of the Scylla paramamosain fry in the breeding pond, to obtain the behavior characteristic parameters of the Scylla paramamosain fry; S2, when the activity degree of the motion trajectory in the behavior characteristic parameters exceeds the historical data threshold value, determining that the Scylla paramamosain fry is in a real-time hunger state; S3, determining the real-time feeding amount corresponding to the real-time hunger state according to the linear relationship between the historical feeding amount and the historical hunger state; S4, controlling the feeding device to scatter feed according to the real-time feeding amount; S5, after the feed scattering is completed, obtaining the fry feeding image again, and analyzing the feeding behavior of the fry feeding image to obtain the feedback data of the scattered feed; S6, based on the feedback data, dynamically adjusting the linear relationship, and updating the real-time feeding amount of the feeding device according to the dynamically adjusted linear relationship.

2. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 1, characterized in that, The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises:

3. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 1, characterized in that, The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method comprises: The method 4. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 3, characterized in that, The historical behavior database of the Scylla paramamosain fry is constructed based on the typical behavior characteristic parameters of the Scylla paramamosain fry in a normal feeding state and a starvation state obtained in a historical breeding cycle, and the historical behavior database comprises: The motion trajectory activity data in the historical behavior database is segmented according to a breeding time window to obtain time series data segments of the Scylla paramamosain fry obtained in the historical breeding cycle; The activity peak value and the activity fluctuation range of each data segment in the time series data segments are extracted; Periodic characteristics of the activity peak value between the continuous time series data segments with the change of the breeding cycle are identified; Based on the distribution statistical result of the activity peak value, an upper limit threshold of the activity value representing the boundary of the normal behavior is determined; According to the periodic characteristics and the activity fluctuation range, the activity change trend characteristics reflecting the change of the behavior rule of the Scylla paramamosain fry are established; The activity upper limit threshold and the activity change trend characteristics are associated and stored to form a reference standard for real-time starvation state determination.

5. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 1, characterized in that, The real-time feeding amount corresponding to the real-time starvation state is determined according to the linear relationship between the historical feeding amount and the historical starvation state, and the real-time starvation state determination comprises: Actual feeding amount data corresponding to different starvation state levels are extracted from historical breeding records to obtain a historical data set of the historical feeding amount and the historical starvation state; Correlation analysis is performed on the historical data set to obtain a corresponding relationship model of the real-time historical feeding amount and the real-time historical starvation state; According to the intensity level of the real-time starvation state, a corresponding reference feeding amount is matched in the corresponding relationship model; The reference feeding amount is adaptively adjusted in combination with the temperature parameter of the current breeding water body and the fry growth stage; Based on the adaptive adjustment result, the real-time feeding amount corresponding to the real-time starvation state is determined.

6. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 5, characterized in that, The correlation analysis on the historical data set to obtain the corresponding relationship model of the real-time historical feeding amount and the real-time historical starvation state comprises: The historical starvation state data and the historical feeding amount data in the historical data set are standardized to generate a regularized data set of the historical feeding amount and the real-time historical starvation state; The correlation degree between different starvation state levels and the corresponding feeding amount in the regularized data set is analyzed to generate an association matrix of the historical feeding amount and the real-time historical starvation state; Based on the association matrix, the core matching relationship between the starvation state level and the feeding amount is extracted to form an initial relationship model of the historical feeding amount and the real-time historical starvation state; The stability of the core matching relationship in the initial relationship model is evaluated using a verification data set; According to the verification result, the parameter weight of the core matching relationship is adjusted to generate an optimized initial relationship model; The optimized initial relationship model is associated and calibrated with the fry feeding response effect to obtain the corresponding relationship model of the real-time historical feeding amount and the real-time historical starvation state.

7. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 6, characterized in that, The calculation formula of the relationship function in the initial relationship model is as follows: ； In the formula, is the matching strength value, is the weight coefficient of the correlation between the th hunger state and the th feeding amount, is the characteristic factor of the th hunger state, is the characteristic factor of the th feeding amount, and 8. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 7, characterized in that, After the bait scattering is completed, the fry feeding image is obtained again, and the fry feeding image is analyzed to obtain feedback data of the scattered bait, comprising: acquire a larva feeding image and an image sequence of the breeding pond after a preset time interval; enhance a contrast between the larva individuals and the bait in the larva feeding image, to obtain an enhanced feeding image; identify an aggregation density of the Scylla paramamosain larvae in a bait area based on the enhanced feeding image, to obtain a Scylla paramamosain larva aggregation distribution map; analyze a feeding action frequency of the Scylla paramamosain larvae in the image sequence, to obtain a feeding activity parameter of the Scylla paramamosain larvae; evaluate a bait residual amount of a bait residual area in the enhanced feeding image, to obtain a bait consumption index of the bait residual area; integrate the aggregation distribution map, the feeding activity parameter, and the bait consumption index, to generate feedback data of the scattered bait.

9. The intelligent feeding method for Scylla paramamosain sea water fry according to claim 1, characterized in that, The linear relationship is dynamically adjusted based on the feedback data, and a real-time bait feeding amount of the bait feeding device is updated according to the dynamically adjusted linear relationship, including: The Scylla paramamosain larva aggregation distribution map, the feeding activity parameter, and the bait consumption index in the feedback data are analyzed, to obtain a bait feeding effect evaluation result of the scattered bait; a deviation degree between the bait feeding effect evaluation result and an expected feeding effect in the bait feeding amount and the actual demand is identified; a linear relationship adjustment instruction is generated according to the deviation degree; the linear relationship between the historical bait feeding amount and the historical hunger state is updated based on the adjustment instruction, to generate a dynamically adjusted linear relationship; a real-time bait feeding amount control parameter of the bait feeding device is updated based on the dynamically adjusted linear relationship.

10. An intelligent feeding system for Scylla paramamosain marine fry, characterized in that, The system includes: an image extraction module configured to extract a motion trajectory and a distribution feature of the Scylla paramamosain larvae from a real-time activity image of the Scylla paramamosain larvae in the breeding pond, to obtain a behavior feature parameter of the Scylla paramamosain larvae; a hunger state judgment module configured to determine that the Scylla paramamosain larvae are in a real-time hunger state when an activity degree of the motion trajectory in the behavior feature parameter exceeds a historical data threshold value; a real-time bait feeding amount generation module configured to determine a real-time bait feeding amount corresponding to the real-time hunger state according to a linear relationship between a historical bait feeding amount and a historical hunger state; a bait scattering module configured to control a bait feeding device to scatter bait according to the real-time bait feeding amount; a bait feeding feedback module configured to acquire a larva feeding image again after the bait scattering is completed, and analyze a feeding behavior of the larva feeding image, to obtain feedback data of the scattered bait; a bait feeding update module configured to dynamically adjust the linear relationship based on the feedback data, and update a real-time bait feeding amount of the bait feeding device according to the dynamically adjusted linear relationship.

Images