Apparatus and method for separating individuals of macrobrachium rosenbergii at the molt stage from non-molt stage individuals

Abstract

The present application relates to shrimp individual separation device technical field, especially be used for macrobrachium rosenbergii molting period and non molting period individual separation's device and method. The rearing pond is the groove type pool body of cuboid, the bottom is provided with a plurality of support legs, the both ends edges position on the top of rearing pond is fixed with first cross beam board in coordination, the first cross beam board is provided with second cross beam board below, the both ends of second cross beam board are welded in the inner wall of rearing pond, a plurality of separation mechanisms are installed on second cross beam board; The separation mechanism includes a suspension column, the suspension column is welded at the bottom of second cross beam board through a plurality of hanging columns, the inside of suspension column is hollow structure, the overall light weight of device is realized, and the through hole is formed in the top center position, is used for docking type through installation covering drive column. The present application can realize the individual separation of molting period and non molting period of macrobrachium rosenbergii, reduce the frequency of mixed culture between individuals, improve the efficiency and quality of culture.

Description

Technical Field

[0001] This invention relates to the field of shrimp individual separation equipment technology, and in particular to an apparatus and method for separating macrophage individuals during and outside the molting stage. Background Technology

[0002] The giant freshwater crustacean, *Macrobrachium rosenbergii*, is a high-value freshwater crustacean aquaculture species whose growth and development are accompanied by periodic molting. Molting is a necessary physiological process for the giant freshwater prawn's growth; however, under high-density artificial culture conditions, differences in molting time among individuals often lead to the co-cultivation of individuals in the same pond who are molting and those who are not. During the molting period, the prawn's exoskeleton is not yet hardened, its mobility is weak, and its defense capabilities are poor, making it highly susceptible to attack from non-molting prawns, resulting in cannibalism, increased mortality, and decreased yield. Especially in large-scale farms, manually separating individuals in the molting and non-molting periods is not only time-consuming and labor-intensive but also prone to causing stress, further affecting survival rates and growth rates.

[0003] Current separation methods primarily rely on manual observation and catching. Manually identifying molting states is time-consuming and labor-intensive, making it unsuitable for the automated management needs of large-scale aquaculture. Secondly, relying on visual judgment of individual molting states is easily affected by light, water quality, and individual differences, leading to misjudgments and missed detections. Alternatively, simple mechanical separation facilities may be used, but existing mechanical separation devices are mostly based on size or weight differences, failing to distinguish individuals at different physiological stages and hindering intelligent separation at the behavioral level. Furthermore, frequent catching or physical screening with existing mechanical separation equipment can easily cause shrimp damage and stress, affecting the survival rate of individuals during molting. Therefore, there is an urgent need to design a device that combines attraction functions with a controllable separation mechanism, enabling automated separation management of giant freshwater prawns during and outside of molting stages, thereby effectively reducing cannibalism and improving aquaculture efficiency. Summary of the Invention

[0004] This invention overcomes the shortcomings of the prior art and provides a device and method for separating individuals of giant freshwater prawns during and outside the molting stage.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: This invention provides a device for separating individuals of giant freshwater prawns during and outside the molting stage, the device comprising a rearing tank: The feeding pool is a rectangular trough-shaped pool with multiple support legs at the bottom. A first crossbeam plate is fixed to both ends of the top edge of the feeding pool. A second crossbeam plate is located directly below the first crossbeam plate. The two ends of the second crossbeam plate are welded to the inner wall of the feeding pool. Several separation mechanisms are installed on the second crossbeam plate. The separation mechanism includes a suspension column, which is welded to the bottom of the second crossbeam plate by several hanging columns. The suspension column has a hollow structure inside, which makes the whole device lightweight. A through hole is opened at the top center for docking through installation of the cover drive column. An extension rod is provided at the top of the cover drive column. The end of the extension rod passes through the second crossbeam plate and is fixed with an electromagnetic device. The side column of the suspension column is provided with a circular slide rail. The circular slide rail is located close to the bottom of the suspension column. The circular slide rail is connected to a separation baffle plate. Both the circular slide rail and the separation baffle plate have a matching groove structure of the same size. The circular slide rail and the separation baffle plate are interlocked with each other through the matching groove structure, so that the separation baffle plate can slide in a ring around the circular slide rail on the suspension column. The top of the breeding pool is equipped with a multi-dimensional separation detection module, which includes an arc-shaped guide rail, a pulley guide block is fitted and connected on the arc-shaped guide rail, and an industrial wide-angle camera is installed at the bottom of the pulley guide block.

[0006] Furthermore, in a preferred embodiment of the present invention, the first crossbeam plate and the second crossbeam plate are arranged in parallel alignment on the feeding pool, and a relative distance is maintained between the first crossbeam plate and the second crossbeam plate.

[0007] Furthermore, in a preferred embodiment of the present invention, the covering drive column is made of transparent PVC polyvinyl chloride material, and an attraction protection cover is provided at the bottom end of the covering drive column. The attraction protection cover has several drainage holes on its side for covering and protecting the molting prawns and attracting the non-molting prawns to gather under the suspension column. The bottom edge of the attraction protection cover is provided with several circumferentially arranged turbulence teeth.

[0008] Furthermore, in a preferred embodiment of the present invention, a sonar imaging sensor is installed on the lure protective cover for sonar imaging to detect the phenotypic characteristics of the giant freshwater prawn during its molting period and its relative position within the area protected by the lure protective cover.

[0009] Furthermore, in a preferred embodiment of the present invention, a return spring is sleeved on the outside of the extension rod. One end of the return spring is fixed to the top of the suspension column, and the other end is fixed to the bottom of the second crossbeam plate. A plurality of metal blocks are provided at the bottom of the first crossbeam plate. The number and arrangement of the metal blocks are consistent with the separation mechanism, and each metal block is located directly above the electromagnetic device corresponding to each separation mechanism.

[0010] Furthermore, in a preferred embodiment of the present invention, the separation barrier plate is a semi-circular arc-shaped structure, and a plurality of first driven teeth are provided on the outer side of the arc. The first driven teeth mesh with the teeth on the first driving gear. The first driving gear is fixed to the end of the first servo motor, and the first servo motor is mounted on the suspension column.

[0011] Furthermore, in a preferred embodiment of the present invention, a second driving gear is installed in the middle of the pulley guide block. The second driving gear is fixed to the output end of the second servo motor. The second driving gear meshes with the second driven gear, and the second driven gear is disposed on the outer edge of the arc-shaped guide rail.

[0012] A second aspect of the present invention provides a method for separating macrophage individuals during and outside the molting stage, applicable to any of the apparatuses described in the present invention, specifically comprising the following steps: S1: Obtain the clusters of giant freshwater prawns in molting and non-molting stages to be separated, put all the clusters to be separated into the rearing pond, and simultaneously start the separation mechanism and multi-dimensional separation detection module. S2: The industrial camera of the multi-dimensional separation detection module takes all-round, multi-dimensional pictures of the cluster of giant freshwater prawns to be separated in the breeding pond to obtain the aggregated image data of individual giant freshwater prawns. S3: Using standard image features of giant freshwater prawns at different growth stages as noise terms, a variational encoder is used to learn and train the noise of the cluster center on the clustered image data based on the noise terms and perform reverse noise reduction to obtain the actual feature vector located at the cluster center. The analysis of whether the actual feature vector is a molting period feature controls the opening of the inducing protective cover to cover the giant freshwater prawn individuals during the molting period. S4: Based on the aggregation characteristics of non-molting stage shrimp attracted by molting stage shrimp at different non-molting stages, fractal processing is performed on the non-molting stage pixel feature points in the aggregated image data, and the main and secondary gradient directions are calculated to determine the fractal slope of the aggregation of non-molting stage shrimp. According to the fractal slope, the separation barrier is controlled to open another area, so that the non-molting stage shrimp individuals attracted by the molting stage shrimp individuals can enter the area. S5: Upon entry, release the covered molting macrophage individuals, thus separating molting macrophage individuals from non-molting macrophage individuals.

[0013] Furthermore, in a preferred embodiment of the present invention, step S3 specifically includes the following steps: The separation index of the molting period of giant freshwater prawn is obtained. Based on the separation index, the molting period image data of giant freshwater prawn and the text semantic prompts of the molting period image data are retrieved in the big data network. The standard image feature set of giant freshwater prawn at different growth stages is also obtained. A latent random noise space is constructed based on a standard image feature set. A pre-trained variational encoder and variational decoder are introduced. The variational encoder is used to map the clustered image data to the probability distribution matrix of the latent random noise space, and the latent vector of the cluster center is constructed based on the semantic prompts of the text. A variational decoder is used to extract the latent vector of the clustered image data during the mapping process. Gaussian noise about the standard image feature set is continuously added to the latent vector of the clustered center in the latent noise space, and the noise addition time step is recorded. The training process involves gradually injecting Gaussian noise into the aggregated image data at multiple noise-adding time steps to learn the random perturbation diffusion distribution of the aggregated images that have the standard image features of the giant freshwater prawn, and to generate a perturbation diffusion law model. A priori condition diffusion network is introduced to construct a backward denoising operator. Based on the embedded preset guidance coefficients of text semantic prompts, the backward denoising operator is used to discretize and eliminate noise in the noisy latent vector of the disturbance diffusion law model based on the guidance coefficients, and finally obtains the actual feature vector of the giant freshwater prawn located at the aggregation center in the aggregation image data. Based on big data, a set of molting period feature vectors of giant freshwater prawns is obtained. If at least one molting period feature vector in the set has a greater degree of agreement with the actual feature vector than a preset degree of agreement, then the electromagnetic device is controlled to open the protective cover to cover the molting giant freshwater prawn individuals.

[0014] Furthermore, in a preferred embodiment of the present invention, step S4 specifically includes the following steps: Obtain the preset resolution scheduling strategy for spatial shooting by industrial wide-angle cameras, and construct a non-spatial grid array based on the preset resolution scheduling strategy; Binary pixel conversion aggregated image data, and non-spatial grid array overlay and matte onto the aggregated image data after binary pixel conversion. At the same time, based on big data, pixel feature points of giant freshwater prawns during non-molting period and aggregation characteristics and aggregation index of prawns in different non-molting periods affected by molting period attraction are obtained. Only count the number of bins with at least one non-molting pixel feature point to obtain a number of non-molting prominent bins. Based on the aggregation characteristics and aggregation index, a scaling scale is preset. The non-molting prominent bins are scaled according to the scaling scale to obtain a series of fractal sizes of non-molting pixel feature points corresponding to prominent bins. The Prewitt operator is introduced to extract multidimensional gradients from binary pixels in aggregated image data. The horizontal and vertical gradients of each non-molting pixel feature point are obtained. The gradient covariance matrix is ​​calculated by combining the gradient magnitude and direction expressed by the horizontal and vertical gradients, and the structural tensor between two adjacent non-molting pixel feature points is obtained. The principal and secondary directions of the clustering gradient distribution of non-molting pixel features are calculated based on the structural tensor, and the principal and secondary directions of energy are obtained. Based on the principal and secondary directions of energy, each non-molting pixel feature point is fitted in the linear regression equation according to the fractal size to obtain the fractal slope of the clustering of non-molting giant freshwater prawns in the clustered image data. If the fractal slope is greater than the preset fractal slope, the non-molting stage giant freshwater prawns in the aggregated image data are calibrated as having high aggregation. The separation mechanism in the area is then activated by the corresponding first servo motor to drive the separation barrier to open another area.

[0015] The beneficial technical effects of this invention are as follows: The control electromagnetic device is activated, causing the suspended column, which is relatively fixed, to rise vertically upwards. This, in turn, causes the attraction protective cover and the cover driving column to rise synchronously. At this time, the molting giant freshwater prawns with poor mobility will gradually and slowly enter the area opened by the rising attraction protective cover over time. When the sonar imaging sensor detects that they have completely entered the attraction protective cover, the control electromagnetic device is turned off, canceling the electromagnetic force. This causes the cover driving column to fall back into place under the action of the reset elastic force, thereby quickly covering and protecting the molting giant freshwater prawns directly below, achieving temporary separation and protection of the molting giant freshwater prawns. Subsequently, the first servo motor is controlled to rotate the first drive gear. The first drive gear further drives the separation barrier plate to slide in a circular guide along the annular slide rail, thus opening the separation barrier plate to create an independent rearing area for non-molting giant freshwater prawns. At this time, the attracted and gathered non-molting giant freshwater prawns can enter the separation area through the opened entrance, achieving the separation effect between molting and non-molting giant freshwater prawns. This avoids harming molting giant freshwater prawns by non-molting individuals, increasing the survival rate of reared giant freshwater prawns, reducing the frequency of cannibalism, and improving rearing efficiency and quality. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained from these drawings without creative effort.

[0017] Figure 1 This is a three-dimensional structural diagram of the device; Figure 2 for Figure 1 Enlarged structural diagram at point AA; Figure 3 A three-dimensional structural diagram of the separation mechanism located inside the feeding tank; Figure 4 for Figure 3 Enlarged structural diagram at point BB; Figure 5 This is a partial structural diagram of the separation mechanism.

[0018] The annotations in the attached figures are explained as follows: 101. Feeding tank; 102. Support leg; 103. First crossbeam plate; 104. Second crossbeam plate; 105. Suspension column; 106. Cover drive column; 107. Extension rod; 108. Electromagnetic device; 109. Lure protection cover; 201. Drainage hole; 202. Turbidator tooth; 203. Return spring; 204. Metal block; 205. Circular slide rail; 206. Separation barrier plate; 207. First driven tooth; 208. First driving gear; 209. First servo motor; 301. Arc-shaped guide rail; 302. Pulley guide block; 303. Industrial wide-angle camera; 304. Second driving gear; 305. Second servo motor; 306. Second driven tooth. Detailed Implementation

[0019] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. These drawings are simplified schematic diagrams, which only illustrate the basic structure of the present invention in a schematic manner. Therefore, they only show the components related to the present invention. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of this application can be combined with each other.

[0020] In the description of this application, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships based on the orientation or positional relationships shown in the accompanying drawings, are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as limiting the scope of protection of this application. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0021] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this application based on the specific circumstances.

[0022] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0023] like Figures 1-5 As shown, the first aspect of the present invention provides an apparatus for separating individuals of giant freshwater prawns during and outside the molting stage, the apparatus comprising a rearing tank 101.

[0024] The feeding pool 101 is a rectangular trough-shaped pool with multiple support legs 102 at the bottom. A first crossbeam plate 103 is fixed to both ends of the top edge of the feeding pool 101. A second crossbeam plate 104 is located directly below the first crossbeam plate 103. The two ends of the second crossbeam plate 104 are welded to the inner wall of the feeding pool 101. Several separation mechanisms are installed on the second crossbeam plate 104.

[0025] The first crossbeam plate 103 and the second crossbeam plate 104 are arranged in parallel alignment on the feeding pool 101, and the first crossbeam plate 103 and the second crossbeam plate 104 maintain a relative distance.

[0026] It should be noted that each separation mechanism is arranged in a compact linear pattern with equal spacing and is positioned at the center line of the rearing pond 101, maintaining a relatively parallel pattern with the wide side of the rearing pond 101. This allows several separation mechanisms to form a "separation dam" that divides the rearing pond 101 into two identical and independent rearing spaces, used to separate molting-stage giant freshwater prawns from non-molting-stage giant freshwater prawns.

[0027] The separation mechanism includes a suspension column 105, which is welded to the bottom of the second crossbeam plate 104 by several hanging columns. The suspension column 105 has a hollow structure inside, which makes the whole device lightweight. A through hole is opened at the center of the top for docking and through-installation of the cover drive column 106. An extension rod 107 is provided on the top of the cover drive column 106. The end of the extension rod 107 passes through the second crossbeam plate 104 and is fixed with an electromagnetic device 108.

[0028] The covering drive column 106 is made of transparent PVC polyvinyl chloride material. An attraction protection cover 109 is provided at the bottom of the covering drive column 106. The attraction protection cover 109 has several drainage holes 201 on its side, which are used to cover and protect the giant freshwater prawns in the molting period and attract the giant freshwater prawns in the non-molting period to gather under the suspension column 105. The bottom edge of the attraction protection cover 109 is provided with several circumferentially arranged baffle teeth 202.

[0029] The sonar imaging sensor is installed on the lure protection cover 109 to detect the phenotypic characteristics of the giant freshwater prawn during the molting period and its relative position in the area protected by the lure protection cover 109.

[0030] A return spring 203 is sleeved on the outside of the extension rod 107. One end of the return spring 203 is fixed to the top of the suspension column 105, and the other end is fixed to the bottom of the second crossbeam plate 104. Several metal blocks 204 are provided at the bottom of the first crossbeam plate 103. The number and arrangement of the metal blocks 204 are consistent with the separation mechanism, and each metal block 204 is located directly above the electromagnetic device 108 corresponding to each separation mechanism.

[0031] It should be noted that when molting and non-molting giant freshwater prawns are mixed and poured into the rearing tank 101, the giant freshwater prawns in the rearing tank 101 are captured and photographed by a multi-dimensional separation and detection module. When the detection identifies a molting giant freshwater prawn in the area where the separation mechanism is located, the electromagnetic device 108 is activated to generate electromagnetic force. Under the action of electromagnetic force, the electromagnetic device 108 attracts the metal block 204 directly above it, thereby causing the electromagnetic device 108 to drive the relatively fixed suspension of the covering drive column 106. The column 105 rises vertically upwards, and during the ascent, the top of the cover drive column 106 compresses the return spring 203, causing the return spring 203 to apply a downward return force to the cover drive column 106. The attraction protection cover 109 rises synchronously with the cover drive column 106. Since the giant freshwater prawns in the molting period have poor mobility and usually stay at the bottom of the pond, the giant freshwater prawns in the molting period will gradually and slowly enter the area opened after the attraction protection cover 109 rises over time. Simultaneously, a sonar imaging sensor detects in real time whether the phenotypic characteristics of a complete molting-stage giant freshwater prawn (Macrobrachium rosenbergii) are present in the open area directly below the attraction protection cover 109. If present, it indicates that the molting-stage giant freshwater prawn has completely entered the coverage and protection range of the attraction protection cover 109. The electromagnetic device 108 is then turned off, and the electromagnetic force is removed. The electromagnetic device 108 and the metal block 204 can no longer attract each other, causing the reset spring force to drive the covering drive column 106 to descend back to its original position. This allows the attraction protection cover 109 to quickly cover and protect the molting-stage giant freshwater prawn directly below, achieving temporary separation and protection of the molting-stage giant freshwater prawn. This avoids damage to the molting-stage giant freshwater prawn by non-molting-stage giant freshwater prawns and increases the survival rate of giant freshwater prawns at different stages of cultivation.

[0032] It should be noted that because molting-stage giant freshwater prawns have poor mobility, they attract attacks from non-molting-stage giant freshwater prawns. At this time, molting-stage prawns are densely surrounded by non-molting-stage prawns, forming a clustering phenomenon. Therefore, when the attraction shield 109 rapidly descends to cover molting-stage giant freshwater prawns, it may harm the surrounding non-molting-stage prawns. Therefore, if the disturbance flow teeth 202 can disturb the aquatic environment to a certain extent during the rapid descent of the attraction shield 109, the molting-stage giant freshwater prawns, due to their poor mobility, will not be affected by the water disturbance and will not exhibit stress behavior to escape the open coverage area. Conversely, non-molting-stage giant freshwater prawns will sense the threat of disturbance in the surrounding aquatic environment and will exhibit stress response behavior, rapidly spreading out to escape. Therefore, the rapidly descending attraction shield 109 will not harm either molting-stage or non-molting-stage giant freshwater prawns, significantly improving the separation safety factor and reducing separation mortality.

[0033] It should be noted that, because non-molting giant freshwater prawns are attracted not only by the reduced activity of molting individuals but also by their sex pheromones or odors, if the attraction shield 109 is completely sealed, the aforementioned pheromones emitted by molting giant freshwater prawns will be blocked and isolated when covered, thus significantly reducing the attraction and aggregation effect on non-molting giant freshwater prawns. To address this, the several drainage holes 201 on the attraction shield 109 of this device allow the pheromones emitted by molting individuals to maintain continuous diffusion in the aquaculture environment, thereby better attracting non-molting giant freshwater prawns to gather in large numbers, increasing the aggregation scale, and effectively accelerating the separation rate between molting and non-molting giant freshwater prawns.

[0034] The side column of the suspension column 105 is provided with a circular slide rail 205. The circular slide rail 205 is located close to the bottom end of the suspension column 105. The circular slide rail 205 is connected to a separation baffle plate 206. Both the circular slide rail 205 and the separation baffle plate 206 have a matching groove structure of the same size. The circular slide rail and the separation baffle plate are interlocked by the matching groove structure, so that the separation baffle plate 206 can slide in a ring around the circular slide rail 205 on the suspension column 105.

[0035] The separation barrier plate 206 has a semi-circular arc structure, and a plurality of first driven teeth 207 are provided on its arc exterior. The first driven teeth 207 mesh with the teeth on the first driving gear 208. The first driving gear 208 is fixed to the end of the first servo motor 209, and the first servo motor 209 is mounted on the suspension column 105.

[0036] It should be noted that when non-molting giant freshwater prawns gather in large numbers due to being attracted by molting individuals covered by the attraction shield 109, the first servo motor 209 is activated. The output of the first servo motor 209 causes the first drive gear 208 to rotate. With the engagement groove structure between the annular slide rail 205 and the separation barrier plate 206, the first drive gear 208, through the first driven teeth 207, further drives the separation barrier plate 206 to slide in a ring-shaped guide along the annular slide rail 205. It is worth noting that the separation barrier plate... The initial position of the separation barrier 206 is located on one side of the independent separation area of ​​the non-molting giant freshwater prawn. At this time, the separation barrier 206 is gradually rotated on the suspension column 105 from one side of the independent separation area of ​​the non-molting giant freshwater prawn to the other side, that is, rotated to the side where the molting giant freshwater prawn attracts a high concentration of non-molting giant freshwater prawns (the independent separation area of ​​the molting giant freshwater prawn). It is equivalent to opening the independent separation area of ​​the non-molting giant freshwater prawn. At this time, the attracted non-molting giant freshwater prawn individuals can enter the independent separation area through the entrance opened by the separation barrier 206. Once all non-molting giant freshwater prawns have passed through, the first servo motor 209 is reversed, similarly causing the separation barrier 206 to rotate back to its initial position. This closes the entrance to the independent separation area for non-molting giant freshwater prawns. After closing, the electromagnetic device 108 is reactivated to attract and raise the attraction protection cover 109, releasing the temporarily protected molting giant freshwater prawns. Because the independent separation area for non-molting giant freshwater prawns is closed, they can only slowly swim back to their original area (the separation area for molting giant freshwater prawns), achieving the separation of molting and non-molting giant freshwater prawns, reducing cannibalism frequency, and improving farming efficiency and quality.

[0037] It should be noted that each suspension column 105 is equipped with a baffle on its side. The width of the baffle is equal to the distance between two adjacent separation mechanisms. It can effectively prevent molting and non-molting giant freshwater prawns from swimming into the gaps between the separation mechanisms during the attraction and aggregation process, thereby reducing the probability of confusion in the separation of molting and non-molting individuals and improving the accuracy of giant freshwater prawn separation.

[0038] The top of the breeding pool 101 is equipped with a multi-dimensional separation detection module, which includes an arc-shaped guide rail 301, a pulley guide block 302 is fitted onto the arc-shaped guide rail 301, and an industrial wide-angle camera 303 is installed at the bottom of the pulley guide block 302.

[0039] The pulley guide block 302 has a second drive gear 304 installed in the middle. The second drive gear 304 is fixed to the output end of the second servo motor 305. The second drive gear 304 meshes with the second driven tooth 306, and the second driven tooth 306 is located on the outer edge of the arc-shaped guide rail 301.

[0040] It should be noted that the multidimensional separation detection module is located at the front end of the first crossbeam plate 103, thus enabling it to accurately detect whether there is aggregation behavior of giant freshwater prawns in the molting period at the front end of the separation mechanism. When this occurs, the industrial wide-angle camera 303 is activated, allowing it to take high-definition zoom wide-angle photos of the rearing pond directly below. The second servo motor 305 is started to drive the second drive gear 304 to rotate. Since the arc-shaped guide rail 301 is fixed, the rotation of the second drive gear 304 will be around the second driven tooth 306, thereby driving the pulley guide rail 302 to slide semi-circularly on the arc-shaped guide rail 301. This drives the industrial wide-angle camera 303 to slide to different circumferential positions to take multi-dimensional pictures of the giant freshwater prawns in the rearing pond from another angle. This achieves a multi-angle and all-round capture effect to attract and gather the giant freshwater prawns, providing a highly reliable and accurate image analysis basis for the subsequent separation control mechanism. This reduces the false detection and missed detection of giant freshwater prawns gathering during the molting and non-molting periods, reduces separation confusion and error rate, and improves separation reliability.

[0041] A second aspect of the present invention provides a method for separating macrophage individuals during and outside the molting stage, applicable to any of the apparatuses described in the present invention, specifically comprising the following steps: S1: Obtain the clusters of giant freshwater prawns in molting and non-molting stages to be separated, put all the clusters to be separated into the rearing pond, and simultaneously start the separation mechanism and multi-dimensional separation detection module. S2: The industrial camera of the multi-dimensional separation detection module takes all-round, multi-dimensional pictures of the cluster of giant freshwater prawns to be separated in the breeding pond to obtain the aggregated image data of individual giant freshwater prawns. S3: Using standard image features of giant freshwater prawns at different growth stages as noise terms, a variational encoder is used to learn and train the noise of the cluster center on the clustered image data based on the noise terms and perform reverse noise reduction to obtain the actual feature vector located at the cluster center. The analysis of whether the actual feature vector is a molting period feature controls the opening of the inducing protective cover to cover the giant freshwater prawn individuals during the molting period. S4: Based on the aggregation characteristics of non-molting stage shrimp attracted by molting stage shrimp at different non-molting stages, fractal processing is performed on the non-molting stage pixel feature points in the aggregated image data, and the main and secondary gradient directions are calculated to determine the fractal slope of the aggregation of non-molting stage shrimp. According to the fractal slope, the separation barrier is controlled to open another area, so that the non-molting stage shrimp individuals attracted by the molting stage shrimp individuals can enter the area. S5: Upon entry, release the covered molting macrophage individuals, thus separating molting macrophage individuals from non-molting macrophage individuals.

[0042] Furthermore, in a preferred embodiment of the present invention, step S3 specifically includes the following steps: The separation index of the molting period of giant freshwater prawn is obtained. Based on the separation index, the molting period image data of giant freshwater prawn and the text semantic prompts of the molting period image data are retrieved in the big data network. The standard image feature set of giant freshwater prawn at different growth stages is also obtained. A latent random noise space is constructed based on a standard image feature set. A pre-trained variational encoder and variational decoder are introduced. The variational encoder is used to map the clustered image data to the probability distribution matrix of the latent random noise space, and the latent vector of the cluster center is constructed based on the semantic prompts of the text. A variational decoder is used to extract the latent vector of the clustered image data during the mapping process. Gaussian noise about the standard image feature set is continuously added to the latent vector of the clustered center in the latent noise space, and the noise addition time step is recorded. The training process involves gradually injecting Gaussian noise into the aggregated image data at multiple noise-adding time steps to learn the random perturbation diffusion distribution of the aggregated images that have the standard image features of the giant freshwater prawn, and to generate a perturbation diffusion law model. A priori condition diffusion network is introduced to construct a backward denoising operator. Based on the embedded preset guidance coefficients of text semantic prompts, the backward denoising operator is used to discretize and eliminate noise in the noisy latent vector of the disturbance diffusion law model based on the guidance coefficients, and finally obtains the actual feature vector of the giant freshwater prawn located at the aggregation center in the aggregation image data. Based on big data, a set of molting period feature vectors of giant freshwater prawns is obtained. If at least one molting period feature vector in the set has a greater degree of agreement with the actual feature vector than a preset degree of agreement, then the electromagnetic device is controlled to open the protective cover to cover the molting giant freshwater prawn individuals.

[0043] It should be noted that molting-stage individuals often attract non-molting-stage individuals to form clusters. However, when the clusters reach a certain size, the density of non-molting-stage giant freshwater prawns may become too high, completely obscuring the molting-stage prawns that act as the attraction centers. Furthermore, the turbidity of the aquaculture water environment makes it difficult for industrial wide-angle cameras to determine whether molting-stage giant freshwater prawns are present in local shrimp populations based on captured images. This hinders precise decision-making regarding whether to activate the attraction shield to further confine molting-stage individuals. To address this, this method sets the standard features of non-molting-stage giant freshwater prawn images as the Gaussian noise factor to be eliminated, thus constructing a latent random noise space to provide an adaptive domain for subsequent noise removal training. Subsequently, a pre-trained variational encoder maps the clustered image data to the probability distribution matrix of this latent random noise space. Notably, the size of this latent random noise space is much smaller than the pixel space of the image, thereby preserving the high-dimensional core semantic and structural features of textual semantic cues in the clustered image data starting from the random noise latent vectors. The latent random noise space is defined by the distribution range of standard Macrobrachium rosenbergii images (non-molting stage), allowing features from different growth stages to be distinguished within the latent random noise space. This provides a noise and statistical prior for the subsequent "noise learning-denoising" process of clustering centers of non-molting features. Then, a variational decoder is used to extract latent vectors of clustering centers from the latent space and add Gaussian noise from the standard non-molting images. This learns to simulate the perturbation diffusion law of image degradation under different non-molting Macrobrachium rosenbergii clustering noise intensities, thus characterizing the degree of distortion of clustering centers in image features. Finally, the textual semantic pre-set guiding coefficients of the molting stage Macrobrachium rosenbergii image features further guide the backward denoising operator to discretize and eliminate the noisy latent vectors of the perturbation diffusion law model, enhancing the correlation search between the generated image and the target molting stage features. This achieves the extraction of the most representative true feature vectors of molting stage individuals from high-density clustered, noisy image data of non-molting Macrobrachium rosenbergii individuals.

[0044] It should be noted that if at least one molting-stage feature vector in the molting stage feature vector set has a greater degree of agreement with the actual feature vector than a preset degree of agreement, it indicates that the activity scale of non-molting-stage Macrobrachium rosenbergii at the front end of the separation mechanism shown in the aggregated image data contains an image feature with molting-stage Macrobrachium rosenbergii individuals as the aggregation center point. This indicates that these non-molting-stage Macrobrachium rosenbergii are attracted and aggregated by molting-stage individuals, hence the electromagnetic device is controlled to open the attraction protective cover upwards to cover the molting-stage Macrobrachium rosenbergii individuals. This method can perform noise learning and denoising analysis on the Macrobrachium rosenbergii cluster images in front of the separation mechanism, thereby accurately extracting the features of molting-stage individuals from the aggregation center point under the noise disturbance of a large number of non-molting-stage Macrobrachium rosenbergii individual features. This provides image basis for decision-making in subsequent separation control, reduces the separation error between molting-stage and non-molting-stage Macrobrachium rosenbergii caused by high-density noise aggregation obscuring key features, optimizes the recognition accuracy of target features, and improves the separation control accuracy of the separation mechanism.

[0045] Furthermore, in a preferred embodiment of the present invention, step S4 specifically includes the following steps: Obtain the preset resolution scheduling strategy for spatial shooting by industrial wide-angle cameras, and construct a non-spatial grid array based on the preset resolution scheduling strategy; Binary pixel conversion aggregated image data, and non-spatial grid array overlay and matte onto the aggregated image data after binary pixel conversion. At the same time, based on big data, pixel feature points of giant freshwater prawns during non-molting period and aggregation characteristics and aggregation index of prawns in different non-molting periods affected by molting period attraction are obtained. Only count the number of bins with at least one non-molting pixel feature point to obtain a number of non-molting prominent bins. Based on the aggregation characteristics and aggregation index, a scaling scale is preset. The non-molting prominent bins are scaled according to the scaling scale to obtain a series of fractal sizes of non-molting pixel feature points corresponding to prominent bins. The Prewitt operator is introduced to extract multidimensional gradients from binary pixels in aggregated image data. The horizontal and vertical gradients of each non-molting pixel feature point are obtained. The gradient covariance matrix is ​​calculated by combining the gradient magnitude and direction expressed by the horizontal and vertical gradients, and the structural tensor between two adjacent non-molting pixel feature points is obtained. The principal and secondary directions of the clustering gradient distribution of non-molting pixel features are calculated based on the structural tensor, and the principal and secondary directions of energy are obtained. Based on the principal and secondary directions of energy, each non-molting pixel feature point is fitted in the linear regression equation according to the fractal size to obtain the fractal slope of the clustering of non-molting giant freshwater prawns in the clustered image data. If the fractal slope is greater than the preset fractal slope, the non-molting stage giant freshwater prawns in the aggregated image data are calibrated as having high aggregation. The separation mechanism in the area is then activated by the corresponding first servo motor to drive the separation barrier to open another area.

[0046] It should be noted that because the attraction activities of non-molting *Macrobrachium rosenbergii* individuals at different molting stages are random, unevenly aggregated *Macrobrachium rosenbergii* clusters with varying degrees of aggregation and size will appear in the rearing pond. To improve separation efficiency, the separation mechanism of this device is activated when non-molting *Macrobrachium rosenbergii* reach a high aggregation density, thereby maximizing the number of molting and non-molting prawns separated in a single cycle and reducing the consumption and separation deviation caused by repeated device activation and deactivation. However, the control methods used in existing separation equipment are difficult to accurately analyze the aggregation degree of non-molting individuals in aggregated images. To address this, this method constructs a non-spatial grid array using a preset resolution scheduling strategy for industrial wide-angle cameras, enabling multi-scale observation of binary pixels and capturing the self-similarity complexity of hierarchical structural features. By statistically analyzing the spatial filling quantity of non-molting pixel feature points in aggregated image data at the current scale, a series of non-molting prominent binaries are obtained. These binaries reflect the changes in the density and complexity of local aggregation of non-molting Macrobrachium rosenbergii individuals as a function of aggregation characteristics and aggregation index. A higher fractal scale indicates a denser and more complex aggregation of non-molting individuals, while a lower scale indicates a relatively sparse aggregation distribution. Fractal geometry is used to characterize the spatial aggregation of shrimp groups, compensating for the inability of traditional pixel density to represent complex aggregation patterns. However, due to the subjective transformation of fractal size, judging the degree of aggregation solely based on this fractal size is often too one-sided and limited, easily leading to a misalignment of the scale benchmark for aggregation gradients. Therefore, this method further uses the Prewitt operator to calculate the horizontal and vertical gradients of binary images to obtain a structural tensor of the statistical relationship of gradients (gradient covariance matrix). This structural tensor describes the aggregation direction and energy distribution of local feature textures, quantifying the directional energy information hidden in the aggregated image data, and providing reliable guidance for judging aggregation strength and directionality. Subsequently, the directional energy of the principal direction (maximum eigenvalue) and the secondary direction (minimum eigenvalue) is calculated from the structural tensor. This method uses the principal and secondary directional energies as clustering characteristics to further fit each non-molting pixel feature point based on the fractal size, thereby fusing structural direction information with fractal complexity to establish a more stable aggregation metric for non-molting Macrobrachium rosenbergii, namely the fractal slope. This fractal slope reflects the gradient trend of local aggregation as the aggregation characteristics and aggregation index change at the corresponding scale. If the fractal slope is greater than the preset fractal slope, it indicates that the non-molting Macrobrachium rosenbergii individual aggregation density or scale at the front end of the separation mechanism is high, meeting the maximum separation requirements of this device, and the non-molting Macrobrachium rosenbergii in this aggregated image data are calibrated as having high aggregation.

[0047] It should be noted that this method can calculate the fractal geometry characterization and structural gradient tracking of texture features in images of non-molting Macrobrachium rosenbergii individuals in aggregated image data. This enables multi-dimensional analysis of the aggregation complexity of non-molting Macrobrachium rosenbergii individuals at the front end of the separation mechanism, making the aggregation density positioning of different local non-molting Macrobrachium rosenbergii individuals attracted by molting Macrobrachium rosenbergii more accurate. This improves the open control precision of the separation mechanism, maximizes the separation quantity of molting and non-molting Macrobrachium rosenbergii, and improves separation efficiency.

[0048] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A device for separating individuals of giant freshwater prawns during and outside the molting stage, the device comprising a rearing tank, characterized in that: The feeding pool is a rectangular trough-shaped pool with multiple support legs at the bottom. A first crossbeam plate is fixed to both ends of the top edge of the feeding pool. A second crossbeam plate is located directly below the first crossbeam plate. The two ends of the second crossbeam plate are welded to the inner wall of the feeding pool. Several separation mechanisms are installed on the second crossbeam plate. The separation mechanism includes a suspension column, which is welded to the bottom of the second crossbeam plate by several hanging columns. The suspension column has a hollow structure inside, which makes the whole device lightweight. A through hole is opened at the top center for docking through installation of the cover drive column. An extension rod is provided at the top of the cover drive column. The end of the extension rod passes through the second crossbeam plate and is fixed with an electromagnetic device. The side column of the suspension column is provided with a circular slide rail. The circular slide rail is located close to the bottom of the suspension column. The circular slide rail is connected to a separation baffle plate. Both the circular slide rail and the separation baffle plate have a matching groove structure of the same size. The circular slide rail and the separation baffle plate are interlocked with each other through the matching groove structure, so that the separation baffle plate can slide in a ring around the circular slide rail on the suspension column. The top of the breeding pool is equipped with a multi-dimensional separation detection module, which includes an arc-shaped guide rail, a pulley guide block is fitted and connected on the arc-shaped guide rail, and an industrial wide-angle camera is installed at the bottom of the pulley guide block; The first crossbeam and the second crossbeam are arranged in parallel alignment on the feeding pool, and a relative distance is maintained between the first crossbeam and the second crossbeam. The covering drive column is made of transparent PVC polyvinyl chloride material. An attraction protection cover is provided at the bottom of the covering drive column. Several drainage holes are opened on the side of the attraction protection cover to cover and protect the giant freshwater prawns in the molting period and to attract giant freshwater prawns in the non-molting period to gather under the suspension column. Several circumferentially arranged turbulence teeth are provided on the bottom edge of the attraction protection cover. A return spring is sleeved on the outside of the extension rod. One end of the return spring is fixed to the top of the suspension column, and the other end is fixed to the bottom of the second crossbeam plate. Several metal blocks are arranged at the bottom of the first crossbeam plate. The number and arrangement of the metal blocks are consistent with the separation mechanism, and each metal block is located directly above the electromagnetic device corresponding to each separation mechanism.

2. The device for separating molting and non-molting individuals of Macrobrachium rosenbergii according to claim 1, characterized in that: The lure protection cover is equipped with a sonar imaging sensor, which is used to detect the phenotypic characteristics of the giant freshwater prawn during the molting period and its relative position in the area protected by the lure protection cover.

3. The device for separating molting and non-molting individuals of Macrobrachium rosenbergii according to claim 1, characterized in that: The separation barrier plate has a semi-circular arc structure, and a number of first driven teeth are provided on the outer side of the arc. The first driven teeth mesh with the teeth on the first driving gear. The first driving gear is fixed to the end of the first servo motor, and the first servo motor is mounted on the suspension column.

4. The device for separating molting and non-molting individuals of Macrobrachium rosenbergii according to claim 1, characterized in that: A second driving gear is installed in the middle of the pulley guide block. The second driving gear is fixed to the output end of the second servo motor. The second driving gear meshes with the second driven gear, and the second driven gear is located on the outer edge of the arc-shaped guide rail.

5. A method for separating individuals of *Macrobrachium rosenbergii* during and outside the molting stage, applied to the apparatus for separating individuals of *Macrobrachium rosenbergii* during and outside the molting stage as described in any one of claims 1-4, characterized in that... Specifically, the following steps are included: S1: Obtain the clusters of giant freshwater prawns in molting and non-molting stages to be separated, put all the clusters to be separated into the rearing pond, and simultaneously start the separation mechanism and multi-dimensional separation detection module. S2: The industrial camera of the multi-dimensional separation detection module takes all-round, multi-dimensional pictures of the cluster of giant freshwater prawns to be separated in the breeding pond to obtain the aggregated image data of individual giant freshwater prawns. S3: Using standard image features of giant freshwater prawns at different growth stages as noise terms, a variational encoder is used to learn and train the noise of the cluster center on the clustered image data based on the noise terms and perform reverse noise reduction to obtain the actual feature vector located at the cluster center. The analysis of whether the actual feature vector is a molting period feature controls the opening of the inducing protective cover to cover the giant freshwater prawn individuals during the molting period. S4: Based on the aggregation characteristics of non-molting stage shrimp attracted by molting stage shrimp at different non-molting stages, fractal processing is performed on the non-molting stage pixel feature points in the aggregated image data, and the main and secondary gradient directions are calculated to determine the fractal slope of the aggregation of non-molting stage shrimp. According to the fractal slope, the separation barrier is controlled to open another area, so that the non-molting stage shrimp individuals attracted by the molting stage shrimp individuals can enter the area. S5: Upon entry, release the covered molting giant freshwater prawns, thus separating the molting and non-molting giant freshwater prawns.

6. The method for separating individuals of the giant freshwater prawn during and outside the molting stage according to claim 5, characterized in that, S3 specifically includes the following steps: The separation index of the molting period of giant freshwater prawn is obtained. Based on the separation index, the molting period image data of giant freshwater prawn and the text semantic prompts of the molting period image data are retrieved in the big data network. The standard image feature set of giant freshwater prawn at different growth stages is also obtained. A latent random noise space is constructed based on a standard image feature set. A pre-trained variational encoder and variational decoder are introduced. The variational encoder is used to map the clustered image data to the probability distribution matrix of the latent random noise space, and the latent vector of the cluster center is constructed based on the semantic prompts of the text. A variational decoder is used to extract the latent vector of the clustered image data during the mapping process. Gaussian noise about the standard image feature set is continuously added to the latent vector of the clustered center in the latent random noise space, and the noise addition time step is recorded. The training process involves gradually injecting Gaussian noise into the aggregated image data at multiple noise-adding time steps to learn the random perturbation diffusion distribution of the standard image features of giant freshwater prawn in the aggregated images, and to generate a perturbation diffusion law model. A priori condition diffusion network is introduced to construct a backward denoising operator. Based on the embedded preset guidance coefficients of text semantic prompts, the backward denoising operator is used to discretize and eliminate noise in the noisy latent vector of the disturbance diffusion law model based on the guidance coefficients, and finally obtains the actual feature vector of the giant freshwater prawn located at the aggregation center in the aggregated image data. Based on big data, a set of molting period feature vectors of giant freshwater prawns is obtained. If at least one molting period feature vector in the set has a greater degree of agreement with the actual feature vector than a preset degree of agreement, then the electromagnetic device is controlled to open the protective cover to cover the molting giant freshwater prawn individuals.

7. The method for separating molting and non-molting individuals of Macrobrachium rosenbergii according to claim 5, characterized in that, S4 specifically includes the following steps: Obtain the preset resolution scheduling strategy for spatial shooting by industrial wide-angle cameras, and construct a non-spatial grid array based on the preset resolution scheduling strategy; Binary pixel conversion aggregated image data, and non-spatial grid array overlay and matte onto the aggregated image data after binary pixel conversion. At the same time, based on big data, pixel feature points of giant freshwater prawns during non-molting period and aggregation characteristics and aggregation index of prawns in different non-molting periods affected by molting period attraction are obtained. Only count the number of bins with at least one non-molting pixel feature point to obtain a number of non-molting prominent bins. Based on the aggregation characteristics and aggregation index, a scaling scale is preset. The non-molting prominent bins are scaled according to the scaling scale to obtain a series of fractal sizes of non-molting pixel feature points corresponding to prominent bins. The Prewitt operator is introduced to extract multidimensional gradients from binary pixels in aggregated image data. The horizontal and vertical gradients of each non-molting pixel feature point are obtained. The gradient covariance matrix is ​​calculated by combining the gradient magnitude and direction expressed by the horizontal and vertical gradients, and the structural tensor between two adjacent non-molting pixel feature points is obtained. The principal and secondary directions of the clustering gradient distribution of non-molting pixel features are calculated based on the structural tensor, and the principal and secondary directions of energy are obtained. Based on the principal and secondary directions of energy, each non-molting pixel feature point is fitted in the linear regression equation according to the fractal size to obtain the fractal slope of the clustering of non-molting giant freshwater prawns in the clustered image data. If the fractal slope is greater than the preset fractal slope, the non-molting stage giant freshwater prawns in the aggregated image data are calibrated as having high aggregation. The separation mechanism in the control area is activated by the first servo motor to drive the separation barrier to open another area.

Images