An integrated device and method for intelligent sea cucumber pond inversion, growth monitoring and prediction

Abstract

The present invention provides an integrated device and method for intelligent sea cucumber pond emptying, growth monitoring and prediction, which organically combines the sea cucumber pond emptying process with sea cucumber growth monitoring. During the sea cucumber pond emptying process, a conveyor belt arranged between the original pond and the new pond is used to transport the sea cucumbers, and an image acquisition module is arranged on the conveyor belt to obtain sea cucumber images. The sea cucumbers are identified by using a pre-built visual recognition model through the image recognition module, and an identification frame is drawn with the longest and widest parts of the sea cucumber edge as critical points. The data processing module calculates the actual size of the identification frame. The prediction model detects and predicts the sea cucumber growth trait data based on the relationship between the actual size of the identification frame and the growth traits of the sea cucumber. Multiple fitting models with the highest confidence and confidence greater than a threshold are selected, and the average value of the fitting models is calculated to obtain the growth monitoring data of the sea cucumber size and the prediction data of the wet weight and dry weight, thereby realizing intelligent monitoring of the growth status of the sea cucumber.

Description

Technical Field

[0001] The present invention relates to the technical field of sea cucumber farming, and provides a device and method for integrating intelligent sea cucumber pond inversion, growth monitoring and prediction. Background Art

[0002] Sea cucumbers have become a valuable fishery resource due to their unique nutritional value and market demand. With economic development and increased acceptance in the future, the demand for sea cucumber products among Chinese residents will further increase. During the sea cucumber farming process, sea cucumbers are kept in the breeding ponds and need to be regularly drained. The main purpose is to clean the sediment at the bottom of the pond, improve the breeding environment, prevent the occurrence of diseases, and provide a healthier and cleaner growth environment for sea cucumbers. The time interval for draining sea cucumber ponds is usually controlled by factors such as the water temperature, feed feeding, and the amount of food and excretion consumed by sea cucumbers. During the normal feeding period of sea cucumbers, the ponds usually need to be drained once every 1-2 weeks. If the sea cucumbers eat a lot or the water quality deteriorates, the draining cycle needs to be further shortened.

[0003] During the sea cucumber farming process, it is necessary to regularly monitor the growth status of sea cucumbers to determine whether their growth conditions are in line with expectations. If abnormal conditions such as growth stagnation are found, timely measures can be taken such as adjusting feed formulas, improving water quality, etc. to promote the normal growth of sea cucumbers.

[0004] Currently, monitoring the growth status of sea cucumbers usually requires visual inspection by staff, which has a low level of intelligence and a large workload. Regular pond emptying is a necessary part of the sea cucumber farming process. If the sea cucumber pond emptying process can be combined with the sea cucumber growth status detection process, the workload of sea cucumber farming can be reduced. Summary of the Invention

[0005] In order to solve the problems existing in the prior art, the first object of the present invention is to provide an integrated device for intelligent sea cucumber pond inversion and growth monitoring and prediction, which combines the sea cucumber pond inversion process with the sea cucumber growth monitoring process, including:

[0006] Aquaculture ponds, comprising original ponds currently culturing sea cucumbers and new ponds for replacement;

[0007] A conveyor belt is provided between the original pool and the new pool for transporting sea cucumbers;

[0008] An image acquisition module, the image acquisition module being arranged on the conveyor belt and being used for acquiring images containing sea cucumbers;

[0009] An image recognition module is configured to obtain an image containing a sea cucumber, identify the sea cucumber using a pre-built visual recognition model, draw an identification frame using the longest and widest points of the sea cucumber edge as critical points, record the size of the identification frame, sequentially sort each identification frame within the identification area, and count the total number of sea cucumber individuals;

[0010] A data processing module, wherein the data processing module obtains the size of the recognition frame and calculates the actual size of the recognition frame;

[0011] A prediction module obtains the actual size of the identification frame, uses a pre-built prediction model, takes the actual size of the sea cucumber identification frame as a prediction parameter, selects multiple fitting models with the highest confidence and a confidence greater than a threshold, and calculates the average value of the fitting models to obtain monitoring data of sea cucumber size and predicted data of wet weight and dry weight.

[0012] Specifically, the conveyor belt is a U-shaped conveyor belt, which is arranged on the outside of the breeding pond. The original pond and the new pond are spaced apart. A fishing mechanism is provided in the breeding pond. The fishing mechanism includes a transverse track, a movable frame and a fishing shovel. The transverse track is provided on the breeding pond. A movable frame that can move along the transverse track is provided on the transverse track. A fishing shovel that can be connected to the movable frame by a lifting rope is provided in the breeding pond.

[0013] Specifically, the mobile frame includes a longitudinal track arranged on the transverse track and a frame body installed on the longitudinal track, a rack is provided on one side of the longitudinal track, a driving motor fixedly installed in the frame body is provided above the rack, a gear is provided on the output shaft of the driving motor, the gear is engaged in the rack, a first rope winding shaft and a second rope winding shaft are provided in the frame body, two rope winding drums are provided in each of the first rope winding shaft and the second rope winding shaft, a lifting rope that can be connected to the fishing shovel is wound in the rope drum, and a first lifting motor and a second lifting motor are provided at the ends of the first rope winding shaft and the second rope winding shaft respectively.

[0014] Specifically, the fishing shovel includes a bottom plate, a rear plate and side plates. An opening is provided on the front side of the fishing shovel. The width of the fishing shovel matches the width of the breeding pond. A plurality of drainage holes are provided in the bottom plate of the fishing shovel, a plurality of water filtering holes are provided in the rear plate of the fishing shovel, and a hanging ring is provided on the top of the fishing shovel.

[0015] A second object of the present invention is to provide a method for integrating intelligent sea cucumber pond inversion with growth monitoring and prediction, comprising:

[0016] The sea cucumbers are caught and placed on a conveyor belt, and the calibrated image acquisition module captures images containing the sea cucumbers;

[0017] The pre-built visual recognition model recognizes the sea cucumber image, draws a recognition frame with the longest and widest points of the sea cucumber edge as critical points, and records the size of the recognition frame;

[0018] Obtain the actual length and width of a single pixel in the image acquisition module, calculate the actual size of the identification frame, and sort each identification frame in turn to count the total number of sea cucumber individuals;

[0019] The actual size of the identification frame is loaded as a prediction parameter into a pre-built prediction model. The prediction model monitors and predicts the sea cucumber growth trait data based on the relationship between the actual size of the identification frame and the growth traits of the sea cucumber. Multiple fitting models with the highest confidence and a confidence greater than a threshold are selected, and the average value of the fitting model is calculated to obtain the monitoring data of the sea cucumber size and the prediction data of the wet weight and dry weight.

[0020] Specifically, the image recognition module uses an improved YOLOv8-seacucumber visual recognition model to identify sea cucumbers, identifies targets with confidence levels greater than a threshold as sea cucumbers, and draws an identification box. The YOLOv8-seacucumber visual recognition model includes an input end, a backbone network, and a detection head. Based on the original YOLOv8 visual recognition model, the C2f-F-SA module is used to replace the C2f module in the backbone network and the detection head. The C2f-F-SA module includes a convolutional layer, a segmentation layer, multiple FasterBlock structures, a connection layer, and a convolutional layer arranged in sequence. The bottleneck layer in the original C2f module is replaced by a FasterBlock structure, and a Shuffle Attention mechanism is added to the FasterNet Block module. Some convolutional layers in the detection head are replaced by Adown modules.

[0021] Specifically, the Shuffle Attention mechanism includes an input end, a segmentation layer, an upper branch, a lower branch, a connection layer, and an output end;

[0022] Input sea cucumber characteristics Figure X ∈R (C，H，W) , the sea cucumber feature maps are divided into G groups according to the channel dimension, which can be expressed as:

[0023] X=[X1,…,X G ], X k ∈R (C / G×H×W) ,

[0024] Where C is the number of channels of the sea cucumber feature map, H is the spatial height of the sea cucumber feature map, W is the spatial width of the sea cucumber feature map, R represents the set of all images in the dataset, X k is the input feature;

[0025] The segmentation layer takes as input the feature Xk Divided into upper branch X along the channel dimension K1 With the lower branch X K2 , X K1 , X K2 ∈R (C / 2G×H×W) , the upper branch adopts channel attention, and generates channel information S by scaling and shifting the transformation parameters w1, b1. First, global average pooling (GAP) is performed to embed global information to generate channel information S∈R (C / 2G×1×1) , the channel information is obtained by k1 Calculate by shrinking on the spatial dimension H×W,

[0026]

[0027] Use the fully connected layer to transform the channel information S and apply the σ activation function (sigmoid) to generate the weight matrix and multiply it by X k1 , and get the weighted features Figure X ' x1 :

[0028] X′ k1 =σ(F C (S))·X k1 =σ(w1S+b1)·X k1 ,

[0029] Where w1∈R (C / 2G×1×1) , b1∈R (C / 2G×1×1) ;

[0030] The lower branch uses spatial attention to capture the spatial dependencies between features, scales and shifts the amount of spatial information by transforming parameters w2 and b2, and uses the group norm (GN) to generate statistics of spatial attention:

[0031] X′ k2 =σ(w2·GN(X k2 )+b2)·X k2 ,

[0032] Where w2∈R (C / 2G×1×1) , b2∈R (C / 2G×1×1) ;

[0033] The connection layer processes X′ obtained by the upper and lower branches K1 and X′ K2 Combined, the output feature map has the same size as the input,

[0034] X′ k =[X′ k1 , X′ k2 ]R (C / G×H×W) .

[0035] Specifically, the ADown module includes an input end, an average pooling layer, a segmentation layer, a first path, a second path, a connection layer, and an output end.

[0036] Input characteristics Figure X ∈R (C，H，W) , the average pooling layer takes the input features Figure X The average pooling feature is obtained by performing an average pooling operation through an average pooling layer with a pooling kernel k of size 2, a step size s of 1, and no padding (p = 0). Figure X avg ,

[0037] X avg =Avgpool(X, k=2, s=1, p=0),

[0038] The segmentation layer averages the pooled features Figure X avg Split into two sub-features along the channel dimension Figure X 1 and X2, sub-features Figure X 1 and X2 are downsampled simultaneously through the first path and the second path respectively to obtain feature maps Y1 and Y2.

[0039] The first path is,subfeature Figure X 1 directly downsamples through a traditional convolutional layer to obtain the feature map Y1,

[0040] Y1=Conv1(X1, k=3, s=2, p=1);

[0041] The second path is,subfeature Figure X 2 First, a maximum pooling layer with a pooling kernel size k of 3, a step size s of 2, and a padding p of 1 is used to extract the maximum value X2 in the local window and highlight the significant features. Subsequently, the feature map is further extracted and refined through a convolution layer with a convolution kernel size k of 1, a step size s of 1, and a padding p of 0 to obtain the feature map Y2.

[0042] X 2，max =MaxPool(X2, k=3, s=2, p=1),

[0043] Y2=Conv2(X 2，max , k=1, s=1, p=0),

[0044] Two learnable weight factors α and β are introduced in the connection layer. Through the training process optimization, the downsampling results are weighted fused. The weight factors α and β control the contribution of Y1 and Y2 in the fusion process respectively. The dimensions of the weight factors α and β are both 1×1×1. The feature map Y after weighted fusion is fusion for:

[0045] Yfusion =α·Y1+β·Y2,

[0046] In the formula, · represents element-by-element multiplication,

[0047] Normalize the weight factors α and β to ensure that their sum is 1.

[0048]

[0049] The weighted fusion feature map is recalculated using the normalized weight factor:

[0050] Y fusion =α norm Y1+β norm Y2,

[0051] From the output end, the weighted fusion feature map Y fusion As output, its dimension is Where C' is the number of output channels.

[0052] Specifically, the YOLOv8-seacucumber visual recognition model construction method is as follows: sea cucumbers are grouped according to their region and species, and sampling training is performed, photos of each sea cucumber are obtained under different shooting angles, different lighting and different backgrounds under natural conditions, and the pictures are proportionally divided into training sets and test sets. The sea cucumbers in the training set are manually labeled using bounding boxes, and a convolutional neural network is used to train a model for all labeled sea cucumber image information. The model is tested using a test set, and the training results are manually reviewed to obtain a visual recognition model.

[0053] Specifically, the prediction model construction method is: based on the marked sea cucumber images and sea cucumber growth trait data, the actual size of the identification box of the same species of sea cucumber in the same area is fitted with its growth trait data, and multiple regression models and machine learning are used to perform model training to obtain a prediction model.

[0054] Compared with the prior art, the present invention has the following beneficial effects:

[0055] 1. The present invention organically combines the process of sea cucumber ponding with sea cucumber growth monitoring, monitors the growth status of sea cucumbers during the process, and provides intelligent sea cucumber ponding equipment, which can automatically capture sea cucumbers and put them on a conveyor belt. The improved YOLOv8-seacucumber visual recognition model is used to identify sea cucumbers, and an identification frame is drawn with the longest and widest points of the sea cucumber edge as critical points. The real size of the identification frame is calculated based on the size of the identification frame, and the real size of the identification frame is used as a prediction parameter. A pre-built prediction model is used to select multiple fitting models to obtain growth monitoring data of sea cucumber size and prediction data of wet weight and dry weight, thereby realizing intelligent monitoring of the growth status of sea cucumbers and effectively reducing the workload of sea cucumber farming.

[0056] 2. The present invention adopts an improved YOLOv8-seacucumber visual recognition model to identify sea cucumbers. On the basis of the original YOLOv8 visual recognition model, the C2f-F-SA module is used to replace the C2f module in the backbone network and the detection head. The C2f-F-SA module includes a convolutional layer, a segmentation layer, a FasterBlock structure, a connection layer and a convolutional layer arranged in sequence. The bottleneck layer in the original C2f module is replaced by the FasterBlock structure to improve the feature extraction efficiency of the network and reduce the computational complexity in the backbone network, thereby more effectively utilizing the computing resources of the device and further enhancing the ability of the backbone network in spatial feature extraction. It can effectively extract the key features of sea cucumbers in the image and enhance the model's perception of the input sea cucumber spatial features. At the same time, it can reduce the amount of calculation and increase the detection speed, thereby improving the overall performance of the model in the sea cucumber image feature extraction task. The Shuffle Attention mechanism is added to the FasterNet Block module, and the convolutional layer of the detection head is replaced with the Adown module to prevent potential information loss during the downsampling process and improve the model's sensitivity to detailed features. At the same time, a weight factor learning mechanism is introduced to achieve more refined downsampling result fusion. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 This is a schematic diagram of the overall structure of the first embodiment of the present invention;

[0058] Figure 2 This is a structural diagram of the fishing mechanism according to the first embodiment of the present invention;

[0059] Figure 3 This is a schematic structural diagram of a frame and a fishing shovel according to a first embodiment of the present invention;

[0060] Figure 4 This is a schematic diagram of the overall structure of the second embodiment of the present invention;

[0061] Figure 5Demarcate the chessboard pattern for the present invention;

[0062] Figure 6 This is a structural diagram of the YOLOv8-seacucumber visual recognition model of the present invention;

[0063] Figure 7 This is a structural block diagram of the C2f-F-SA module of the present invention;

[0064] Figure 8 This is a structural block diagram of the FasterNet Block structure of the present invention;

[0065] Figure 9 This is a structural diagram of the Shuffle Attention mechanism of the present invention;

[0066] Figure 10 This is a structural block diagram of the ADown module of the present invention;

[0067] Figure 11 This is the recognition effect diagram of the YOLOv8-seacucumber visual recognition model of the present invention;

[0068] Figure 12 It is the fitting model of the sea cucumber mass and the length and width of the identification frame of the present invention.

[0069] Figure numerals: 1. breeding pond; 11. original pond; 12. new pond; 2. conveyor belt; 21. mounting frame; 22. camera; 23. partition plate; 3. fishing mechanism; 31. horizontal track; 32. fishing shovel; 321. bottom plate; 322. side plate; 323. back plate; 324. drain hole; 325. filter hole; 326. lifting ring; 33. movable frame; 331. longitudinal track; 332. frame; 333. rack; 334. gear; 335. driving motor; 336. first rope winding shaft; 337. second rope winding shaft; 338. first lifting motor; 339. second lifting motor; 3310. rope winding drum; 3311. lifting rope; 4. guide slope; 5. guide rail. DETAILED DESCRIPTION

[0070] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0071] A device integrating intelligent pond emptying, growth monitoring and prediction of sea cucumbers comprises a breeding pond 1, a conveyor belt 2, an image acquisition module and a host computer arranged on the conveyor belt 2. The image acquisition module takes photos of sea cucumbers transported by the conveyor belt 2 during the pond emptying process, obtains the size of the sea cucumbers to monitor their growth status, and uses a constructed prediction model to predict the dry weight and wet weight of the sea cucumbers.

[0072] As a first embodiment of the present invention, Figure 1-Figure 3 As shown, the integrated intelligent sea cucumber pond inversion and growth monitoring and prediction device includes a breeding pond 1, which comprises an original pond 11 and a new pond 12. Due to long-term use, the water quality in the original pond 11 has deteriorated, and pollutants such as sea cucumber feces and leftover bait may have settled and adsorbed on the bottom and walls of the pond. The new pond 12 is used to receive sea cucumbers harvested from the original pond 11, providing a clean cultivation environment. A harvesting mechanism 3 is provided within the breeding pond 1. A U-shaped conveyor belt 2 is provided on the outside of the breeding pond 1, connecting the original pond 11 and the new pond 12. An image acquisition module is provided on the conveyor belt 2. A host computer for processing data and information is also provided.

[0073] The fishing mechanism 3 includes a transverse track 31, a fishing shovel 32, and a mobile frame 33. The transverse track 31 is set on the wall of the aquaculture pond 1. The mobile frame 33 can move along the transverse track 31. The transverse track 31 is used to transport the mobile frame 33. The mobile frame 33 moves along the transverse track 31 to the original pond 11 to be emptied. The mobile frame 33 is set on the transverse track 31, and the fishing shovel 32 is set in the aquaculture pond 1.

[0074] The shovel 32 includes a bottom plate 321, a back plate 323, and side plates 322. Its front side has an opening, and both the bottom plate 321 and the side plates 322 are provided with inclined surfaces. The width of the shovel 32 matches the width of the aquaculture pond 1. As the shovel 32 moves along the aquaculture pond 1, it scoops up and collects sea cucumbers adsorbed on the bottom and walls of the pond. The bottom plate 321 of the shovel 32 is provided with multiple drainage holes 324. A drainage trough is provided at the angle between the bottom plate 321 and the back plate 323 to drain water from the shovel 32 when it is lifted. The back plate 323 is provided with multiple water filter holes 325 to balance the front and rear water pressure of the shovel 32 as it moves within the aquaculture pond 1. Lifting rings 326 are provided at the four corners of the top of the shovel 32. The shovel 32 is connected to the mobile frame 33 located above the aquaculture pond 1 via lifting ropes 3311. In the normal breeding state of sea cucumbers, the catching shovel 32 is placed at one end of the breeding pond 1 , and when the pond is turned over, the hanging ring 326 is connected to the movable frame 33 to catch the sea cucumbers cultured in the breeding pond 1 .

[0075] The movable frame 33 includes a longitudinal track 331, which is disposed on the transverse track 31. A frame 332 is mounted on the longitudinal track 331, which can be used to push the movable frame 333 to move it above the pool to be emptied. A rack 333 is disposed on one side of the longitudinal track 331. A drive motor 335 is fixedly mounted in the frame 332 above the rack 333. A gear 334 is fixedly mounted on the output shaft of the drive motor 335. The gear 334 engages with the rack 333. The drive motor 335 causes the frame 332 to move along the longitudinal track 331. Within the frame 332, a first rope reel 336 and a second rope reel 337 are located above the lifting ring 326. A rope reel 3310 is located within the first and second rope reels 336, 337 at locations corresponding to the lifting ring 326. A rope 3311 is wound around the rope reel 3310. One end of the rope 3311 is located within the reel, and the other end is connected to the shovel 32. A first lifting motor 338 and a second lifting motor 339 are located at the ends of the first and second rope reels 336, 337, respectively. These motors control the rotation of the first and second rope reels 336, 337, respectively. The first and second lifting motors 338, 339 control the raising and lowering and tilting of the shovel 32.

[0076] During use, the driving motor 335 drives the frame 332 to move along the breeding pond 1, and the frame 332 pulls the fishing shovel 32 from one end of the breeding pond 1 to the other end. The sea cucumbers on the bottom and wall of the pond are all caught in the breeding pond 1 into the fishing shovel 32, and the fishing shovel 32 is lifted by the first lifting motor 338 and the second lifting motor 339. First, the first lifting motor 338 controls the first rope shaft 336 to rotate, lifting the front side of the fishing shovel 32 to prevent the sea cucumbers in the fishing shovel 32 from falling. Then, the first lifting motor 338 and the second lifting motor 339 move together to lift the fishing shovel 32 from the breeding pond 1 and lift it until the height of the fishing shovel 32 matches the height of the conveyor belt 2. Finally, the second lifting motor 339 controls the rear side of the movable frame 33 to lift, so that the sea cucumbers collected in the fishing shovel 32 are slowly dumped onto the conveyor belt 2, thereby completing the harvesting of the sea cucumbers.

[0077] Conveyor belt 2 is a U-shaped conveyor belt 2 that can rotate forward and reverse, adjusting its direction according to the position of the original pool 11 and the new pool 12. Conveyor belt 2 is installed on the ground outside of the culture pond 1, with the original pool 11 and the new pool 12 spaced apart. The input end of conveyor belt 2 is located near the original pool 11, and the output end of conveyor belt 2 is located near the new pool 12. Casters are installed at the bottom of conveyor belt 2 to facilitate its movement. A mounting frame 21 is located in the middle of conveyor belt 2, and an image acquisition module and a fill light are mounted in mounting frame 21. The image acquisition module is a camera 22, located at a known height from conveyor belt 2. Camera 22 captures images of sea cucumbers transported by conveyor belt 2. There are partition columns on both sides of the conveyor belt 2. According to the movement direction of the conveyor belt, a partition plate 23 is installed in the partition column on the output side of the conveyor belt to prevent the sea cucumbers from being stacked when the catching shovel 32 dumps the sea cucumbers onto the conveyor belt 2, which affects the image recognition of the sea cucumbers. A plurality of mounting grooves are provided in the vertical direction in the partition column, and the partition plate 23 is engaged in the mounting groove. The height of the partition plate can be adjusted according to the size of the sea cucumbers cultured in the breeding pond 1.

[0078] A guide ramp 4 is provided at the output end of the conveyor belt. The guide ramp 4 can be engaged with the track or installed at the output end of the conveyor belt 2. One end of the guide ramp is connected to the output end of the conveyor belt 2, and the other end extends to the bottom of the new tank 12. The sea cucumbers transported by the conveyor belt 2 slide along the guide ramp 4 into the new tank 12, preventing the sea cucumbers from falling and getting injured.

[0079] As a second embodiment of the present invention, Figure 4 As shown, the original pond 11 and the new pond 12 are spaced apart, and a conveyor belt 2 is installed above the culture pond 1 between the original pond 11 and the new pond 12. A guide rail 5 is provided on the culture pond 1 to facilitate the movement of the conveyor belt 2. The conveyor belt 2 is mounted on the guide rail 5 and can be pushed along the guide rail 5. A mounting frame 21 is located in the middle of the conveyor belt 2. Mounting frame 21 houses an image acquisition module and a fill light. The image acquisition module is a camera 22, which is positioned at a predetermined height from the conveyor belt 2. A guide ramp 4 is provided at the end of the conveyor belt 2. Workers use a net bag to scoop sea cucumbers onto the conveyor belt 2, and the camera 22 takes photos of the sea cucumbers as they are transported along the conveyor belt 2.

[0080] Before shooting, the camera is calibrated using Zhang's Calibration Method to obtain the camera's actual external and internal parameters. The camera's actual internal parameters include the camera sensor length, camera sensor width, image length pixel number, image width pixel number, focal length, and distortion coefficient. The camera's actual external parameters include: horizontal and vertical viewing angles. The specific steps for camera calibration are:

[0081] 1. Prepare the calibration plate:

[0082] A high-precision checkerboard pattern of known size is used, such as Figure 5 As shown, the checkerboard pattern is a real-sized square block with each square alternating between black and white.

[0083] 2. Collect calibration images:

[0084] Fix the camera in one position and ensure that it does not move during the entire calibration process. Change the position and angle of the calibration plate and take multiple images from different perspectives. Take at least 10 images to obtain more accurate results.

[0085] During the shooting process, ensure that different parts of the chessboard are in the camera's field of view; ensure that the calibration plate is shot at different angles and distances to increase diversity and obtain rich projection information; ensure that there is no obvious blur or distortion in the image, maintain clarity, and that the corners are clearly visible.

[0086] 3. Extract checkerboard corner points:

[0087] In the captured image, the checkerboard corner point extraction algorithm in OpenCV is used to extract the inner corner points of the checkerboard in each image (that is, the four corner points of each small square). The position of the corner point of each checkerboard is used as the input data for calibration. These corner points will be used as the geometric relationship data between the camera and the calibration plate. The extraction steps are as follows:

[0088] (1) Corner detection: Use the corner detection algorithm, i.e. cv::findChessboardCorners in OpenCV, to automatically identify the corners of the chessboard from the image;

[0089] (2) Sub-pixel corner positioning: To improve accuracy, the cv::cornerSubPix function in OpenCV is used to accurately locate corners at the sub-pixel level;

[0090] (3) Storing corner point coordinates: The corner points of each image need to be saved and converted into two-dimensional coordinates in the image coordinate system.

[0091] 4. Construct the calibration matrix:

[0092] Convert the corner coordinates detected in each image into coordinates in the world coordinate system, and construct the camera's intrinsic parameter matrix (including parameters such as focal length and principal point) and extrinsic parameter matrix (including rotation and translation vectors);

[0093] The mathematical model of Zhang's calibration is:

[0094] Assume that point P in the world coordinate system w =(X, Y, Z) T , point P in the camera coordinate system c =(X′, Y′, Z′)T , its projected coordinates on the image plane are p = (x, y) T Transformed by the camera projection matrix P:

[0095]

[0096] Where K is the camera's intrinsic parameter matrix, [R|t] is the camera's extrinsic parameter matrix, where R is the rotation matrix, t is the translation vector, and (x, y) is the image coordinate.

[0097] The relationship between image coordinates (x, y) and normalized coordinates (x', y') is:

[0098]

[0099] Where (c x , c y ) is the principal point coordinate, f x and f y is the focal length in the x and y directions;

[0100] The distortion model is:

[0101] x distorted =x′(1+k1r 2 +k2r 4 +k3r 6 )+2p1x′y′+p2(r 2 +2x′ 2 ),

[0102] y distorted =y′(1+k1r 2 +k2r 4 +k3r 6 )+p1(r 2 +2y′ 2 )+2p2x′y′,

[0103] where r 2 =x′ 2 +y′ 2 , k1, k2, k3 are radial distortion coefficients, p1, p2 are tangential distortion coefficients;

[0104] The reprojection error is:

[0105]

[0106] where p i is the observation point in the image, P i is a point in the world coordinate system, K is the internal parameter matrix, R i and t i is the external parameter matrix of the i-th image;

[0107] During the calibration process, the Levenberg-Marquardt algorithm is used to minimize the error between the actual corner points and the reprojected points in each image, thereby estimating the camera's intrinsic and extrinsic parameters. This is then iteratively optimized, continuously adjusting the intrinsic and extrinsic parameters to minimize the overall reprojection error.

[0108] 5. Calculate camera parameters:

[0109] After calibration, the following important parameters are obtained:

[0110] Camera Intrinsic Parameters:

[0111]

[0112] c x and c y are the principal point coordinates, usually close to the center of the image;

[0113] Distortion Coefficients: including radial distortion coefficients (k1, k2, k3) and tangential distortion coefficients (P1, P2);

[0114] Camera extrinsic parameters: including the rotation matrix R and translation vector t of each image, used to transform points in the world coordinate system to the camera coordinate system;

[0115] Reprojection Error: It refers to the error between the theoretical projection and the actual corner position in the image.

[0116] 6. Reprojection error analysis and result verification:

[0117] Using the obtained camera parameters, the corner coordinates in the world coordinate system are projected back to the image plane, and the error between the reprojected points and the actual detected corner points is calculated to evaluate the accuracy of the calibration. Finally, the calibration results are verified by placing the calibration plate at a known position and reconstructing its position in 3D space using the parameters obtained from the calibration.

[0118] Since the calibrated camera is installed at a fixed position on the conveyor belt, the camera parameters and the distance between the camera and the conveyor belt are fixed. By determining the internal and external parameters of the camera through calibration, the actual length and width of a single pixel in the photo taken by the calibrated camera can be obtained.

[0119] The host is connected to the camera for communication, and the photos taken by the camera are transmitted to the host. The host includes an image recognition module, a data processing module and a prediction module.

[0120] The image recognition module obtains images containing sea cucumbers and uses a pre-built visual recognition model to identify the sea cucumbers. Sea cucumbers with the same feature points in consecutive images are recorded only once.

[0121] like Figure 6 As shown, the image recognition module uses an improved YOLOv8-seacucumber visual recognition model to identify sea cucumbers and draw an identification frame. The improved YOLOv8-seacucumber visual recognition model includes an input terminal (Input), a backbone network (Backbone) and a detection head (Head). The present invention uses the FasterNet Block module to replace the Bottleneck module in C2f to obtain a C2f-F (C2f-Faster) structure, and adds a ShuffleAttention (SA) attention mechanism in the FasterNet Block module to design a new C2f-F-SA (C2f-Faster-ShuffleAttention) structure, and replaces the C2f structure in the traditional YOLOv8 visual recognition model with this structure; the Conv module of the Head part in the YOLOv8 visual recognition model is also replaced with an Adown module. It provides strong technical support for the rapid and accurate identification of sea cucumbers.

[0122] like Figure 7 As shown in the figure, the C2f-F-SA module consists of a convolutional layer (Conv), a split layer (Split), multiple FasterBlock structures, a connection layer (Concat), and a convolutional layer (Conv). The convolutional layer extracts features from the input data, which are then divided into multiple branches by the split layer. Each branch undergoes further feature extraction and information fusion through a FasterBlock structure composed of multiple sub-FasterBlock structures. The branches are then spliced ​​together through the connection layer, and finally the convolutional layer performs the final feature fusion and output.

[0123] The traditional C2f architecture consists of a series of bottleneck layers, which are responsible for extracting and enhancing local features. In this technical solution, the bottleneck layer structure in the YOLOv8 model is replaced with the FasterBlock structure to improve the network's feature extraction efficiency. This replacement strategy aims to reduce the computational complexity in the backbone network, thereby more effectively utilizing device computing resources and further improving the backbone network's ability to extract spatial features. Through this replacement, the FasterBlock structure replaces the bottleneck layer structure in C2f, providing strong technical support for the rapid and accurate identification of sea cucumbers.

[0124] The specific steps for replacing the bottleneck layer structure are:

[0125] Step 1: Identification and positioning

[0126] First, we conduct an in-depth analysis of the C2f module of the YOLOv8 model to accurately locate the existing bottleneck layer and prepare for the subsequent replacement process.

[0127] Step 2: Replacement and Integration

[0128] (1) Removal operations, gradually removing the existing bottleneck layers in the C2f module to free up space for the integration of the new FasterBlock;

[0129] (2) Insertion operation: insert the FasterBlock structure into the C2f module according to the position order of the original bottleneck layer to ensure the coherence of the network structure;

[0130] (3) Dimension matching: To ensure the compatibility of FasterBlock with the original network architecture, the input and output dimensions of FasterBlock are adjusted. This process includes adding or removing 1x1 convolutional layers (linear layers) when necessary to achieve dimensional consistency;

[0131] S3: Model reconstruction

[0132] After completing the integration of the FasterBlock structure, recompile the entire model to ensure that all network layers are connected correctly and that information flows smoothly in the model.

[0133] Step 4: Training and Optimization

[0134] The updated model structure was used for training, during which hyperparameters were carefully adjusted to optimize the model's performance. This series of training and optimization steps ensured that the improved YOLOv8 model provided by the present invention demonstrated excellent performance in the sea cucumber identification task.

[0135] When identifying sea cucumbers, some sea cucumbers account for a small proportion, resulting in an excessively high proportion of the image background. In order to more accurately and quickly identify small and medium-sized individual sea cucumbers, such as Figure 8 As shown in the figure, the SA attention mechanism (ShuffleAttention) is added to the C2f-F structure. When identifying individual sea cucumbers, the SA attention mechanism can effectively extract the key features of sea cucumbers in the image. When distinguishing between normal and diseased sea cucumbers, it enhances the model's ability to perceive the input sea cucumber spatial features. At the same time, it can reduce the amount of computation and increase the detection speed, thereby improving the model's overall performance in the sea cucumber image feature extraction task.

[0136] like Figure 9As shown in the figure, the SA attention mechanism includes input, split layer, upper branch, lower branch, connection layer and output.

[0137] Input sea cucumber characteristics Figure X ∈R (C，H，W) , the sea cucumber feature map is divided into G groups according to the channel dimension, expressed as X = [X1, ..., X G ], X k ∈R (C / G×H×W) , where C is the number of channels of the sea cucumber feature map, H is the spatial height of the sea cucumber feature map, W is the spatial width of the sea cucumber feature map, R represents the set of all images in the dataset, X k is the input feature.

[0138] The segmentation layer takes as input the feature X k Divided into upper branch X along the channel dimension K1 With the lower branch X K2 , X K1 , X K2 ∈R (C / 2G×H×W) , the upper branch uses channel attention, and generates channel information S by scaling and shifting the transformation parameters w1, b1. It first performs global average pooling (GAP) to embed global information and generate channel information S∈R (C / 2G×1×1) , the channel information is obtained by k1 Calculate by shrinking on the spatial dimension H×W,

[0139]

[0140] Use the fully connected layer to transform the channel information S and apply the σ activation function (sigmoid) to generate the weight matrix and multiply it by X k1 , and get the weighted features Figure X ' k1 :

[0141] X′ k1 =σ(F C (S))·X k1 =σ(w1S+b1)·X k1 ,

[0142] Where w1∈R (C / 2G×1×1) , b1∈R (C / 2G×1×1) ,

[0143] The mathematical expression of the σ activation function is:

[0144] Where z is the input of the function, which can be a scalar, vector, matrix or higher-dimensional tensor;

[0145] The lower branch uses spatial attention to capture the spatial dependencies between features, scales and shifts the amount of spatial information by transforming parameters w2 and b2, and uses the group norm (GN) to generate statistics of spatial attention:

[0146] X′ k2 =σ(w2·GN(X k2 )+b2)·X k2 ,

[0147] Where w2∈R (C / 2G×1×1) , b2∈R (C / 2G×1×1) .

[0148] The connection layer processes X′ obtained by the upper and lower branches K1 and X′ K2 Combined, the output feature map has the same size as the input,

[0149] X′ k =[X′ k1 , X′ k2 ]R (C / G×H×W) .

[0150] To address the challenges faced by the backbone network when performing sea cucumber feature extraction tasks, an innovative ADown module is used to optimize the downsampling process. Traditional standard convolutional downsampling reduces the spatial size of the feature map by increasing the stride. However, this method often results in a significant reduction in feature map size, causing sea cucumber regions that occupy a relatively small portion of the original image to be mistakenly merged into the larger background area, resulting in the loss of critical information. This information loss causes the features of small sea cucumbers to become blurred or diluted during the downsampling process, seriously affecting the model's ability to recognize these objects in subsequent layers.

[0151] like Figure 9 As shown in Figure 2, the ADown downsampling method combines the advantages of average pooling (AvgPool2d) and maximum pooling (MaxPool2d). Average pooling is responsible for preserving global information, while maximum pooling emphasizes salient features. By performing convolution operations on the outputs of these two pooling methods, the ADown module can effectively preserve multi-level information, thereby preventing potential information loss during the downsampling process and improving the model's sensitivity to detailed features. At the same time, a weight factor learning mechanism is introduced to achieve more refined fusion of downsampling results.

[0152] By replacing the traditional downsampling Conv module with the downsampling Adown module, the model's object perception capability is significantly improved, thereby enhancing the accuracy and reliability of sea cucumber detection. This innovative design provides a new solution for sea cucumber feature extraction and detection, with broad application prospects and practical value.

[0153] The ADown module includes an input terminal (Input), an average pooling layer (AvgPool2d), a segmentation layer (Split), a first path (cv1), a second path (cv2), a connection layer (Concat) and an output terminal.

[0154] Input characteristics Figure X ∈R (C，H，W) , the average pooling layer takes the input features Figure X The average pooling feature is obtained by performing an average pooling operation through an average pooling layer with a pooling kernel k of size 2, a step size s of 1, and no padding (p = 0). Figure X avg ,

[0155] X avg =Avgpool(X, k=2, s=1, p=0).

[0156] The segmentation layer averages the pooled features Figure X avg Split into two sub-features along the channel dimension Figure X 1 and X2, sub-features Figure X 1 and X2 are downsampled simultaneously through the first path and the second path respectively to obtain feature maps Y1 and Y2.

[0157] The first path is: sub-feature Figure X 1 directly downsamples through a traditional convolutional layer to obtain the feature map Y1,

[0158] Y1=Conv1(X1, k=3, s=2, p=1);

[0159] The second path is: sub-feature Figure X 2 First, a maximum pooling layer with a pooling kernel size k of 3, a step size s of 2, and a padding p of 1 is used to extract the maximum value X2 in the local window and highlight the significant features. Subsequently, the feature map is further extracted and refined through a convolution layer with a convolution kernel size k of 1, a step size s of 1, and a padding p of 0 to obtain the feature map Y2.

[0160] X 2，max =MaxPool(X2, k=3, s=2, p=1),

[0161] Y2=Conv2(X 2，max , k=1, s=1, p=0),

[0162] Two learnable weight factors α and β are introduced in the connection layer. Through the training process optimization, the down-sampling results are weighted fused. The weight factors α and β control the contribution of Y1 and Y2 in the fusion process respectively. The dimensions of the weight factors α and β are both 1×1×1. The feature map Y after weighted fusion isfusion for:

[0163] Y fusion =α·Y1+β·Y2,

[0164] where · represents element-wise multiplication.

[0165] Normalize the weight factors α and β to ensure that their sum is 1:

[0166]

[0167] The feature map after weighted fusion is recalculated using the normalized weight factor:

[0168] Y fusion =α norm Y1+β norm ·Y2.

[0169] From the output end, the weighted fusion feature map Y fusion As output, its dimension is Where C' is the number of output channels.

[0170] The construction method of the visual recognition model is as follows: grouping sea cucumbers according to their region and species, and making the size of each group of sea cucumbers in a similar range, and the number of sea cucumbers in each group is not less than 100; obtaining photos of each sea cucumber under different shooting angles, different lighting and different backgrounds under natural conditions, the number of photos is not less than 1,000, obtaining the growth trait data of each sea cucumber, the growth trait data including body length, body width, wet weight and dry weight, matching the sea cucumber images with their growth trait data, and constructing a data set; dividing the captured pictures into training set and test set according to proportion, manually using bounding boxes to annotate the sea cucumbers in the training set, using the YOLOv8-seacucumber visual recognition model to train the model for all labeled sea cucumber image information, and using the test set to test the model until the detection accuracy and detection speed of the model meet the requirements, and manually reviewing the training results; performing random tests on the trained model, and if the average accuracy (Average Accuracy) of the test results is Precision, AP) is greater than the threshold, then the visual recognition model is trained successfully, and the confidence average precision AP (IoU = 0.75) threshold is preferably 0.8. This visual recognition model can be used to identify sea cucumbers of the same species in the area. If the area where the sea cucumber is located or the species of the sea cucumber is changed, it is necessary to resample and repeat the above steps to rebuild the model. The YOLOv8-seacucumber visual recognition model identifies the sea cucumbers that may exist in the image and calculates its confidence. The targets with confidence greater than the threshold are identified as sea cucumbers and marked in the figure. The confidence threshold is preferably 0.9. The longest and widest points of the sea cucumber edge are used as critical points to draw the recognition frame, and the recognition frame size is recorded. The size includes the number of pixels of the length and width of the recognition frame. Each recognition frame is sorted in turn, and the final sort value of the recognition frame is recorded as the total number of sea cucumber individuals identified in this round.

[0171] Furthermore, the image recognition module also includes a sea cucumber disease recognition model. This model monitors the health status of sea cucumbers based on sea cucumber images and simultaneously monitors their health during the tank emptying process, enabling timely detection of disease risks. The sea cucumber disease library model is trained using photos of diseased sea cucumbers. The images are proportionally divided into training and test sets. The diseased conditions of the sea cucumbers in the training set are manually annotated using bounding boxes. The model is then trained on all annotated sea cucumber images using the YOLOv8-seacucumber visual recognition model and tested using the test set until the model's detection accuracy and speed meet the requirements. The training results are then manually reviewed. The sea cucumber disease recognition model is also used to identify the health status of sea cucumbers. Diseased or abnormal sea cucumbers are additionally marked and labeled with warning labels for easy identification. Based on the identified health status information, staff can take appropriate measures to ensure the healthy growth of sea cucumbers.

[0172] The data processing module obtains the size of the recognition frame and calculates the actual size of the recognition frame based on the actual length and width of a single pixel of the calibrated camera. The actual length and width of the recognition frame are obtained by multiplying the number of pixels in the length and width of the recognition frame with the actual length and width of a single pixel.

[0173] The prediction model is used to monitor and predict the growth status of sea cucumbers based on the actual size of the identification frame. The prediction model is based on the labeled sea cucumber images and sea cucumber growth trait data. The actual size of the identification frame of the same species of sea cucumbers in the same area is fitted with its growth trait data, and the prediction model is obtained by model training using multiple regression models and machine learning. The method for constructing the prediction model is to use the image of the sea cucumber and its growth trait data as a data set, and use multiple regression models to train the data set based on the actual size of the identification frame in the sea cucumber image. The training models include Multiple Linear Regression, Polynomial Regression, Generalized Linear Models (GLM), Decision Tree Regression, Ridge Regression, Random Forests and Support Vector Regression (SVR). The trained model is tested and the determination coefficient R is used to calculate the predicted value. 2 Evaluate model performance, coefficient of determination R 2 It indicates the degree of explanation of the target variable by the variables explained by the regression model. The larger the determination coefficient is, the smaller the errors are, which means the corresponding model is better. If the evaluation result R 2 If it is greater than 0.9, the model is considered qualified.

[0174] The actual size of each sea cucumber identification frame is loaded into the prediction model as a prediction parameter, and the growth status of the sea cucumber is fitted and predicted. The prediction model selects multiple fitting models with the highest confidence and a confidence level greater than a threshold. The growth trait data of the multiple fitting models that meet the requirements are averaged to obtain the growth monitoring data of the length and width of a single sea cucumber and the prediction data of the current wet weight and dry weight. The predicted value is stored in the database, and the growth trait data of each sea cucumber is recorded. Specifically, the actual size of a target sea cucumber identification frame is obtained and loaded into the prediction model. The prediction model selects three fitting models with the highest confidence level greater than 0.9 based on the actual size of the sea cucumber identification frame, which are the fitting models closest to the sea cucumber. The average of these three fitting models is calculated to obtain the growth trait data of the target sea cucumber.

[0175] Sea cucumber often can be twisted under real growth state, be difficult to measure information such as the real length, width of sea cucumber under the living state of sea cucumber, also can't measure the dry weight of sea cucumber, use as the length and width of sea cucumber calculated based on the distortion radian of sea cucumber also easily because sea cucumber itself has shrinkage and causes its result of calculation and the true length of sea cucumber to have certain error.So the present invention takes into account the distortion condition of sea cucumber, even if sea cucumber is under distortion state, the aspect ratio of its identification frame changes, the high confidence regression model that can be set up according to the growth properties data of identification frame actual size fitting sea cucumber carries out growth monitoring and prediction to the growth properties data of sea cucumber, obtains growth monitoring data and the prediction data of wet weight and dry weight of sea cucumber size.In addition compared with calculating sea cucumber true length, using identification frame actual size to monitor and predict the growth properties data of sea cucumber can significantly reduce computing intensity, improves detection efficiency.

[0176] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for integrating intelligent sea cucumber ponding, growth monitoring and prediction, characterized in that: include: The sea cucumbers are caught and placed on a conveyor belt, and the calibrated image acquisition module captures images containing the sea cucumbers during the process of the sea cucumbers being poured into the pool; The pre-built YOLOv8-seacucumber visual recognition model recognizes sea cucumber images, identifies objects with a confidence score greater than a threshold as sea cucumbers, draws an identification frame, and draws an identification frame with the longest and widest points of the sea cucumber edge as critical points, and records the size of the identification frame; The YOLOv8-seacucumber visual recognition model includes an input end, a backbone network, and a detection head. Based on the original YOLOv8 visual recognition model, the C2f-F-SA module is used to replace the C2f module in the backbone network and the detection head. The C2f-F-SA module includes a convolutional layer, a segmentation layer, multiple FasterBlock structures, a connection layer, and a convolutional layer arranged in sequence. The bottleneck layer in the original C2f module is replaced with a FasterBlock structure, and the ShuffleAttention mechanism is added to the FasterNet Block module. Some convolutional layers in the detection head are replaced with Adown modules. Obtain the actual length and width of a single pixel in the image acquisition module, calculate the actual size of the identification frame, and sort each identification frame in turn to count the total number of sea cucumber individuals; The actual size of the identification frame is loaded as a prediction parameter into a pre-built prediction model. The prediction model monitors and predicts the sea cucumber growth trait data based on the relationship between the actual size of the identification frame and the growth traits of the sea cucumber. Multiple fitting models with the highest confidence and a confidence greater than a threshold are selected, and the average value of the fitting model is calculated to obtain the monitoring data of the sea cucumber size and the prediction data of the wet weight and dry weight.

2. The integrated method for intelligent inverting pond, growth monitoring and prediction of sea cucumbers according to claim 1, characterized in that: The Shuffle Attention mechanism includes an input end, a segmentation layer, an upper branch, a lower branch, a connection layer, and an output end; Input sea cucumber characteristic map , the sea cucumber feature maps are divided into G groups according to the channel dimension, which can be expressed as: ， Where C is the number of channels of the sea cucumber feature map, H is the spatial height of the sea cucumber feature map, W is the spatial width of the sea cucumber feature map, R represents the set of all images in the dataset, X k is the input feature; The segmentation layer takes as input the feature X k Divided into upper branch X along the channel dimension K1 With the lower branch X K2 , , the upper branch uses channel attention, and generates channel information S by scaling and shifting the transformation parameters w1, b1. First, global average pooling is performed to embed global information and generate channel information , the channel information is obtained by k1 Calculated by shrinking on the spatial dimension H×W, ， Use the fully connected layer to transform the channel information S and apply the σ activation function to generate the weight matrix and multiply it by X k1 , get the weighted feature map X' k1 : ， Where, , ; The lower branch uses spatial attention to capture the spatial dependencies between features, scales and shifts the amount of spatial information by transforming parameters w2,b2, and uses the group norm to generate statistics of spatial attention: ， Where, , ; The connection layer processes the X' obtained by the upper and lower branches K1 and X' K2 Combined, the output feature map has the same size as the input, 。 3. The integrated method for intelligent inverting pond, growth monitoring and prediction of sea cucumbers according to claim 1, characterized in that: The ADown module includes an input terminal, an average pooling layer, a segmentation layer, a first path, a second path, a connection layer, and an output terminal. Input feature map , where C is the number of channels of the sea cucumber feature map, H is the spatial height of the sea cucumber feature map, W is the spatial width of the sea cucumber feature map, R represents the set of all images in the dataset, and the average pooling layer performs an average pooling operation on the input feature map X through a pooling kernel k with a size of 2, a step size s of 1, and no padding to obtain the average pooled feature map X avg , ， The segmentation layer averages the pooled feature map X avg The sub-feature maps X1 and X2 are divided into two sub-feature maps X1 and X2 along the channel dimension. The sub-feature maps X1 and X2 are downsampled simultaneously through the first path and the second path respectively to obtain feature maps Y1 and Y2. The first path is that the sub-feature map X1 is directly downsampled through a traditional convolutional layer to obtain the feature map Y1. ； The second path is that the sub-feature map X2 first passes through a maximum pooling layer with a pooling kernel size k of 3, a step size s of 2, and a padding p of 1 to extract the maximum value X2 in the local window and highlight the significant features. Subsequently, the feature map passes through a convolution layer with a convolution kernel size k of 1, a step size s of 1, and a padding p of 0 to further extract and refine the features to obtain the feature map Y2. ， ， Two learnable weight factors α and β are introduced in the connection layer. Through the training process optimization, the downsampling results are weighted fused. The weight factors α and β control the contribution of Y1 and Y2 in the fusion process respectively. The dimensions of the weight factors α and β are both 1×1×1. The feature map Y after weighted fusion is fusion for: ， Where ∙ represents element-wise multiplication, Normalize the weight factors α and β to ensure that their sum is 1. ， ； The weighted fusion feature map is recalculated using the normalized weight factor: ， From the output end, the weighted fusion feature map Y fusion As output, its dimension is , where C' is the number of output channels.

4. The sea cucumber intelligent pond inversion and growth monitoring and prediction integrated method according to claim 1 is characterized in that, The YOLOv8-seacucumber visual recognition model construction method comprises the following steps: grouping sea cucumbers according to their region and species, and conducting sampling training, obtaining photos of each sea cucumber under natural conditions at different shooting angles, different lighting, and different backgrounds, and dividing the photos into a training set and a test set in proportion, manually annotating the sea cucumbers in the training set using bounding boxes, training the model using a convolutional neural network on all labeled sea cucumber image information, testing the model using a test set, and manually reviewing the training results to obtain a visual recognition model.

5. The integrated method for intelligent inverting pond, growth monitoring and prediction of sea cucumbers according to claim 1, characterized in that: The prediction model construction method is as follows: based on the marked sea cucumber images and sea cucumber growth trait data, the actual size of the identification frame of the same species of sea cucumber in the same area is fitted with its growth trait data, and a prediction model is obtained by model training using multiple regression models and machine learning.

6. A device for integrating intelligent inverting and growth monitoring and prediction of sea cucumbers, using the method for integrating intelligent inverting and growth monitoring and prediction of sea cucumbers according to any one of claims 1 to 5, characterized in that: include: Aquaculture ponds, comprising original ponds currently culturing sea cucumbers and new ponds for replacement; A conveyor belt is provided between the original pool and the new pool for transporting sea cucumbers; An image acquisition module, which is arranged on the conveyor belt and is used to acquire images containing sea cucumbers during the process of pouring sea cucumbers into the pool; An image recognition module is configured to obtain an image containing a sea cucumber, identify the sea cucumber using a pre-built visual recognition model, draw an identification frame using the longest and widest points of the sea cucumber edge as critical points, record the size of the identification frame, sequentially sort each identification frame within the identification area, and count the total number of sea cucumber individuals; A data processing module, wherein the data processing module obtains the size of the recognition frame and calculates the actual size of the recognition frame; A prediction module obtains the actual size of the identification frame, uses a pre-built prediction model, takes the actual size of the sea cucumber identification frame as a prediction parameter, selects multiple fitting models with the highest confidence and a confidence greater than a threshold, and calculates the average value of the fitting models to obtain monitoring data of sea cucumber size and predicted data of wet weight and dry weight.

7. The integrated device for intelligent inverting pond, growth monitoring and prediction of sea cucumbers according to claim 6, characterized in that: The conveyor belt is a U-shaped conveyor belt, which is arranged on the outside of the breeding pond. The original pond and the new pond are spaced apart. A fishing mechanism is provided in the breeding pond. The fishing mechanism includes a transverse track, a movable frame and a fishing shovel. The transverse track is provided on the breeding pond. A movable frame that can move along the transverse track is provided on the transverse track. A fishing shovel that can be connected to the movable frame by a lifting rope is provided in the breeding pond.

8. The integrated device for intelligent sea cucumber pond inversion, growth monitoring and prediction according to claim 7, characterized in that: The movable frame includes a longitudinal track arranged on the transverse track and a frame body installed on the longitudinal track, a rack is provided on one side of the longitudinal track, a driving motor fixedly installed in the frame body is provided above the rack, a gear is provided on the output shaft of the driving motor, the gear is engaged in the rack, a first rope winding shaft and a second rope winding shaft are provided in the frame body, two rope winding drums are provided in each of the first rope winding shaft and the second rope winding shaft, a lifting rope that can be connected to the fishing shovel is wound in the rope drum, and a first lifting motor and a second lifting motor are provided at the ends of the first rope winding shaft and the second rope winding shaft respectively.

9. The integrated device for intelligent inverting pond, growth monitoring and prediction of sea cucumbers according to claim 7, characterized in that: The fishing shovel includes a bottom plate, a rear plate and side plates. An opening is provided on the front side of the fishing shovel. The width of the fishing shovel matches the width of the breeding pond. A plurality of drainage holes are provided in the bottom plate of the fishing shovel, a plurality of water filtering holes are provided in the rear plate of the fishing shovel, and a hanging ring is provided on the top of the fishing shovel.

Images