An electronic feeding platform device for river crab farming based on machine vision and its working method

Abstract

The present invention discloses a machine vision-based electronic feeding table device and working method for river crab farming. The device consists of an electronic feeding table and a data center. The electronic feeding table includes an underwater target detection module, a water quality detection module, an observation platform and a bracket, and a wireless communication module, which are used to monitor the remaining amount of bait, the feeding duration of river crabs, the average mass of river crabs and the changes in water quality parameters in real time in the crab pond. The data center includes a Web server and a Web client, which are used to remotely receive, process, save and display the data monitored by the electronic feeding table. Through the water quality detection module, three key factors in the process of river crab farming, namely the dissolved oxygen content, water temperature and pH value in the crab pond, are obtained, and the data center quantitatively adjusts the bait feeding amount. The present invention has the advantages of comprehensive data management, high detection accuracy and strong real-time performance, which can not only ensure that river crabs eat enough, but also improve the bait utilization rate, achieving the purpose of precision farming of river crabs.

Description

Technical Field

[0001] The present invention belongs to the fields of aquaculture and machine vision, and relates to an electronic feeding table device for river crab farming based on machine vision and a working method thereof. Background Art

[0002] In the actual process of river crab farming, the feeding of river crabs is affected by the quality of the river crabs themselves, the distribution of river crabs in the crab pond, and the water quality environment of the crab pond. The determination of the traditional bait feeding amount, the detection of the bait remaining rate, and the detection of the quality of river crabs mainly rely on the breeding experience of crab farmers for manual operation. This method not only has a large labor intensity and a low degree of automation, but also because crab farmers cannot directly observe the remaining situation of bait at the bottom of the crab pond, there is no real-time detection and lack of consideration of environmental factors such as water temperature, dissolved oxygen, and pH value of the water body. The feeding error is large, which is likely to cause river crabs to scramble for food, be injured and die, the bait utilization rate is low, and the breeding efficiency is poor. Therefore, it is necessary to use electronic devices to automatically detect the actual growth and feeding situation of river crabs in real time, and then timely and quantitatively adjust the bait feeding amount to improve the breeding efficiency of river crabs.

[0003] Regarding the research on the application system of electronic feeding tables, foreign countries mainly focus on large-scale deep-sea cage aquaculture. AKVA Company in Norway has developed an intelligent aquaculture system, which is equipped with underwater cameras and various sensing devices and can monitor the aquaculture environment and fish populations. However, this system is suitable for deep-sea cage aquaculture, not suitable for inland river crab farming, and the equipment is expensive; the domestic design of electronic feeding tables is still in its infancy. Zhejiang University has developed an intelligent management and control system for efficient ecological circulating water aquaculture, which realizes the segmentation of bait targets through local adaptive threshold method and Canny edge detection method. However, this system has the disadvantages of slow detection speed of bait targets and large counting errors. Summary of the Invention

[0004] In order to solve the above technical problems, the present invention provides an electronic feeding table device for river crab farming based on machine vision and a working method thereof.

[0005] The present invention is realized through the following technical solutions: An electronic feeding table device for river crab farming based on machine vision, which is composed of an electronic feeding table and a data center. The electronic feeding table includes an underwater target detection module, a water quality detection module, an observation platform and a bracket, and a wireless communication module, which is used to monitor the remaining quantity of bait, the feeding duration of river crabs, the average quality of river crabs and the change of water quality parameters in the crab pond in real time. The data center includes a Web server and a Web client, which are used to remotely receive, process, save and display the data monitored by the electronic feeding table.

[0006] The bottom of the observation platform is a white bottom plate; the underwater target detection module consists of a mobile phone and a waterproof mobile phone case, which is installed at the center of the feeding table, collects underwater video data in real time, detects the remaining amount of bait and the length and width of the crab shell, and uploads the data to the data center Web server through the wireless communication module; the water quality detection module uses a microcontroller to receive signals and make decisions. Through the dissolved oxygen sensor and pH sensor, it monitors the dissolved oxygen content, temperature and pH value of the water body respectively, and uploads the water quality parameters to the data center Web server through the wireless communication module. The Web client reads the data from the server and provides a graphical page to display the data information.

[0007] A working method of an electronic feeding table device for river crab breeding based on machine vision according to the present invention collects data through the electronic feeding table device, including the following steps:

[0008] Step S1: Place the electronic feeding table and randomly select river crab and bait samples in the crab pond;

[0009] Step S2: Use the water quality detection module of the electronic feeding table to obtain the dissolved oxygen content, temperature and pH value parameters of the water body during the river crab breeding process in real time, and upload the water quality parameters to the data center through the wireless communication module;

[0010] Step S3: Use the underwater target detection module of the electronic feeding table to detect the remaining amount of bait and the length and width of the crab shell in real time through an improved target recognition algorithm, and upload the detection results to the data center through the wireless communication module;

[0011] Step S4: The data center Web server receives, processes and saves the data such as the remaining bait rate, river crab feeding duration, river crab quality, and water quality parameters monitored by the electronic feeding table device. The Web client reads the data from the server and provides a graphical page to display the data information.

[0012] Further, step S3 specifically includes the following steps:

[0013] Step S31: Construct a dataset of river crab and bait pictures. After framing and annotating the river crab and bait images, divide the dataset into a training set and a test set;

[0014] Step S32: Feature extraction, build an improved YOLOv5 lightweight network feature extraction model; use GhostNet convolution to replace the ordinary convolution of the backbone network, and first generate the original feature map through 1×1 ordinary convolution operation Then perform depthwise separable convolution operations on the original feature map one by one to generate redundant feature maps Next, the original feature map and the redundant feature map are concatenated to generate a feature map, and the CoordAttention attention mechanism is added between the BackBone and Head structures of the YOLOv5 network to capture cross-channel information, direction-aware information, and location-aware information, enabling the model to more accurately locate and identify the target of interest;

[0015] Step S33: Input the feature map generated in step S32 into the Neck part, obtain three feature maps of different sizes through the bidirectional fusion backbone network (FPN+PAN) structure, generate candidate boxes on the three feature maps of different scales, and then obtain the detection boxes of the crab and bait targets based on the loss function and backpropagation;

[0016] Step S34: Regression target box loss function module, use SIoU to calculate the loss function, and calculate by combining the overlap degree, center distance, aspect ratio, and angle between the real box and the predicted box. The SIoU loss function Loss SIoU The calculation method is:

[0017]

[0018] Among them, B is the predicted box, B gt is the real box; is the loss value combined with distance and angle; is the distance loss, is the abscissa of the center point of the real box, is the abscissa of the center point of the predicted box, is the ordinate of the center point of the real box, is the ordinate of the center point of the predicted box, c w is the horizontal distance between the center points of the real box and the predicted box, c h is the vertical distance between the center points of the real box and the predicted box; γ = 2 - sin(2α), is the angle loss, and α is the angle between the real box and the predicted box; is the aspect ratio loss, θ is the aspect ratio attention parameter, w is the length of the predicted box, w gt is the length of the real box, h is the width of the predicted box, h gt is the width of the real box;

[0019] Step S35: Use the object detection evaluation index to evaluate the improved model, select the parameter model with the highest detection accuracy, deploy it on the mobile phone, the mobile phone detects the target in the water in real time, and uploads the remaining number of baits and the pixel data of the length and width of the crab shell to the data center.

[0020] Furthermore, the specific steps of the step S4 include the following steps:

[0021] Step S41: Calculate the bait remaining rate of the Chinese mitten crabs by the remaining amount of bait, and calculate the feeding duration of the Chinese mitten crabs;

[0022] Step S42: Place a number of reference objects with the same size on the food platform, correct the detected length and width pixels of the crab shell, and convert the length and width pixels into the mass m of the Chinese mitten crab. The conversion method is as follows:

[0023]

[0024] where a w1 , a w2 is the attention parameter of the crab shell width; b w1 , b w2 is the attention parameter of the crab shell length; w rp is the actual width of the corrected crab shell, w op is the pixel width of the crab shell, p(x) is the correction ratio function in the X-axis direction; h rp is the actual length of the corrected crab shell, h op is the pixel length of the crab shell, p(y) is the correction ratio function in the Y-axis direction;

[0025] Step S43: Quantitatively adjust the bait distribution coefficient of each area in the crab pond by combining the bait remaining rate, feeding duration, mass of the Chinese mitten crab and water quality environment parameters. The adjustment method is as follows:

[0026]

[0027] where k 0 (x,y) is the initial bait distribution coefficient of the sub-area in the crab pond, M h (x,y) is the mass distribution density of the Chinese mitten crab, S(x,y) is the area of the sub-area, E d (x,y) is the water quality environment parameter coefficient of the sub-area; c f is the bait remaining rate; c f0 is the bait remaining rate threshold; h f is the feeding duration of the Chinese mitten crab, h f0 is the feeding duration threshold of the Chinese mitten crab. The electronic food platform detects the bait remaining rate, the average mass of the Chinese mitten crab and the water quality parameters in real time. When the bait remaining rate is less than the bait remaining rate threshold, the feeding duration of the Chinese mitten crab is recorded; if the water quality parameters exceed the warning value suitable for the growth of the Chinese mitten crab, an alarm is sent to remind the user through the data center.

[0028] The beneficial effects of the present invention are as follows: The present invention improves the YOLOv5 model based on lightweight, reduces the computational complexity of the model through GhostNet convolution, and speeds up the running speed of the model on mobile devices. By using mobile devices to detect the bait and crab targets on the electronic feeding platform in real time, the remaining rate of bait, the feeding duration of crabs, the size and quality of crabs are obtained, and the pond water quality environment parameters are obtained through sensors, so as to realize quantitative adjustment of the bait distribution in each area of the crab pond, save the breeding cost of crabs while ensuring sufficient feeding of crabs, and prevent water pollution. It has the advantages of comprehensive data management, high detection accuracy and strong real-time performance, and can achieve the purpose of precise crab breeding. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 is a schematic diagram of the composition of the electronic feeding platform system of the present invention;

[0030] Figure 2 is the structure diagram of the electronic feeding platform;

[0031] Figure 3 is the general flow chart of an underwater object detection method based on machine vision;

[0032] Figure 4 is the flow chart of the crab quality estimation method;

[0033] Figure 5 is the flow chart of the bait feeding amount adjustment method. DETAILED DESCRIPTION OF THE INVENTION

[0034] The following further describes the specific implementation of the present invention with reference to the accompanying drawings.

[0035] As Figure 1 shown, the electronic feeding platform system consists of several electronic feeding platform monitoring points, a Web server and a Web client. In the crab pond, the water quality parameters, the remaining quantity of bait and the pixel values of the length and width of the crab shell are monitored by the electronic feeding platform monitoring points and uploaded to the cloud server through the wireless network; the cloud server processes, analyzes and saves the monitoring data to the database, and adjusts the bait feeding amount; the Web client displays the monitoring data in real time, and sends an alarm to the user when the monitoring data is abnormal.

[0036] As Figure 2As shown in the figure, the electronic food platform device consists of an underwater target detection module, a water quality detection module, an observation platform and its support, a wireless communication module, etc. Among them, the observation platform is 1 m long and 1 m wide, and the platform surface is white; the support is made of 304 stainless steel and has corrosion resistance; the underwater target detection module consists of a mobile phone and a waterproof mobile phone case, which is installed at the center of the food platform, 1 m above the observation platform, with the shooting angle perpendicular downward, and it can collect underwater video data in real time, detect the remaining amount of bait and the length and width data of crab shells, and upload them to the Web server of the data center through the wireless communication module; the water quality detection module uses STM32F407ZGT6 as the microcontroller, and through the fluorescence dissolved oxygen sensor and the pH composite electrode sensor, it can measure the dissolved oxygen content, temperature and pH value of the water body respectively, and upload the water quality parameters to the Web server of the data center through the 4G module.

[0037] As Figure 3 shown in the figure, due to the limited memory and computing resources of the mobile phone side, on the premise of ensuring accuracy, the YOLOv5 target detection algorithm is improved. The ordinary convolution of the original network model is replaced by GhostNet convolution, and the original feature map is generated through the 1×1 ordinary convolution operation. Then, the depthwise separable convolution operation is performed on the original feature map one by one to generate redundant feature maps. Next, the original feature map and the redundant feature map are concatenated to generate a feature map, which greatly reduces the number of parameters and the amount of calculation. And the CoordAttention attention mechanism is added between the BackBone and Head structures of the YOLOv5 network to capture cross-channel information, direction-aware information and position-aware information, so that the model can more accurately locate and identify the target of interest and improve the detection accuracy of the model. And in order to improve the convergence speed of training, the SIoU loss function is used, which is calculated by combining the overlap degree, center distance, aspect ratio and angle between the real box and the predicted box. The calculation method of the SIoU loss function is as follows:

[0038]

[0039] Among them, B is the predicted box, and B gt is the real box; is the loss value combined with distance and angle; is the distance loss, is the abscissa of the center point of the real box, is the abscissa of the center point of the predicted box, is the ordinate of the center point of the real box, is the ordinate of the center point of the predicted box, c w is the horizontal distance between the center points of the real box and the predicted box, c his the longitudinal distance between the center points of the ground truth box and the predicted box; γ = 2 - sin(2α), which is the angular loss, and α is the angle between the ground truth box and the predicted box; is the aspect ratio loss, θ is the aspect ratio attention parameter, w is the length of the predicted box, w gt is the length of the ground truth box, h is the width of the predicted box, h gt is the width of the ground truth box.

[0040] Obtain bait and Chinese mitten crab sample images through the feeding platform, process the image samples to make a dataset, use the improved YOLOv5 object detection algorithm to train the picture data, and use the object detection evaluation metrics Precision (precision rate), Recall (recall rate), AP (average precision), and mAP (mean average precision) to evaluate the improved model. As shown in formula (2), TP is the number of correctly predicted positive samples, FP is the number of false alarms that predict negative samples as positive samples, and Precision represents the proportion of true positive samples among all predicted positive sample results; as shown in formula (3), FN is the number of false alarms that predict positive samples as negative samples, and Recall represents the proportion of accurately predicted positive samples among all true positive samples; as shown in formula (4), AP is the area under the PR curve, which is used to measure the quality of the model for each category. The approximate method is used to calculate AP, where N is the total number of samples, k is the index of each sample point, and ΔR(k) = R(k) - R(k - 1); as shown in formula (5), mAP is the average value of AP for all categories. Then select the parameter model with the highest detection accuracy, deploy the model on the mobile phone, and identify the bait and Chinese mitten crab targets.

[0041]

[0042]

[0043]

[0044]

[0045] The flow of the Chinese mitten crab quality estimation method is as Figure 4 shown. Since the length and width pixels of equal-sized objects in the image are not equal, the length and width pixels of the crab shell need to be corrected first. First, place several reference objects of the same size (such as a one-yuan coin with a diameter of 2.5 cm) on the feeding platform, and through the improved YOLOv5 algorithm, detect the center positions (x, y) of each reference object in the image and the length and width pixels (w op , h op) The ratio of the pixel length and width of an object to its actual length and width at different positions of different images is measured. A large amount of data is fitted to obtain the functional relationship of the ratio of the pixel length and width of the object to its actual length and width at different positions. The ratio function in the X-axis direction is shown in formula (6), where w op is the pixel width of the object, and w rp is the actual width of the object; the ratio function in the Y-axis direction is shown in formula (7), where h op is the pixel length of the object, and h rp is the actual length of the object; then, through the improved YOLOv5 algorithm, the central position (x, y) and the pixel length and width (w op , h op ) of the crab shell on the image are detected, and the actual width and actual length of the crab shell are corrected through formulas (8) and (9) respectively; then, the corrected actual length and width of the crab shell are substituted into the relationship between the mass of the Chinese mitten crab and the length and width of the crab shell, as shown in formula (10), to obtain the mass of a single Chinese mitten crab. Finally, the average mass of the Chinese mitten crabs is taken to achieve the purpose of adjusting the bait feeding amount according to the change in the mass of the Chinese mitten crabs.

[0046]

[0047]

[0048]

[0049]

[0050] m = 1.353×w rp 0.874 + 0.1441×h rp 3.551 (10)

[0051] The process flow of the bait feeding amount adjustment method is as shown in Figure 5 , and the feeding amount is adjusted by four factors: the feeding duration of the Chinese mitten crab, the bait remaining rate, the average mass of the Chinese mitten crabs, and the water quality parameters. When the bait remaining rate on the feeding table is less than the bait remaining rate threshold, it is recorded that the Chinese mitten crab has finished eating, and the feeding duration of the Chinese mitten crab is recorded, so that the bait feeding amount is adjusted according to the feeding duration and the bait remaining rate of the Chinese mitten crab; the detection of the average mass of the Chinese mitten crabs is also carried out in real time, so that the bait feeding amount is adjusted according to the change in the mass of the Chinese mitten crabs; the water quality parameters are collected in real time through the water quality detection module. If the water quality parameters exceed the range suitable for the normal growth of the Chinese mitten crabs, an alarm is sent to the user, and the bait feeding amount is adjusted according to the real-time water quality parameters, and the adjustment is shown in formula (11):

[0052]

[0053] where k0 (x, y) is the bait distribution coefficient of the initial crab pond area, M h (x, y) is the mass distribution density of Chinese mitten crabs, S(x, y) is the area of the sub-region, E d (x, y) is the water quality environment parameter coefficient of the sub-region; c f is the bait remaining rate; c f0 is the threshold of the bait remaining rate; h f feeding duration, h f0 feeding duration threshold.

[0054] In summary, an electronic food table device and working method for Chinese mitten crab breeding based on machine vision according to the present invention integrate sensor technology, machine vision algorithms and Internet of Things communication technology, make lightweight improvements to the YOLOv5 target recognition algorithm for realistic application scenarios, and introduce the CoordAttention attention mechanism to enhance the feature extraction ability of the network and the SIoU target box loss function to accelerate the convergence speed during network training. Then, the underwater video is obtained through the underwater target detection module, the number of baits on the food table and the pixel lengths and widths of the crab shells of Chinese mitten crabs are detected based on the improved YOLOv5 target recognition algorithm, and the detection results are uploaded to the data center through a wireless network. The data center processes, analyzes and stores the data to obtain the bait remaining rate, the feeding duration of Chinese mitten crabs and the average mass of Chinese mitten crabs. Through the water quality detection module, three key factors in the process of Chinese mitten crab breeding, namely the dissolved oxygen content, water temperature and pH value in the crab pond, are obtained, and the data center makes quantitative adjustments to the bait feeding amount. The present invention has the advantages of comprehensive data management, high detection accuracy and strong real-time performance, can not only ensure that Chinese mitten crabs eat enough, but also improve the bait utilization rate, reduce water pollution, improve the breeding efficiency, and achieve the purpose of precise breeding of Chinese mitten crabs.

Claims

1. Working method of an electronic feeding table device for Chinese mitten crab farming based on machine vision, characterized in that, the electronic feeding table device consists of an electronic feeding table and a data center. The electronic feeding table includes an underwater target detection module, a water quality detection module, an observation platform and a bracket, and a wireless communication module, which are used to monitor the remaining amount of bait, the feeding duration of Chinese mitten crabs, the average mass of Chinese mitten crabs and the change of water quality parameters in the crab pond in real time. The data center includes a Web server and a Web client, which are used to remotely receive, process, save and display the data monitored by the electronic feeding table; the bottom of the observation platform is a white bottom plate; the underwater target detection module is composed of a mobile phone and a waterproof mobile phone case, which is installed in the center of the feeding table, collects underwater video data in real time, detects the remaining amount of bait and the length and width of the crab shell, and uploads the detection results to the Web server of the data center through the wireless communication module; the water quality detection module uses a microcontroller to receive signals and make decisions. Through a dissolved oxygen sensor and a pH sensor, it monitors the dissolved oxygen content, temperature and pH value of the water body respectively, and uploads the water quality parameters to the Web server of the data center through the wireless communication module. The Web client reads the data from the server and provides a graphical page to display the data information; the method includes: data collection through the electronic feeding table device, including the following steps: Step S1: Place the electronic feeding table and randomly select Chinese mitten crab and bait samples in the crab pond; Step S2: Use the water quality detection module of the electronic feeding table to obtain the dissolved oxygen content, temperature and pH value parameters of the water body during the Chinese mitten crab farming process in real time, and upload the water quality parameters to the data center through the wireless communication module; Step S3: Use the underwater target detection module of the electronic feeding table to detect the remaining amount of bait and the length and width of the crab shell in real time through an improved target recognition algorithm, and upload the detection results to the data center through the wireless communication module; Step S4: The Web server of the data center receives, processes and saves the data of the remaining bait rate, the feeding duration of Chinese mitten crabs, the mass of Chinese mitten crabs and the water quality parameters monitored by the electronic feeding table device. The Web client reads the data from the server and provides a graphical page to display the data information; the specific steps of step S3 are as follows: Step S31: Construct a dataset of Chinese mitten crab and bait pictures. After framing and annotating the Chinese mitten crab and bait images, divide the dataset into a training set and a test set; Step S32: Feature extraction, building a feature extraction model based on the improved lightweight YOLOv5 network; replacing the ordinary convolution of the backbone network with GhostNet convolution, first generating the original feature map through 1×1 ordinary convolution operation Then, perform depthwise separable convolution operations on the original feature map one by one to generate redundant feature maps Next, splice the original feature map and the redundant feature map to generate a feature map, and add a CoordAttention attention mechanism between the BackBone and Head structures of the YOLOv5 network to capture cross-channel information, direction-aware information, and position-aware information, making the model more accurately locate and identify the target of interest; Step S33: Input the feature map generated in step S32 into the Neck part, obtain three feature maps of different sizes through a bidirectional fusion backbone network (FPN+PAN) structure, generate candidate boxes on the three different scale feature maps, and then obtain the detection boxes of Chinese mitten crab and bait targets based on the loss function and backpropagation; Step S34: Return to the target bounding box loss function module, calculate the loss function using SIoU, and calculate it in combination with the overlap degree, center distance, aspect ratio, and angle between the real bounding box and the predicted bounding box. The SIoU loss function Loss SIoU The calculation method is as follows: Among them, B is the predicted bounding box, and B gt is the ground truth bounding box; is the loss value combined with distance and angle; is the distance loss, is the abscissa of the center point of the ground truth bounding box, is the abscissa of the center point of the predicted bounding box, is the ordinate of the center point of the ground truth bounding box, is the ordinate of the center point of the predicted bounding box, c w is the horizontal distance between the center points of the ground truth and predicted bounding boxes, c h is the vertical distance between the center points of the ground truth and predicted bounding boxes; γ = 2 - sin(2α), which is the angle loss, and α is the angle between the ground truth and predicted bounding boxes; is the aspect ratio loss, θ is the aspect ratio attention parameter, w is the length of the predicted bounding box, w gt is the length of the ground truth bounding box, h is the width of the predicted bounding box, h gt is the width of the ground truth bounding box; Step S35: Use the target detection evaluation index to evaluate the improved model, select the parameter model with the highest detection accuracy, deploy it on the mobile phone, and the mobile phone detects the target in the water in real time, and uploads the pixel data of the remaining amount of bait and the length and width of the crab shell to the data center.

2. According to the working method of an electronic feeding table device for Chinese mitten crab farming based on machine vision described in claim 1, characterized in that, the specific steps of step S4 are as follows: Step S41: Calculate the remaining rate of the bait eaten by the Chinese mitten crabs based on the remaining amount of the bait, and calculate the feeding duration of the Chinese mitten crabs. Step S42: Place a number of reference objects of the same size on the food platform, correct the detected length and width pixels of the crab shell, and convert the length and width pixels into the mass m of the Chinese mitten crab. The conversion method is as follows: Among them, a w1 , a w2 is the attention parameter of the crab shell width; b w1 , b w2 is the attention parameter of the crab shell length; w rp is the actual width of the crab shell after correction, w op is the pixel of the crab shell width, and p(x) is the correction ratio function in the X-axis direction; h rp is the actual length of the crab shell after correction, h op is the pixel of the crab shell length, and p(y) is the correction ratio function in the Y-axis direction; Step S43: Quantitatively adjust the bait distribution coefficient of each area in the crab pond by combining the remaining rate of the bait, the feeding duration, the mass of the Chinese mitten crab, and the water quality environment parameters. The adjustment method is as follows: where k 0 (x, y) is the bait distribution coefficient of the initial crab pond area, M h (x, y) is the mass distribution density of Chinese mitten crabs, S(x, y) is the area of the sub-region, E d (x, y) is the water quality environment parameter coefficient of the sub-region; c f is the bait remaining rate; c f0 is the threshold of the bait remaining rate; h f is the feeding duration of Chinese mitten crabs, h f0 is the threshold of the feeding duration of Chinese mitten crabs. The electronic feeding platform detects the bait remaining rate, the average mass of Chinese mitten crabs and the water quality parameters in real time. When the bait remaining rate is less than the threshold of the bait remaining rate, the feeding duration of Chinese mitten crabs is recorded; if the water quality parameters exceed the warning value suitable for the growth of Chinese mitten crabs, an alarm is sent through the data center to remind the user.

Images