CN115843733A - Machine vision-based electronic feeding table device for river crab cultivation and working method - Google Patents
Machine vision-based electronic feeding table device for river crab cultivation and working method Download PDFInfo
- Publication number
- CN115843733A CN115843733A CN202211662847.7A CN202211662847A CN115843733A CN 115843733 A CN115843733 A CN 115843733A CN 202211662847 A CN202211662847 A CN 202211662847A CN 115843733 A CN115843733 A CN 115843733A
- Authority
- CN
- China
- Prior art keywords
- crab
- bait
- river
- data
- water quality
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 38
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 50
- 238000001514 detection method Methods 0.000 claims abstract description 44
- 241000238557 Decapoda Species 0.000 claims abstract description 34
- 235000013305 food Nutrition 0.000 claims abstract description 26
- 238000004891 communication Methods 0.000 claims abstract description 14
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims abstract description 11
- 229910052760 oxygen Inorganic materials 0.000 claims abstract description 11
- 239000001301 oxygen Substances 0.000 claims abstract description 11
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 230000008569 process Effects 0.000 claims abstract description 9
- 230000008859 change Effects 0.000 claims abstract description 5
- 230000006870 function Effects 0.000 claims description 17
- 238000010586 diagram Methods 0.000 claims description 13
- 230000008447 perception Effects 0.000 claims description 6
- 238000012549 training Methods 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000012937 correction Methods 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 230000007613 environmental effect Effects 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 230000002457 bidirectional effect Effects 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 238000009432 framing Methods 0.000 claims description 2
- 230000004927 fusion Effects 0.000 claims description 2
- 238000002372 labelling Methods 0.000 claims description 2
- 238000012360 testing method Methods 0.000 claims description 2
- 239000000203 mixture Substances 0.000 claims 1
- 230000008901 benefit Effects 0.000 abstract description 7
- 238000009395 breeding Methods 0.000 abstract description 7
- 230000001488 breeding effect Effects 0.000 abstract description 7
- 238000013523 data management Methods 0.000 abstract description 3
- 238000009360 aquaculture Methods 0.000 description 7
- 244000144974 aquaculture Species 0.000 description 7
- 230000037406 food intake Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 2
- 235000012631 food intake Nutrition 0.000 description 2
- 238000003911 water pollution Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 239000010963 304 stainless steel Substances 0.000 description 1
- 241000251468 Actinopterygii Species 0.000 description 1
- 229910000589 SAE 304 stainless steel Inorganic materials 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 238000012258 culturing Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 238000002795 fluorescence method Methods 0.000 description 1
- 235000012054 meals Nutrition 0.000 description 1
- 238000011160 research Methods 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a river crab breeding electronic food platform device based on machine vision and a working method thereof, wherein the device is composed of an electronic food platform and a data center, the electronic food platform comprises an underwater target detection module, a water quality detection module, an observation platform, a support and a wireless communication module and is used for monitoring the residual amount of baits, the eating time of river crabs, the average quality of river crabs and the change condition of water quality parameters in a crab pond in real time, and the data center comprises a Web service end and a Web client and is used for remotely receiving, processing, storing and displaying the data monitored by the electronic food platform. Through a water quality detection module, three key factors of the dissolved oxygen content, the water temperature and the pH value in the crab pond in the river crab culture process are obtained, and a data center quantitatively adjusts the bait feeding amount. The method has the advantages of comprehensive data management, high detection accuracy and strong real-time performance, can ensure that the river crabs eat enough food, can improve the utilization rate of bait, and achieves the aim of accurately breeding the river crabs.
Description
Technical Field
The invention belongs to the field of aquaculture and machine vision, and belongs to an electronic feeding table device for river crab culture based on machine vision and a working method.
Background
In the actual process of river crab culture, the feeding of river crabs is influenced by the self-quality of the river crabs, the distribution condition of the river crabs in a crab pond and the water quality environment of the crab pond. The traditional bait feeding amount determination, bait residual rate detection and river crab quality detection mainly depend on the manual operation of the breeding experience of crab farmers, the method has the advantages of high labor intensity and low automation degree, and the river crabs are easily eaten, damaged and killed due to the fact that the crab farmers cannot directly observe the residual condition of the baits at the bottoms of crab ponds and do not detect in real time or consider environmental factors such as water temperature, dissolved oxygen and pH value, the bait feeding error is large, the bait feeding is low in utilization rate and the breeding benefit is poor. Therefore, the actual growth and ingestion conditions of the river crabs need to be automatically detected in real time through electronic equipment, and then the bait feeding amount is timely and quantitatively adjusted, so that the culture benefit of the river crabs is improved.
The research on the electronic food platform application system mainly focuses on large-scale deep sea cage aquaculture abroad, and Norwegian AKVA company develops an intelligent aquaculture system which is provided with a water bottom camera and various sensing devices and can monitor the aquaculture environment and fish school, but the system is suitable for deep sea cage aquaculture, is not suitable for inland river crab aquaculture and is expensive in equipment; the design of an electronic food platform is still in a starting stage in China, a high-efficiency ecological circulating water aquaculture intelligent control system is developed by Zhejiang university, the bait target is segmented by a local adaptive threshold method and a Canny edge detection method, but the system has the defects of low detection speed and large counting error of the bait target.
Disclosure of Invention
In order to solve the technical problems, the invention provides an electronic feeding table device for river crab cultivation based on machine vision and a working method.
The invention is realized by the following technical scheme: an electronic food table device for river crab breeding based on machine vision is composed of an electronic food table and a data center, wherein the electronic food table comprises an underwater target detection module, a water quality detection module, an observation platform, a support and a wireless communication module and is used for monitoring the residual amount of bait, the eating time of river crabs, the average quality of river crabs and the change condition of water quality parameters in a crab pond in real time;
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 shell, is installed in a food platform center, collects underwater video data in real time, detects the residual bait amount and crab shell length and width data, and uploads the data to a data center Web server through the wireless communication module; the water quality detection module receives signals and makes decisions by using a microcontroller, the dissolved oxygen content, the temperature and the pH value of the water body are respectively monitored by the dissolved oxygen sensor and the pH sensor, water quality parameters are uploaded to a data center Web server side by the wireless communication module, and the Web client side reads data from the server side and provides graphical page display data information.
The invention relates to a working method of an electronic food table device for river crab cultivation based on machine vision, which is used for carrying out data acquisition through the electronic food table device and comprises the following steps:
step S1: placing an electronic food table, and randomly extracting river crab and bait samples in the crab pond;
step S2: the method comprises the following steps of using a water quality detection module of an electronic food platform to obtain parameters of dissolved oxygen content, temperature and pH value of a water body in the river crab culture process in real time, and uploading water quality parameters to a data center through a wireless communication module;
and step S3: the method comprises the following steps that an underwater target detection module of an electronic food table is used, the residual amount of baits and the length and width of crab shells are detected in real time through an improved target recognition algorithm, and detection results are uploaded to a data center through a wireless communication module;
and step S4: the data center Web server receives, processes and stores data such as the bait residual rate, the river crab eating duration, the river crab quality, the water quality parameters and the like monitored by the electronic food table device, and the Web client reads the data from the server and provides graphical page display data information.
Further, the step S3 specifically includes the following steps:
step S31: constructing a river crab and bait picture data set, and dividing the data set into a training set and a test set after framing and labeling river crab and bait images;
step S32: extracting features, namely building a lightweight network feature extraction model based on improved YOLOv 5; replacing the ordinary convolution of the backbone network with GhostNet convolution, firstly generating an original feature map by 1 multiplied by 1 ordinary convolution operation
Then, the original feature maps are subjected to deep separable convolution operation one by one to generate redundant feature maps
Splicing the original characteristic diagram and the redundant characteristic diagram to generate a characteristic diagram, adding a CoordAttention attention mechanism between a backhaul structure and a Head structure of the YOLOv5 network, and capturing cross-channel information, direction perception information and position perception information to enable the model to more accurately position and identify an interested target;
step S33: inputting the feature map generated in the step S32 into a Neck part, obtaining three feature maps with different sizes through a bidirectional fusion backbone network (FPN + PAN) structure, generating candidate frames on the three feature maps with different sizes, and then obtaining a detection frame of river crabs and bait targets based on a loss function and reverse propagation;
step S34: the regression target frame Loss function module adopts the SIoU to calculate the Loss function, and calculates by combining the overlapping degree, the central distance, the length-width ratio and the angle between the real frame and the prediction frame, wherein the SIoU Loss function Loss is less SIoU The calculation method comprises the following steps:
wherein,
b is a prediction block, B gt Is a real frame; />
The loss value is the combination of the distance and the angle; />
Based on the distance lost>
Is the abscissa of the center point of the real frame, and>
for predicting the abscissa of the center point of the frame, < > or>
Is the ordinate of the center point of the real frame>
As ordinate of the center point of the prediction box, c w As the lateral distance of the real frame from the center point of the predicted frame, c h The longitudinal distance between the center point of the real frame and the center point of the prediction frame is taken as the distance between the center point of the real frame and the center point of the prediction frame; γ =2-sin (2 α), which is the angular loss, α is the angle between the real and predicted frames; />
For aspect ratio loss, θ is an aspect ratio attention parameter, <' > based on the length of the cover, and >>
w is the length of the prediction box, w gt Is the length of the real box, h is the width of the predicted box, h gt The width of the real frame;
step S35: and evaluating the improved model by using the target detection evaluation index, selecting a parameter model with the highest detection precision, deploying the parameter model on the mobile phone, detecting the target in real time in water by using the mobile phone, and uploading the residual bait quantity and the crab shell length and width pixel data to a data center.
Further, the step S4 specifically includes the following steps:
step S41: calculating the bait residual rate of the river crabs for eating according to the residual bait quantity, and calculating the eating time of the river crabs;
step S42: placing a plurality of reference objects with the same size on a feeding table, correcting the detected crab shell length and width pixels, and converting the length and width pixels into river crab mass m by the following conversion method:
wherein, a w1 ，a w2 The crab shell width attention parameter is obtained; b w1 ，b w2 The length attention parameter of the crab shell is obtained; w is a rp For the corrected actual width of the crab shell,
w op p (X) is a correction ratio function in the X-axis direction and is a crab shell width pixel; h is rp For the corrected actual length of the crab shell>
h op The crab shell length pixel is used, and p (Y) is a correction ratio function in the Y-axis direction;
step S43: quantitatively adjusting the bait distribution coefficients of each region of the crab pond by combining the bait residual rate, the feeding time, the river crab quality and the water quality environmental parameters, wherein the adjusting method comprises the following steps:
wherein k is 0 (x, y) is the bait distribution coefficient of the initial crab pond subregion,
M h (x, y) is the mass distribution density of the river crab, S (x, y) is the area of the subarea, E d (x, y) sub-area water quality environment parameter coefficients; c. C f The residual rate of the bait is; c. C f0 The threshold value of the residual rate of the bait is set; h is f For the river crab to eat for a long time h f0 Is the threshold value of the eating time of the river crabs. The electronic feeding table detects the bait remaining rate, the average quality of the river crabs and the water quality parameters in real time, and when the bait remaining rate is smaller than the threshold value of the bait remaining rate, the feeding duration of the river crabs is recorded; and if the water quality parameter exceeds a suitable river crab growth warning value, an alarm is sent out through the data center to remind a user.
The invention has the beneficial effects that: according to the invention, the YOLOv5 model is improved based on light weight, the calculated amount of the model is reduced through GhostNet convolution, and the running speed of the model on the mobile terminal equipment is accelerated. Bait and river crab targets on the electronic feeding platform are detected in real time through the mobile terminal device, the bait residual rate, the feeding time length and the size and quality of river crabs are obtained, pond water quality environment parameters are obtained through the sensor, quantitative adjustment of bait distribution in each area of the crab pond is achieved, the river crab breeding cost is saved while the sufficient feeding of the river crabs is ensured, water pollution is prevented, the method has the advantages of comprehensive data management, high detection accuracy and strong real-time performance, and the purpose of accurate breeding of the river crabs can be achieved.
Drawings
FIG. 1 is a schematic diagram of the electronic dining table system of the present invention;
FIG. 2 is a block diagram of an electronic dining table;
FIG. 3 is a general flow diagram of an underwater machine vision based target detection method;
FIG. 4 is a flow chart of a river crab quality estimation method;
fig. 5 is a flow chart of a bait feeding amount adjusting method.
Detailed Description
Embodiments of the present invention are further described below with reference to the accompanying drawings.
As shown in fig. 1, the electronic dining table system is composed of a plurality of electronic dining table monitoring points, a Web server and a Web client. In the crab pond, monitoring water quality parameters, the residual amount of baits and crab shell length and width pixels by an electronic dining table monitoring point at the same time, and uploading the parameters to a cloud server through a wireless network; the cloud server processes, analyzes and stores the monitoring data into a database, and adjusts the bait feeding amount; and the Web client displays the monitoring data in real time, and gives an alarm to the user when the monitoring data is abnormal.
As shown in fig. 2, the electronic food table device is composed of an underwater target detection module, a water quality detection module, an observation platform and a bracket, a wireless communication module and the like. Wherein the length of the observation platform is 1m, the width of the observation platform is 1m, and the surface of the observation platform is white; the bracket 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 shell, is arranged in the center of a food platform, is 1m high away from the observation platform, has a vertically downward shooting angle, collects underwater video data in real time, detects the residual bait amount and crab shell length and width data, and uploads the data to a data center Web server through the wireless communication module; the water quality detection module takes STM32F407ZGT6 as a microcontroller, and respectively measures the dissolved oxygen content, temperature and pH value of the water body by a fluorescence method dissolved oxygen sensor and a pH composite electrode sensor, and uploads water quality parameters to a data center Web service end through a 4G module.
As shown in fig. 3, because the memory and the computing resources of the mobile phone end are limited, on the premise of ensuring the accuracy, the yollov 5 target detection algorithm is improved, the common convolution of the original network model is replaced by the GhostNet convolution, and the original characteristic diagram is generated by the 1 × 1 common convolution operation
Then, the original feature maps are subjected to a depth separable convolution operation one by one to generate redundant feature maps->
And then, splicing the original characteristic diagram and the redundant characteristic diagram to generate a characteristic diagram, greatly reducing the parameter quantity and the calculated quantity, adding a CoordAttention attention mechanism between a Backbone and a Head structure of the YOLOv5 network, capturing cross-channel information, direction perception information and position perception information, enabling the model to more accurately position and identify an interested target, and improving the detection precision of the model. In order to improve the convergence rate of training, a loss function is calculated by using the sio, and the calculation is performed by combining the overlapping degree, the center distance, the aspect ratio and the angle between the real frame and the prediction frame, so that the sio loss function calculation method is as follows:
wherein,
b is a prediction box, B gt Is a real frame; />
The loss value is the combination of the distance and the angle;/>
based on the distance lost>
Is the abscissa of the center point of the real frame, and>
for predicting the abscissa of the center point of the frame, < > or>
Is the ordinate of the center point of the real frame>
As ordinate of the center point of the prediction box, c w As the lateral distance of the real frame from the center point of the predicted frame, c h The longitudinal distance between the center point of the real frame and the center point of the prediction frame is taken as the distance between the center point of the real frame and the center point of the prediction frame; γ =2-sin (2 α), which is the angular loss, α is the angle between the real and predicted frames; />
For aspect ratio loss, θ is an aspect ratio attention parameter, <' > based on the length of the cover, and >>
w is the length of the prediction 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.
Bait and river crab sample images are obtained through a feeding platform, an image sample is processed to make a data set, an improved YOLOv5 target detection algorithm is used for training image data, and an improved model is evaluated by target detection evaluation indexes Precision, recall, AP (average accuracy) and mAP (average class AP). As shown in formula (2), TP is the number of correctly predicted positive samples, FP is the number of false positives for predicting negative samples as positive samples, precision represents the proportion of true positive samples in all results of predicting as positive samples; as shown in formula (3), FN is the false alarm number for predicting a positive sample into a negative sample, recall indicates that the exact positive sample proportion is predicted in all true positive samples; as shown in equation (4), AP is the area under the PR curve to measure the quality of the model in each class, and the AP is calculated by using an approximation method, where N is the total number of samples, k is the index of each sample point, Δ R (k) = R (k) -R (k-1); average of AP for all classes of maps as shown in equation (5). And then selecting a parameter model with the highest detection precision, deploying the model on a mobile phone, and identifying the bait and the river crab target.
As shown in fig. 4, the length and width pixels of the object with equal size in the image are not equal, so the length and width pixels of the crab shell are corrected first. Firstly, a plurality of reference objects (such as a unitary coin with the diameter of 2.5 cm) with the same size are placed on a table, and the central position (x, y) and the length and width pixels (w) of each reference object on an image are detected through a modified Yolov5 algorithm op ，h op ) Measuring the ratio of the length and width pixels of the object to the actual length and width of the object at different positions of different images, fitting a large amount of data to obtain a functional relation of the ratio of the length and width pixels of the object to the actual length and width of the object at different positions, wherein the function of the ratio of the length and width pixels of the object to the actual length and width of the object in the X-axis direction is shown in formula (6), wherein w is op Is an object width pixel, w rp Is actually wide of the objectDegree; the ratio function with respect to the Y-axis direction is shown in formula (7), where h op Is an object length pixel, h rp Is the actual length of the object; then detecting the center position (x, y) and the length-width pixel (w) of the crab shell on the image by a modified YOLOv5 algorithm op ，h op ) And correcting the actual width and the actual length of the crab shell respectively through formulas (8) and (9); then, the actual length and width of the corrected crab shell are substituted into a relational expression of the river crab mass and the crab shell length and width, as shown in a formula (10), the mass of a single river crab is obtained, and finally, the average value of the river crab mass is obtained, so that the aim that the bait feeding amount can be adjusted along with the change of the river crab mass is achieved.
m＝1.353×w rp 0.874 +0.1441×h rp 3.551 (10)
The flow of the bait feeding amount adjusting method is shown in figure 5, and the feeding amount is adjusted through four factors of the river crab feeding time, the bait remaining rate, the average river crab quality and the water quality parameter. When the residual rate of the baits on the feeding table is smaller than the threshold value of the residual rate of the baits, recording that the feeding of the river crabs is finished, and recording the feeding time length of the river crabs, so that the bait feeding amount is adjusted along with the feeding time length and the residual rate of the baits; the detection of the average river crab quality is also carried out in real time, so that the bait feeding amount is adjusted along with the change of the river crab quality; 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 normal growth of the river crabs, an alarm is sent to a user, and the bait feeding amount is adjusted according to the real-time water quality parameters, wherein the adjustment is shown as a formula (11):
wherein k is 0 (x, y) is the bait distribution coefficient of the initial crab pond subarea,
M h (x, y) is the mass distribution density of the river crab, S (x, y) is the area of the subarea, E d (x, y) sub-area water quality environment parameter coefficients; c. C f The residual rate of the bait is; c. C f0 Is the threshold value of the residual rate of the bait; h is a total of f Duration of food intake h f0 A meal duration threshold.
In summary, the electronic feeding platform device for river crab cultivation based on machine vision and the working method thereof integrate a sensor technology, a machine vision algorithm and an internet of things communication technology, improve a Yolov5 target recognition algorithm in a light weight manner aiming at a practical application scene, introduce a Coordattention attention mechanism to enhance the feature extraction capability of a network and a SIoU target frame loss function to accelerate the convergence rate during network training, then obtain an underwater video through an underwater target detection module, detect the quantity of feeding platform baits and the length and width pixels of river crab shells based on the improved Yolov5 target recognition algorithm, upload the detection results to a data center through a wireless network, and process, analyze and store data by the data center to obtain the residual rate of the baits, the feeding time of the river crabs and the average quality of the river crabs. Through the water quality detection module, three key factors of the river crab culture process, namely the dissolved oxygen content, the water temperature and the pH value in the crab pond, are obtained, and the data center quantitatively adjusts the bait feeding amount. The method has the advantages of comprehensive data management, high detection accuracy and strong real-time performance, can ensure that the river crabs have sufficient food intake, can improve the bait utilization rate, reduce water pollution, improve the culture benefit and achieve the aim of accurately culturing the river crabs.
Claims (4)
1. An electronic food table device for river crab cultivation based on machine vision is characterized by comprising an electronic food table and a data center, wherein the electronic food table comprises an underwater target detection module, a water quality detection module, an observation platform, a support and a wireless communication module and is used for monitoring the residual amount of bait, the eating time of river crabs, the average quality of river crabs and the change condition of water quality parameters in a crab pond in real time;
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 shell, is arranged in a food table center, acquires underwater video data in real time, detects the residual amount of baits and the length and width data of crab shells, and uploads the data to a data center Web server through the wireless communication module; the water quality detection module receives signals and makes decisions by using a microcontroller, the dissolved oxygen content, the temperature and the pH value of the water body are respectively monitored by the dissolved oxygen sensor and the pH sensor, water quality parameters are uploaded to a data center Web server side by the wireless communication module, and the Web client side reads data from the server side and provides graphical page display data information.
2. A working method of an electronic feeding table device for river crab cultivation based on machine vision is characterized in that data acquisition is carried out through the electronic feeding table device, and the working method comprises the following steps:
step S1: placing an electronic food table, and randomly extracting river crab and bait samples in the crab pond;
step S2: the method comprises the following steps of using a water quality detection module of an electronic food platform to obtain parameters of dissolved oxygen content, temperature and pH value of a water body in the river crab culture process in real time, and uploading water quality parameters to a data center through a wireless communication module;
and step S3: the method comprises the following steps that an underwater target detection module of an electronic food table is used, the residual amount of baits and the length and width of crab shells are detected in real time through an improved target recognition algorithm, and detection results are uploaded to a data center through a wireless communication module;
and step S4: the data center Web server receives, processes and stores data such as the bait remaining rate, the river crab eating duration, the river crab quality, the water quality parameters and the like monitored by the electronic dining table device, and the Web client reads the data from the server and provides graphical page display data information.
3. The working method of the electronic feeding platform device for river crab cultivation based on machine vision as claimed in claim 2, wherein the step S3 specifically comprises the following steps:
step S31: constructing a river crab and bait picture data set, and dividing the data set into a training set and a test set after framing and labeling river crab and bait images;
step S32: extracting features, namely building a lightweight network feature extraction model based on improved YOLOv 5; replacing the common convolution of the main network by GhostNet convolution, firstly generating an original feature map by 1 multiplied by 1 common convolution operation
Then, carrying out depth separable convolution operation on the original feature maps one by one to generate redundant feature maps->
Then, splicing the original characteristic diagram and the redundant characteristic diagram to generate a characteristic diagram, adding a CoordAttention attention mechanism between the Backbone and the Head structure of the YOLOv5 network, and capturing cross-channel information, direction perception information and position perception information to enable the model to more accurately position and identify an interested target;
step S33: inputting the feature map generated in the step S32 into a Neck part, obtaining three feature maps with different sizes through a bidirectional fusion backbone network (FPN + PAN) structure, generating candidate frames on the three feature maps with different sizes, and then obtaining a detection frame of river crabs and bait targets based on a loss function and reverse propagation;
step S34: a regression target frame loss function module, which adopts SIoU to calculateCalculating a Loss function by combining the overlapping degree, the central distance, the length-width ratio and the angle between the real frame and the prediction frame, and obtaining the SIoU Loss function Loss SIoU The calculation method comprises the following steps:
wherein,
b is a prediction block, B gt Is a real frame; />
The loss value is the combination of the distance and the angle; />
Based on the distance lost>
Is the abscissa of the center point of the real frame, and>
for predicting the abscissa of the center point of the frame, < > or>
Is the ordinate of the center point of the real frame>
As ordinate of the center point of the prediction box, c w Is the lateral distance of the real frame from the center point of the predicted frame, c h The longitudinal distance between the center point of the real frame and the center point of the prediction frame is taken as the distance between the center point of the real frame and the center point of the prediction frame; γ =2-sin (2 α), which is the angular loss, α is the angle between the real and predicted frames; />
For aspect ratio loss, θ is an aspect ratio attention parameter, <' > based on the length of the cover, and >>
w is the length of the prediction box, w gt Is the length of the real box, h is the width of the predicted box, h gt The width of the real frame;
step S35: and evaluating the improved model by using the target detection evaluation index, selecting a parameter model with the highest detection precision, deploying the parameter model on the mobile phone, detecting the target in real time in water by using the mobile phone, and uploading the residual bait quantity and the crab shell length and width pixel data to a data center.
4. The working method of the electronic feeding platform device for river crab cultivation based on machine vision as claimed in claim 3, wherein the step S4 specifically comprises the following steps:
step S41: calculating the bait residual rate of the river crab eating according to the residual bait quantity, and calculating the eating time of the river crab;
step S42: placing a plurality of reference objects with the same size on a feeding table, correcting the detected crab shell length and width pixels, and converting the length and width pixels into river crab mass m, wherein the conversion method comprises the following steps:
wherein, a w1 ，a w2 The crab shell width attention parameter is obtained; b is a mixture of w1 ，b w2 The crab shell length attention parameter is obtained; w is a rp For the corrected actual width of the crab shell,
w op for crab shell width pixels, p (X) is a correction ratio function in the X-axis direction; h is rp For the corrected actual length of the crab shell>
h op The crab shell length pixel is used, and p (Y) is a correction ratio function in the Y-axis direction;
step S43: quantitatively adjusting the bait distribution coefficient of each area of the crab pond by combining the bait residual rate, the feeding time, the river crab quality and the water quality environmental parameters, wherein the adjusting method comprises the following steps:
wherein k is 0 (x, y) is the bait distribution coefficient of the initial crab pond subregion,
M h (x, y) is the mass distribution density of the river crab, S (x, y) is the area of the subarea, E d (x, y) sub-area water quality environment parameter coefficients; c. C f The residual rate of the bait is; c. C f0 The threshold value of the residual rate of the bait is set; h is a total of f For the river crab to eat for a long time h f0 Is the threshold value of the eating time of the river crabs. The electronic feeding table detects the bait remaining rate, the average quality of the river crabs and the water quality parameters in real time, and when the bait remaining rate is smaller than the threshold value of the bait remaining rate, the feeding duration of the river crabs is recorded; and if the water quality parameter exceeds the suitable river crab growth warning value, an alarm is sent out through the data center to remind the user. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211662847.7A CN115843733A (en) | 2022-12-23 | 2022-12-23 | Machine vision-based electronic feeding table device for river crab cultivation and working method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211662847.7A CN115843733A (en) | 2022-12-23 | 2022-12-23 | Machine vision-based electronic feeding table device for river crab cultivation and working method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115843733A true CN115843733A (en) | 2023-03-28 |
Family
ID=85654183
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211662847.7A Pending CN115843733A (en) | 2022-12-23 | 2022-12-23 | Machine vision-based electronic feeding table device for river crab cultivation and working method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115843733A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117256545A (en) * | 2023-11-21 | 2023-12-22 | 安徽农业大学 | Intelligent feeding monitoring device and monitoring system thereof |
-
2022
- 2022-12-23 CN CN202211662847.7A patent/CN115843733A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117256545A (en) * | 2023-11-21 | 2023-12-22 | 安徽农业大学 | Intelligent feeding monitoring device and monitoring system thereof |
| CN117256545B (en) * | 2023-11-21 | 2024-02-02 | 安徽农业大学 | Intelligent feeding monitoring device and monitoring system thereof |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107094683B (en) | Automatic feeding and water quality monitoring control system for aquaculture | |
| CN109460753B (en) | Method for detecting floating object on water | |
| US20210073573A1 (en) | Ship identity recognition method based on fusion of ais data and video data | |
| CN113537106B (en) | Fish ingestion behavior identification method based on YOLOv5 | |
| CN111222574B (en) | Ship and civil ship target detection and classification method based on multi-model decision-level fusion | |
| CN112213962A (en) | Intelligent feeding system and method based on growth model and sonar feedback | |
| CN110826514A (en) | Construction site violation intelligent identification method based on deep learning | |
| CN114724022B (en) | Method, system and medium for detecting farmed fish shoal by fusing SKNet and YOLOv5 | |
| Huang et al. | The prototype of a smart underwater surveillance system for shrimp farming | |
| CN206281972U (en) | Maritime radar monitoring system | |
| CN115843733A (en) | Machine vision-based electronic feeding table device for river crab cultivation and working method | |
| CN116778310A (en) | Acoustic-optical image fusion monitoring method and system for aquaculture | |
| CN113484867A (en) | Imaging sonar-based fish school density detection method in closed space | |
| CN114898405B (en) | Portable broiler chicken anomaly monitoring system based on edge calculation | |
| CN115578423A (en) | Fish key point detection, individual tracking and biomass estimation method and system based on deep learning | |
| CN117115688A (en) | Dead fish identification and counting system and method based on deep learning under low-brightness environment | |
| CN112883969A (en) | Rainfall intensity detection method based on convolutional neural network | |
| Liu et al. | Evaluation of body weight of sea cucumber Apostichopus japonicus by computer vision | |
| CN107907645A (en) | A kind of water quality early-warning device and method of aquatic animal behavioural information driving | |
| CN116128822A (en) | Grassland vegetation coverage monitoring method based on continuous photography | |
| CN115661081A (en) | Aquaculture automatic control water changing method based on image processing | |
| CN114119662A (en) | Image processing method and system in fish detection visual system | |
| CN115600829A (en) | Method and device for predicting culture water quality | |
| CN112651958A (en) | Aquaculture environment real-time online monitoring method based on Internet of things and artificial intelligence and big data cloud platform | |
| CN113284164A (en) | Shrimp swarm automatic counting method and device, electronic equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |