Device and method for fish organism intelligent identification tracking and state judgment

Abstract

The embodiment of the invention provides a device and a method for intelligent identification tracking and state judgment of fish organisms, which are used for collecting and identifying fish school information of a water area to be detected, a tracking identification platform, and a water area processing assembly, a bait throwing assembly and an identification assembly which are respectively arranged on the tracking identification platform, the tracking identification platform is arranged on the bank of a water area to be detected; the bait throwing assembly is used for throwing bait into the water area to be detected so as to attract fish schools; the water area treatment assembly is used for being arranged in a water area to be detected to perform water flow disturbance on the water area; and the identification assembly is used for collecting a fish school video of the water area to be detected and identifying the type and growth state information of each fish in the fish school video. According to the device and the method, a large amount of original data is obtained through long-time identification and tracking, and the fish type and fish state information in the water area to be detected is efficiently obtained based on the identification assembly.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the technical field of fish identification, and more particularly to a device and method for intelligent identification, tracking and status judgment of fish organisms, and a computer-readable storage medium. Background Art

[0002] In the fish farming industry, the growth status of fish and the water quality will directly affect the final output of aquatic products. The growth status of fish can only be achieved through manual sampling and testing. This solution is difficult to implement, and the sample status results finally obtained are not universal.

[0003] On the other hand, for larger rivers and lakes, it is impossible to monitor fish organisms through traditional manual sampling schemes, and long-term dynamic tracking through existing dynamic image monitoring technology can only make a certain degree of judgment on the number of fish schools, but cannot obtain effective information on the actual status of fish organisms. It is difficult to judge the growth of fish populations and water quality in a certain water area through the collected data. Summary of the invention

[0004] The purpose of the embodiments of the present disclosure is to provide a device and method for intelligent identification, tracking and status judgment of fish organisms, as well as a computer-readable storage medium, so as to solve the aforementioned problems existing in the prior art.

[0005] In order to achieve the above objectives, the technical solutions adopted by the embodiments of the present disclosure are as follows:

[0006] The present disclosure provides a device for intelligent identification, tracking and status judgment of fish organisms, which is used to collect and identify information about fish schools in a water area to be tested. The device includes: a tracking and identification platform, and a water area processing component, a bait delivery component and an identification component respectively arranged on the tracking and identification platform;

[0007] Tracking and identification platform, used to be set up on the shore of the water area to be tested;

[0008] A bait delivery component is used to deliver bait to the water area to be tested to attract fish;

[0009] A water area processing component is used to be arranged in the water area to be tested to drive the water flow in the water area;

[0010] The identification component is used to collect fish school videos in the water area to be tested and identify the type and growth status information of each fish in the fish school video.

[0011] Optionally, the tracking and identification platform includes: a vertical bracket and a horizontal bracket, one end of the vertical bracket is used to be set on the shore of the water area to be measured, and the other end is connected to the horizontal bracket;

[0012] The horizontal support is located above the water area, wherein the horizontal support is provided with the identification component and the bait delivery component.

[0013] Optionally, the tracking and identification platform further comprises: a vertical mounting plate, one end of which is connected to the horizontal bracket and the other end of which is used to extend into the water area;

[0014] A connecting plate is provided between the vertical mounting plate and the vertical bracket, so a supporting plate is provided on the other end of the vertical mounting plate facing away from the shore of the water area;

[0015] The water area treatment component includes: a water pump, a pumping pipe and a drainage pipe; the water pump is arranged on the support plate, the water inlet of the water pump is connected to one end of the pumping pipe, the water outlet is connected to one end of the drainage pipe, the other end of the pumping pipe is close to the shore of the water area, and the other end of the drainage pipe is far away from the shore of the water area.

[0016] Optionally, the axial direction of the water pump is perpendicular to the vertical mounting plate, the water inlet of the water pump faces the shore of the water area, and the water outlet of the water pump is away from the shore of the water area;

[0017] The other end of the pumping pipe passes through the support plate and extends along the extending direction of the water area shore;

[0018] The other end of the drainage pipe extends away from the shore of the water area, and the drainage pipe is at a preset angle to the vertical mounting plate.

[0019] Optionally, the bait delivery assembly includes: a bait trough, a channel, a solenoid valve, a floating plate and a partition net;

[0020] The channel passes through the horizontal bracket; the bait trough is provided at one end of the channel located above the horizontal bracket, the other end of the channel is used to extend into the water area, the other end of the channel is provided with the float plate, and the other end is provided with the partition net, the solenoid valve is provided in the other end of the channel, and is used to adjust the opening and closing of the channel.

[0021] Optionally, the identification component includes: an image acquisition device, a lighting device and a scanner respectively arranged on the horizontal support, the image acquisition device is used to acquire a video of the water area to be measured;

[0022] The lighting device is arranged within a preset distance range of the image acquisition device, and is used to provide lighting or attract fish schools when the image acquisition device is capturing videos;

[0023] The scanner is electrically connected to the lighting device and the image acquisition device respectively to control the start and stop of the lighting device, control the image acquisition device to capture video, and identify and analyze the video to obtain the type and growth status of each fish in the video. The scanner is equipped with a visual recognition module; when the bait delivery component includes a solenoid valve, the scanner is electrically connected to the solenoid valve to control the opening and closing of the solenoid valve.

[0024] Another aspect of the present disclosure provides a method for intelligent identification, tracking and status judgment of fish organisms, the method comprising:

[0025] Build the above-mentioned device and set the tracking and identification platform on the shore of the water area to be tested;

[0026] The water area treatment component and the identification component are respectively started, and the water area treatment component drives the water flow in the water area so that the water close to the shore is drawn in and then discharged in a direction away from the shore;

[0027] The bait delivery component delivers the bait into the water area, and the water treatment component drives the water flow, so that the bait spreads in the water area to attract fish.

[0028] The recognition component collects video of the water area to be tested, and controls the frequency of the bait delivery component delivering bait according to the acquisition frequency of the original image, so as to fully obtain the fish information.

[0029] Another aspect of the present disclosure provides a method for intelligent identification, tracking and status judgment of fish organisms, the method comprising:

[0030] Obtaining a video of a school of fish in a water area where the fish are to be tracked and identified, wherein the video of the school of fish is obtained based on an identification component of the device as described above built on the shore of the water area;

[0031] Input each image frame in the video into a pre-trained fish state comparison model in sequence to obtain the type and state results of each fish in the corresponding image; wherein the fish state comparison model is based on: YOLOv5 network training;

[0032] Determine whether the fish state is abnormal. If so, mark the abnormal fish and save the image to which the abnormal fish belongs; if not, save the image to which the normal fish belongs.

[0033] Optionally, the training process of the fish state control model includes:

[0034] Acquire an image data set having at least one fish, and mark each fish in the fish data set with a circumscribed rectangular frame, species or type, and state, wherein the state includes normal and abnormal;

[0035] The convolutional neural network is trained with the labeled fish image data set to obtain a trained fish state comparison model.

[0036] Another aspect of the present disclosure provides a device for intelligent identification, tracking and status judgment of fish organisms, the device comprising:

[0037] An acquisition module is used to acquire a video of a school of fish in the water area to be tested, wherein the video of the school of fish is acquired according to the recognition component of the above-mentioned device built on the shore of the water area to be tested;

[0038] The detection module is used to input each image frame in the video into a pre-trained fish state comparison model in sequence to obtain the type and state results of each fish in the corresponding image; wherein the fish state comparison model is based on: YOLOv5 network training;

[0039] The judgment module is used to judge whether the state of the fish is abnormal. If so, the abnormal fish will be marked and the image of the abnormal fish will be saved; if not, the image of the normal fish will be saved.

[0040] On the other hand, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for intelligent identification, tracking and status judgment of fish organisms as described above.

[0041] The beneficial effects of the embodiments of the present disclosure are:

[0042] The device for intelligent identification, tracking and status judgment of fish organisms in the disclosed embodiment has a simple structure and can efficiently judge the growth status of fish organisms in the water area to be tested. The water area processing component, i.e., the identification area processing system, is used to divert the bottom of the water area to be tested, and the bait delivery component delivers bait to attract fish, both of which improve the efficiency of obtaining effective data of the water area to be tested. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of a structure for intelligent identification, tracking and status judgment of fish organisms proposed in an embodiment of the present disclosure;

[0044] Figure 2 for Figure 1 Schematic diagram of the structure of the mid-water treatment component;

[0045] Figure 3 for Figure 1 Schematic diagram of the structure of the bait delivery component;

[0046] Figure 4 A schematic diagram of a method flow for intelligent identification, tracking and status judgment of fish organisms proposed in an embodiment of the present disclosure;

[0047] Figure 5 A schematic flow chart of another method for intelligent identification, tracking and status judgment of fish organisms proposed in an embodiment of the present disclosure;

[0048] Figure 6 This is a schematic diagram of the structure of another device for intelligent identification, tracking and status judgment of fish organisms proposed in an embodiment of the present disclosure.

[0049] In the figure: 1. Tracking and identification platform; 2. Water area treatment component; 3. Bait delivery component; 4. Image acquisition equipment; 5. Scanner; 6. Lighting equipment; 7. Pallet; 8. Water pump; 9. Pumping pipe; 10. Drainage pipe; 11. Bait trough; 12. Channel; 13. Float; 14. Partition net. DETAILED DESCRIPTION

[0050] In order to make the purpose, technical solution and advantages of the embodiment of the present disclosure more clear, the embodiment of the present disclosure is further described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation methods described here are only used to explain the embodiment of the present disclosure and are not used to limit the embodiment of the present disclosure.

[0051] like Figure 1 As shown, on the one hand, an embodiment of the present disclosure proposes a device for intelligent identification, tracking and status judgment of fish organisms, which is used for collecting and identifying information of fish schools in a water area to be tested. The device includes: a tracking and identification platform 1, and a water area treatment component 2, a bait delivery component 3 and an identification component (not shown in the figure) respectively arranged on the tracking and identification platform; the tracking and identification platform is used to be set on the shore of the water area to be tested; the bait delivery component is used to deliver bait to the water area to be tested to attract fish schools; the water area treatment component 2 is used to be set in the water area to be tested to disturb the water flow in the water area; the identification component is used to collect fish school videos in the water area to be tested, and identify the type and growth status information of each fish in the fish school video.

[0052] The device of the embodiment of the present disclosure is set in the water area to be tested, and multiple devices can be set according to actual conditions. The device is convenient for collecting videos of fish schools in the water area to be tested, and can identify the type and growth status information of each fish from the video. Specifically, the type and size of each fish are identified, and the corresponding fish growth status is obtained according to the type and size of each fish. The growth status refers to abnormal or normal.

[0053] The water area processing component drives the water environment of the water area to be tested according to the actual situation, so that the water body keeps flowing continuously; the bait can also be released through the bait release component to spread the bait and increase the probability of fish approaching. The identification component is used to collect fish organisms in the water area to be tested, which is convenient for subsequent identification and tracking. The identification component collects the original image of the fish school in the water area to be tested. The identification component collects video, and cuts out pictures from the video to identify each fish, and realizes tracking in the collected video screen, etc. Specifically, the identification component obtains and saves the video in real time, and then identifies and analyzes the collected video to achieve interception and archiving of effective content such as the fish photographed.

[0054] As a specific example of a tracking and identification platform, the tracking and identification platform includes: a vertical bracket (not shown in the figure) and a horizontal bracket (not shown in the figure), one end of the vertical bracket is used to be set on the shore of the water area to be tested, and a rib is provided on the side facing the shore, and the other end is connected to the horizontal bracket; the horizontal bracket is located above the water area, wherein the horizontal bracket is provided with the identification component and the bait delivery component 3.

[0055] The device of the disclosed embodiment is arranged on the shore of the water area to be tested, and collects and identifies the fish schools near the shore. The tracking and identification platform has a simple structure and is easy to build. The ribs are arranged on the side of the vertical bracket facing the shore. There can be two ribs, which are arranged on both sides of the vertical bracket. The ribs are arranged on the shore, which can strengthen the structural strength of the vertical bracket and avoid bending of the vertical bracket. The ribs can be right-angled triangles, and the two right-angled sides are respectively arranged on the vertical bracket and the shore ground. The vertical bracket can be a vertical bracket, and the horizontal bracket can be a horizontal bracket. The holes in the triangle plate can improve the balance and make it more stable when in use. One end of the tracking and identification platform can be directly embedded into the ground of the shore of the water area to be tested using long bolts. The specific fixing method can be carried out according to the actual situation and will not be repeated here.

[0056] like Figure 2 As shown, as a specific example of a tracking and identification platform, the tracking and identification platform also includes: a vertical mounting plate, one end of which is connected to the horizontal bracket, and the other end is used to extend into the water area; a connecting plate is arranged between the vertical mounting plate and the vertical bracket, so a support plate is arranged on the other end of the vertical mounting plate away from the water area shore; the water area treatment component includes: a water pump, a water suction pipe and a drainage pipe; the water pump is arranged on the support plate, the water inlet of the water pump is connected to one end of the water suction pipe, the water outlet is connected to one end of the drainage pipe, the other end of the water suction pipe is close to the water area shore, and the other end of the drainage pipe is far away from the water area shore.

[0057] The vertical mounting plate in the disclosed embodiment should be longer than the vertical bracket, one end of the vertical bracket is set on the shore, and the other end of the vertical mounting plate is set in the water, and the length of the vertical mounting plate is at least longer than the length of the vertical bracket above the ground surface. The vertical mounting plate is also a vertical mounting plate. Specifically, the length of the vertical bracket and the vertical mounting plate can be set according to actual conditions, the distance between the vertical mounting plate and the shore can be set according to actual conditions, and the distance between the vertical mounting plate and the vertical bracket can also be set according to actual conditions.

[0058] The connecting plate makes the tracking and identification platform structure stable. No matter how the water pump is arranged, in order to increase the flow area of ​​the driven water flow, the water inlet of the water pumping pipe is close to the shore of the water area, and the water outlet of the water discharge pipe is far away from the shore.

[0059] Preferably, two connecting plates are provided, one ends of the two connecting plates are arranged close to each other on the vertical bracket and one of the vertical mounting plates, and the other ends are arranged at a preset distance on the vertical mounting plate and the other of the vertical mounting plates.

[0060] As a specific example of a water area treatment component, the axial direction of the water pump is perpendicular to the vertical mounting plate, its water inlet faces the shore of the water area, and its water outlet is away from the shore of the water area; the other end of the pumping pipe passes through the support plate and extends along the extension direction of the shore of the water area; the other end of the drainage pipe extends away from the shore of the water area, and the drainage pipe is at a preset angle to the vertical mounting plate.

[0061] Specifically, the extension line of the drainage pipe is perpendicular to the vertical mounting plate to increase the area for driving the water flow. The water area treatment component drives the water flow to the water area to be tracked and identified through a water pump. The water pump draws water near the water suction port of the pumping pipe and discharges it toward the water area to be tracked and identified. It is best that the water suction port is not directly opposite the shore. Multiple drainage ports can be set at the other end of the drainage pipe, and distributed in the direction away from the shore, which is not limited here. After the water pump is placed on a support plate, the pumping pipes and drainage pipes at both ends are respectively connected to the water area at the bottom. The water close to the shore is pumped out with the help of the water pump and discharged outward into the water area.

[0062] A support plate is installed at the bottom of the vertical mounting plate of the tracking and identification platform, and a pumping pipe is connected to the side of the support plate. One end of the pumping pipe is fixedly connected to the water inlet of the water pump, and the water outlet of the water pump is connected to one end of the drainage pipe. The water pump draws water from the side of the support plate through the position of the pumping pipe and then discharges it out from the position of the drainage pipe, thereby achieving the flow effect of this part of the water body.

[0063] like Figure 3As shown, as a specific example of a bait delivery component, the bait delivery component 3 includes: a bait tank 11, a channel 12, a solenoid valve, a floating plate 13 and a partition net 14;

[0064] The channel 12 passes through the horizontal bracket; the bait trough 11 is provided at one end of the channel 12 located above the horizontal bracket, and the other end of the channel 12 is used to extend into the water area. The other end of the channel 12 is provided with the float 13, and the other end is provided with the partition net 14. The solenoid valve is provided in the other end of the channel for adjusting the opening and closing of the channel.

[0065] When the bait delivery assembly of the disclosed embodiment is used, the bait is delivered into the top bait trough, and the bait inside enters the inner side of the partition net along the bottom channel and is directly immersed below the water surface. An electromagnetic valve can be arranged in the channel and electrically connected to the rear scanner to realize automatic control of bait delivery.

[0066] Specifically, the partition net can be directly opposite to the outlet of the drainage pipe. The floating plate can reduce the pulling force on the horizontal bracket after the bait is put in. The horizontal bracket of the tracking and identification platform is interspersed with a channel 12, and a bait trough 11 is provided at the top of the channel 12. After the bait is put into the bait trough 11, it can enter the partition net 14 at the bottom along the channel 12. Since the bait is directly immersed under the water surface, it can be melted by the water flow. After the bait is melted, the melted bait can be dispersed by the water flow discharged from the drainage outlet of the rear end drainage pipe, and the melted bait is diffused toward a distance to achieve the effect of attracting fish. This process can be used and controlled according to the collection efficiency of the recognition component for the video of fish in the water area. The use state of the bait delivery component is controlled according to the collection frequency of the original picture collected. This process can be used and controlled according to the collection efficiency of the bottom fish photos by the picture acquisition camera 4. When the fish photos cannot be recognized for a long time, the solenoid valve inside the channel can be opened to release the bait.

[0067] As a specific example of the identification component, the identification component includes: an image acquisition device 4, a lighting device 6 and a scanner 5 respectively arranged on the horizontal support, the image acquisition device 4 is used to collect videos of the water area to be tested; the lighting device 6 is arranged within a preset distance range of the image acquisition device 4, and is used to provide lighting or attract fish schools when the image acquisition device 4 collects videos; the scanner 5 is electrically connected to the lighting device and the image acquisition device 4 respectively to control the start and stop of the lighting device, control the image acquisition device 4 to collect videos, and identify and analyze the videos to obtain the type and growth status of each fish in the video, wherein the scanner is equipped with a visual recognition module. When the bait delivery component includes a solenoid valve, the scanner is electrically connected to the solenoid valve to control the opening and closing of the solenoid valve.

[0068] like Figure 1 As shown, the horizontal support includes a first-step support and a second-step support, one end of the first-step support is connected to the vertical support and the vertical mounting plate respectively, the other end of the first-step support is connected to one end of the second-step support through two fixing plates on the side facing the water surface, the second-step support is lower than the first-step support, that is, the second-step support is closer to the water surface than the first-step support, and the two fixing plates can be tilted and arranged in parallel. The double-layer of the horizontal support increases the structural strength and improves the aesthetics.

[0069] The image acquisition device and the lighting lamp of the lighting device are arranged on the side of the horizontal bracket facing the water surface, the scanner is arranged on the side of the horizontal bracket facing away from the water surface, and the scanner is close to the image acquisition device. Specifically, the image acquisition device and the lighting lamp of the lighting device are arranged on the side of the second-stage bracket facing the water surface, and the scanner is arranged on the side of the second-stage bracket facing away from the water surface. The channel of the bait delivery component is passed through the second-stage bracket. A fixed tube is arranged on the side of the other end of the first-stage bracket facing away from the water surface, and the fixed tube is used to pass through a connecting tube. The connecting line of the image acquisition device is arranged in the connecting tube and connected to the power supply; the connecting tube is wound above the scanner and passes through the fixed tube. The connecting tube can be a hose or a hard tube. The lighting device can also be connected to the power supply through a cable, and similarly, the cable can also be connected to the power supply through a connecting tube.

[0070] The scanner is equipped with a recognition module, and the recognition module can be a K210 visual recognition module, for example. The visual recognition module recognizes and analyzes the image through edge computing, and the scanner can also be embedded with a wireless communication module to wirelessly connect to the host computer. The image acquisition device may include a camera; the lighting device may include a searchlight.

[0071] In the disclosed embodiment, a searchlight 6 is built at the end of the horizontal support of the tracking and identification platform, that is, the searchlight 6 is built at the end of the second-stage support away from the shore, and the searchlight 6 is used to illuminate the bottom area to be collected by the camera 4. A scanner 5 is placed on the horizontal support of the tracking and identification platform. After the camera 4 collects and shoots videos of fish in the water area at the bottom, the scanner 5 completes subsequent feature comparison and status judgment.

[0072] The tracking and identification platform may also be installed with a water quality detection device, which may be installed on a vertical mounting plate or a horizontal bracket and hoisted above the water surface, and the water quality monitoring probe may be immersed in water to obtain water quality data.

[0073] like Figure 4 As shown, another aspect of the present disclosure provides a method for intelligent identification, tracking and status judgment of fish organisms, the method comprising:

[0074] Step S110: construct the above-mentioned device and place the tracking and identification platform on the shore of the water area to be tested.

[0075] The method of the embodiment of the present disclosure is performed by fixing a vertical bracket on the shore of the water area to be tracked and identified, and adjusting the image acquisition device of the identification component so that the acquisition range covers the water area to be measured.

[0076] Step S120, respectively start the water area processing component and the identification component, the water area processing component drives the water flow in the water area, so that the water close to the shore is drawn in and then discharged in the direction away from the shore.

[0077] In step S120, the image acquisition device of the recognition component is turned on, and the lighting device, i.e., the searchlight, built at the end of the tracking and recognition platform can also be turned on to illuminate the bottom of the recognition component image acquisition device, i.e., the camera, to facilitate the camera to collect video. Lighting can also attract fish.

[0078] Step S130: the bait delivery component delivers the bait into the water area, and the water area treatment component drives the water flow, so that the bait spreads in the water area to attract fish.

[0079] The disclosed embodiment can attract fish by placing baits through the bait placing component, and can also gather fish at the bottom of the image acquisition device, that is, the camera, by means of the phototropism of organisms. Both the lighting device and the bait placing can be turned on according to the actual situation.

[0080] Step S140, the recognition component collects the video of the water area to be tested, and controls the frequency of the bait delivery component delivering bait according to the acquisition frequency of the original image, so as to fully and accurately obtain the fish information.

[0081] A conventional solenoid valve structure is installed inside the channel of the bait delivery component, and the scanner is electrically connected to the solenoid valve. When the scanner detects that no fish photo can be recognized within the preset time, the solenoid valve inside the channel can be controlled to open, and the bait inside the bait tank can be delivered into the water to attract fish. The preset time can be set according to actual conditions.

[0082] like Figure 5 As shown, another aspect of the present disclosure provides a method for intelligent identification, tracking and status judgment of fish organisms, the method comprising:

[0083] Step S210: obtaining a video of a school of fish in the water area where the fish are to be tracked and identified, wherein the video of the school of fish is obtained based on the identification component of the above-mentioned device built on the shore of the water area.

[0084] Before step S210, fish schools may be attracted to the water area to be tested by illumination or bait placement, and a video of the water area to be tested may be acquired by using an image acquisition device of the recognition component.

[0085] Step S220, input each image frame in the video into a pre-trained fish state control model in sequence to obtain the type and state results of each fish in the corresponding image; wherein the fish state control model is based on: YOLOv5 network training.

[0086] In the disclosed embodiment, pictures in the video can also be extracted according to a preset frame extraction frequency, which can speed up the detection efficiency, and the pictures obtained by frame extraction are detected and analyzed to obtain the type of each fish and the corresponding target detection frame. The target detection is the circumscribed rectangular frame of the fish, and the fish status is obtained according to the size and type of the target detection frame of the corresponding fish. The target detection frame is regarded as the size of the fish. For pictures where no fish is detected, the pictures are directly abandoned and the next picture is selected.

[0087] Step S230, determine whether the state of the fish is abnormal. If the state of the fish is abnormal, mark the abnormal fish and save the image of the abnormal fish; if the state of the fish is normal, also save the image of the normal fish.

[0088] As a specific example, after the image frames in the video are input into a pre-trained fish status control model in sequence, the method also includes: determining whether fish status detection is being performed, and if so, feeding back the fish status image; otherwise, continuing to input the next frame of image to detect the status of each fish in the next frame of image, and saving the final execution result.

[0089] The disclosed embodiment can detect and identify fish after they have grown up. By determining whether fish status detection is being performed, it is possible to avoid the image from mistakenly identifying other branches or garbage. Steps S220 and S230 of the disclosed embodiment are both implemented by the video recognition module in the scanner. Determine the type and size of each fish in each frame of the image, and judge the growth status of the fish based on the type and corresponding size of the fish. If a fish is too large or too small and exceeds the normal range, then the fish is growing abnormally. Each fish in a frame of the image needs to be detected, and the abnormal fish are marked. If there is an abnormal fish, this abnormal image is also saved if all the fish in this frame are normal, and continue to detect the next frame of the image. If no fish is detected in the image, the image is not saved.

[0090] As a specific example of the fish identification method, after obtaining the video of the school of fish in the water area where the fish to be tracked and identified is located, the method further includes:

[0091] The frequency of bait delivery is controlled according to the frequency of original image acquisition, so as to control the number of fish collected, until the frequency of original image acquisition is within the preset time range, the fish approaching the water area to be tested are identified and tracked, and the original images of the fish in the water area are collected. The original image frequency is the number of times the image acquisition device collects video.

[0092] As a specific example of the fish state control model, the training process of the fish state control model includes:

[0093] Step S221, obtaining an image data set having at least one fish, and marking an outer rectangular frame, species or type, and state of each fish in the fish data set, wherein the state includes normal and abnormal;

[0094] In step S221, the disclosed embodiment independently labels each fish in each picture, and labels different abnormality types according to the acquired fish abnormality big data model. There is at least one fish in an image. When there are multiple fish, each one is labeled, and the circumscribed rectangular frame is labeled based on experience, and the size of the fish is determined according to the area of ​​the labeled frame. Depending on the type of fish, the body size of the fish is too large or too small, and the incomplete phenomenon is abnormal. That is to say, for a certain type of fish, the body size of the fish is larger than the first preset size or smaller than the second preset size, and the presence of incompleteness is abnormal, otherwise it is normal. The body size can be the length or width of the fish. The disclosed embodiment can label adult fish.

[0095] Step S222: Use the labeled image data set to train the YOLOv5 network to obtain a trained fish state comparison model.

[0096] The device constructed according to the embodiment of the present disclosure can not only obtain videos of fish schools in water areas, but also identify the species and detect the status of individual fish in the fish school video. The growth status of individual fish can be determined based on the size of the detection frame of each fish. Here, the growth status mainly refers to whether the individual is normal or abnormal, and the individual size is determined by the area of ​​the detection frame.

[0097] like Figure 6 As shown, the embodiment of the present disclosure proposes a device for intelligent identification, tracking and status judgment of fish organisms, the device comprising:

[0098] The acquisition module 210 is used to acquire the video of the school of fish in the water area to be tested, wherein the video of the school of fish is acquired according to the recognition component of the above-mentioned device built on the shore of the water area to be tested.

[0099] The detection module 220 is used to input each image frame in the video into a pre-trained fish state control model in sequence to obtain the type and state results of each fish in the corresponding image; wherein the fish state control model is based on: YOLOv5 network training;

[0100] The judging module 230 is used to judge whether the state of the fish is abnormal. If so, the abnormal fish is marked and the image to which the abnormal fish belongs is saved; if not, the image to which the normal fish belongs is saved.

[0101] The detection module 220 and the determination module 230 are both disposed in the scanner.

[0102] On the other hand, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for intelligent identification, tracking and status judgment of fish organisms as described above.

[0103] By adopting the above technical solution disclosed in the embodiment of the present disclosure, the following beneficial effects are obtained:

[0104] 1. In the method for intelligent tracking, identification and status judgment of fish organisms, a tracking and identification platform is built on the shore of the water area to be tested. A large amount of original video data is obtained through long-term tracking and identification, and detection is performed based on the trained fish status control model, so as to efficiently judge the growth status of fish organisms and water quality data in the water area.

[0105] 2. The method for intelligent tracking, identification and status judgment of fish organisms uses the identification area processing system carried by the tracking and identification platform to divert the bottom of the water area to be tested, so that the water surface of the top identification area flows continuously. The flow effect prevents non-fish organisms from blocking and covering the tracking and identification area, provides protection for the tracking and identification area, and thus improves the efficiency of the method in obtaining effective data.

[0106] 3. The method for intelligent tracking, identification and status judgment of fish organisms is implemented by monitoring the waters near the shore area, so it is easy to identify the situation that makes it difficult for fish to approach. By setting a bait delivery component at the bottom of the tracking and identification area to accelerate the gathering of fish schools, the image acquisition device, that is, the efficiency of sample video obtained by the camera, is improved, and the fish biological status results obtained within a certain period of time are more accurate and reliable.

[0107] The following is a detailed description of the fish tracking and status judgment using the device for intelligent identification, tracking and status judgment of fish organisms according to the embodiment of the present disclosure, which is mainly implemented according to the following steps:

[0108] S1. First, collect and prepare a data set required for training, wherein the data set includes images of at least one fish.

[0109] The quality and quantity of the data set have an important impact on the training results. The process includes data acquisition and data preprocessing. Data acquisition is obtained through various channels, including network download, direct import of existing professional data sets, or subsequent large-scale video acquisition and image extraction through the device of the embodiment of the present disclosure. Ensure that the data source is reliable and meets the training objectives. In this embodiment, a large number of fish state pictures are downloaded from the network to complete; data preprocessing: build a suitable model architecture. The model architecture used in this embodiment includes convolutional neural network (CNN) and recurrent neural network (RNN).

[0110] S2. Manually label the data set: clarify the types of information that need to be labeled, including: labeling the outer frame, type and status of each fish, using the fish outer frame as the specified rule, and the labeling format is: {(class), (x1, y1), (x2, y2)}, where (x1, y1) is the coordinate point of the upper left corner of the largest circumscribed rectangle of a single fish, and (x2, y2) is the coordinate point of the lower right corner of the largest circumscribed rectangle of the same single fish. And label the type and status of the fish together.

[0111] S3. Establish a big data model, train the convolutional neural network model according to the data set, and obtain a high-accuracy fish state control model. The fish state control model can be obtained based on YOLOv5 network training. The training process includes model initialization: initializing the convolutional neural network model parameters for pre-training model; defining the loss function using cross entropy loss, mean square error, etc.; and using the Adam optimization algorithm to adjust the model parameters to minimize the loss function.

[0112] S4. Build a tracking and identification device. First, select the water area for subsequent fish tracking and identification, and install a tracking and identification platform 1 on the shore beside the water area. The bottom of the tracking and identification platform 1 uses the water area processing component 2 to continuously flow the water environment of the tracking and identification area. At the same time, the bait delivery component 3 is used to increase the probability of fish approaching. The identification component image acquisition device 4, that is, the camera, is used to identify and track the approaching fish organisms, and the corresponding original video images are collected. The video is collected, and the pictures captured from the video are used for target detection.

[0113] Specifically, the tracking and identification platform 1 is constructed by fixing a vertical bracket on the shore of the water area to be tracked and identified, and the water treatment component 2 drives the water flow in the water area to be tracked and identified through a water pump 8. The water pump 8 draws water near the water inlet of the water pumping pipe 9 and discharges it toward the area to be tracked and identified. A searchlight 6 is built at the end of the horizontal bracket of the tracking and identification platform 1, and the bottom of the image acquisition device 4, i.e., the camera, is illuminated by the searchlight 6. A scanner 5 is placed on the top of the horizontal bracket of the tracking and identification platform 1. After the image acquisition device 4, i.e., the camera, collects video of the fish at the bottom of the horizontal bracket, the scanner 5 completes subsequent feature comparison and state judgment processes.

[0114] Among them, a support plate 7 is installed at the bottom of the vertical bracket of the tracking and identification platform 1, and a pumping pipe 9 is connected to the side of the support plate 7. One end of the pumping pipe 9 is fixedly connected to the water inlet of the water pump 8, and the water outlet of the water pump 8 is connected to the drainage pipe 10. The water pump 8 draws water from the side through the position of the pumping pipe 9 and then discharges it outward from the position of the drainage pipe 10, thereby realizing the flow effect of this part of the water body.

[0115] In this embodiment, fish are gathered at the bottom of the image acquisition device 4, that is, the camera, by means of the phototaxis of organisms. At the same time, a bait delivery component 3 is built at the bottom of the tracking and identification platform 1. The use state of the bait delivery component 3 is controlled according to the frequency of the original image collected. After the bait is delivered into the bait trough 11, it is accumulated inside the partition net 14 at the bottom. The bait is diffused by means of the melting of the bait and the water area treatment component 2, and the number of fish in the image acquisition device camera 4 is controlled until the image acquisition device frequency of the image acquisition device camera 4 is controlled within the set range.

[0116] A channel 12 is interspersed at the bottom of the tracking and identification platform 1, and a bait trough 11 is set at the top of the channel 12. After the bait is thrown into the inside through the bait trough 11, it can enter the bottom isolation net 14 along the channel 12. Since the bait is directly immersed under the water surface, it can be melted by the water flow. After the bait is melted, the water flow can be discharged from the inside of the drainage pipe 10 with the help of the water pump 8 at the rear end, and the melted bait can be diffused to a distance to achieve the effect of attracting fish. This process can be used and controlled according to the image acquisition device 4, that is, the camera's acquisition efficiency of the bottom fish video.

[0117] S7, based on dynamic recognition technology, uses the self-attention mechanism. The bottom video screen captured by the camera. Make sure to capture pictures from the video.

[0118] S8. Perform feature comparison based on the fish state control model. In the original video collected, local extreme points can form feature points, so by comparing the pixel points with all adjacent points; the direction of the key points is statistically analyzed using the gradient histogram. With the key point as the origin, the main direction is determined by the distribution of the key directions of the pixel points in a specific area. The detection frame of the fish in each frame of the video is obtained in the above manner, i.e., the outer frame.

[0119] S9. Use the trained fish state control model to judge the obtained state, display the tracking and identification species and life state judgment results in each frame image, and save the fish abnormal state frame image at the same time to determine whether fish state detection is being performed. If so, feedback the fish state and abnormal state frame image, otherwise continue to extract pictures from the video and repeat the recognition process, and save the final execution result.

[0120] Determine whether fish status detection is in progress to avoid misidentifying other branches and garbage in the image. The outer frame of the fish in different states is used as the main reference direction of the big data model. The real-time fish movement state collected is compared with the outer frame of the established training data in different states to determine the fish state.

[0121] The above is only a preferred implementation of the embodiment of the present disclosure. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the embodiment of the present disclosure. These improvements and modifications should also be considered as the protection scope of the embodiment of the present disclosure.

Claims

1. A device for intelligent identification, tracking and status judgment of fish organisms, used for collecting and identifying fish school information in the water area to be tested, characterized in that: The device comprises: a tracking and identification platform, and a water area processing component, a bait delivery component and an identification component respectively arranged on the tracking and identification platform; Tracking and identification platform, used to be set up on the shore of the water area to be tested; A bait delivery component is used to deliver bait to the water area to be tested to attract fish; A water area processing component is used to be arranged in the water area to be tested to disturb the water flow in the water area; The identification component is used to collect fish school videos in the water area to be tested and identify the type and growth status information of each fish in the fish school video.

2. The device according to claim 1, characterized in that The tracking and identification platform includes: a vertical bracket and a horizontal bracket, one end of the vertical bracket is used to be set on the shore of the water area to be measured, and the other end is connected to the horizontal bracket; The horizontal support is located above the water area, wherein the horizontal support is provided with the identification component and the bait delivery component.

3. The device according to claim 2, characterized in that The tracking and identification platform further includes: a vertical mounting plate, one end of which is connected to the horizontal bracket and the other end of which is used to extend into the water area; A connecting plate is provided between the vertical mounting plate and the vertical bracket, so a supporting plate is provided on the other end of the vertical mounting plate facing away from the shore of the water area; The water area treatment component includes: a water pump, a pumping pipe and a drainage pipe; the water pump is arranged on the support plate, the water inlet of the water pump is connected to one end of the pumping pipe, the water outlet is connected to one end of the drainage pipe, the other end of the pumping pipe is close to the shore of the water area, and the other end of the drainage pipe is far away from the shore of the water area.

4. The device according to claim 3, characterized in that The axial direction of the water pump is perpendicular to the vertical mounting plate, the water inlet of the water pump faces the shore of the water area, and the water outlet of the water pump is far away from the shore of the water area; The other end of the pumping pipe passes through the support plate and extends along the extending direction of the water area shore; The other end of the drainage pipe extends away from the shore of the water area, and the drainage pipe is at a preset angle to the vertical mounting plate.

5. The device according to claim 1, characterized in that The bait delivery assembly comprises: a bait trough, a channel, a solenoid valve, a floating plate and a partition net; The channel passes through the horizontal bracket; the bait trough is provided at one end of the channel located above the horizontal bracket, the other end of the channel is used to extend into the water area, the other end of the channel is provided with the float plate, and the other end is provided with the partition net, the solenoid valve is provided in the other end of the channel, and is used to adjust the opening and closing of the channel.

6. The device according to any one of claims 2 to 5, characterized in that: The recognition component comprises: an image acquisition device, a lighting device and a scanner respectively arranged on the horizontal support, wherein the image acquisition device is used to acquire a video of the water area to be measured; The lighting device is arranged within a preset distance range of the image acquisition device, and is used to provide lighting or attract fish schools when the image acquisition device is capturing videos; The scanner is electrically connected to the lighting device and the image acquisition device respectively to control the lighting device to start and stop, control the image acquisition device to collect video, and perform recognition and analysis on the video to obtain the type and growth status of each fish in the video, wherein the scanner is equipped with a visual recognition module; When the bait delivery assembly includes a solenoid valve, the scanner is electrically connected to the solenoid valve to control the opening and closing of the solenoid valve.

7. A method for intelligent identification, tracking and status judgment of fish organisms, characterized in that: The method comprises: Build the device described in any one of claims 1 to 6 above, and place the tracking and identification platform on the shore of the water area to be tested; The water area treatment component and the identification component are respectively started, and the water area treatment component drives the water flow in the water area so that the water close to the shore is drawn in and then discharged in a direction away from the shore; The bait delivery component delivers the bait into the water area, and the water treatment component drives the water flow, so that the bait spreads in the water area to attract fish. The recognition component collects video of the water area to be tested, and controls the frequency of bait release by the bait release component according to the acquisition frequency of the original image, so as to fully obtain fish information.

8. A method for intelligent identification, tracking and status judgment of fish organisms, characterized in that: The method comprises: Acquire a video of a school of fish in a water area where fish are to be tracked and identified, wherein the video of the school of fish is acquired by using an identification component of the device according to any one of claims 1 to 6 built on the shore of the water area; Input each image frame in the video into a pre-trained fish state comparison model in sequence to obtain the type and state results of each fish in the corresponding image; wherein the fish state comparison model is based on: YOLOv5 network training; Determine whether the fish state is abnormal. If so, mark the abnormal fish and save the image to which the abnormal fish belongs; if not, save the image to which the normal fish belongs.

9. A device for intelligent identification, tracking and status judgment of fish organisms, characterized in that: The device comprises: An acquisition module, used for acquiring a video of a school of fish in the water area to be tested, wherein the video of the school of fish is acquired according to the recognition component of any one of the apparatuses of claims 1 to 6 built on the shore of the water area to be tested; The detection module is used to input each image frame in the video into a pre-trained fish state comparison model in sequence to obtain the type and state results of each fish in the corresponding image; wherein the fish state comparison model is based on: YOLOv5 network training; The judgment module is used to judge whether the state of the fish is abnormal. If so, the abnormal fish will be marked and the image of the abnormal fish will be saved; if not, the image of the normal fish will be saved.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the steps of the method as claimed in claim 8 are implemented.

Images