Portable fish detection device and method

Abstract

The invention discloses a portable fish detection device and method, and relates to the field of fish detection, and the device comprises a fish luring module which is used for periodically emitting blue light and green light for fish luring; the data acquisition module is used for acquiring fish image data after fish luring; the data processing module is connected with the data acquisition module and is used for identifying the type of the fish by adopting a built-in fish image identification model according to the fish image data; the fish image recognition model is obtained by training an improved YOLOv5 model by a training sample; the improved YOLOv5 model comprises a coordinate attention mechanism and an adaptive spatial feature fusion module. According to the invention, the fish detection speed and precision can be improved.

Description

Technical Field

[0001] This application relates to the field of fish detection, and particularly to a portable fish detection device and method. Background Art

[0002] In current fish surveys, there are various survey methods and fishing gears, including net fences, gillnets, trawls, fish traps, electrofishing, etc. Among them, electrofishing causes great damage to fish bodies and the surrounding water bodies, and electrofishing is prohibited at many sampling points; methods such as casting nets have high requirements for net-casting techniques, often resulting in empty nets, low work efficiency, and excessive time waste for multi-point sampling surveys. In addition, existing sonar fish finders can judge the size, quantity, and position of fish, but cannot well judge the species of fish. Summary of the Invention

[0003] To solve the above problems, this application provides a portable fish detection device and method.

[0004] To achieve the above object, this application provides the following solutions:

[0005] In a first aspect, this application provides a portable fish detection device, including:

[0006] A fish attracting module, configured to periodically emit blue light and green light to attract fish;

[0007] A data acquisition module, configured to acquire fish image data after attracting fish;

[0008] A data processing module, connected to the data acquisition module, configured to identify the species of fish according to the fish image data by using a built-in fish image recognition model; the fish image recognition model is obtained by training an improved YOLOv5 model with training samples; the improved YOLOv5 model includes a coordinate attention mechanism and an adaptive spatial feature fusion module.

[0009] Optionally, the fish attracting module includes:

[0010] A color-changing LED lamp, configured to periodically emit blue light and green light;

[0011] A white light illuminating lamp, configured to irradiate white light for auxiliary illumination;

[0012] A central control system, respectively connected to the color-changing LED lamp and the white light illuminating lamp, configured to control the color-changing LED lamp and the white light illuminating lamp.

[0013] Optionally, the data acquisition module includes: an underwater high-definition camera, configured to acquire fish image data after attracting fish.

[0014] Optionally, the data acquisition module further includes:

[0015] A multibeam sonar sensor, which uses frequency modulation continuous wave technology and echo analysis technology to determine the distance between fish and the image acquisition point and the number of fish;

[0016] A current meter, which measures the water flow velocity through the Doppler effect;

[0017] A water quality sensor, which measures the water environment conditions where the fish are located;

[0018] A GPS, which records the position of the image acquisition point and the time of image acquisition.

[0019] Optionally, the data processing module is further configured to perform positioning and tracking of fish using the DeepSORT algorithm according to the recognition result of the fish image recognition model.

[0020] Optionally, the portable fish detection device further includes:

[0021] A display control module, which is respectively connected to the fish attracting module and the data processing module, and is used to control the working mode of the fish attracting module and visualize the processing result of the data processing module.

[0022] Optionally, the portable fish detection device further includes:

[0023] A power supply module, which provides electrical energy for the fish attracting module, the data acquisition module, the data processing module and the display control module.

[0024] Optionally, the carrier of the portable fish detection device is a telescopic rod, which includes a telescopic first rod body, a second rod body and a third rod body. The power supply module and the data processing module are arranged on the first rod body. A protective cover is arranged at one end of the third rod body far away from the second rod body. The fish attracting module and the data acquisition module are arranged in the protective cover.

[0025] In a second aspect, the present application provides a portable fish detection method, including:

[0026] Attracting fish by periodically emitting blue light and green light;

[0027] Collecting fish image data after attracting fish;

[0028] According to the fish image data, using a fish image recognition model to identify the species of fish; the fish image recognition model is obtained by training an improved YOLOv5 model with training samples; the improved YOLOv5 model includes a coordinate attention mechanism and an adaptive spatial feature fusion module.

[0029] According to the specific embodiments provided by this application, this application has the following technical effects:

[0030] (1) By periodically emitting blue light and green light through the fish attracting module to attract fish, it will not cause damage to the fish body and the surrounding water body.

[0031] (2) By introducing the CA (Coordinate Attention) mechanism and the ASFF (Adaptive Spatial Feature Fusion) module into the improved YOLOv5 model, the introduction of the CA mechanism solves the problem of feature extraction and recognition of fish under complex backgrounds and occlusion conditions, and the ASFF module can adaptively fuse features of different scales to improve the overall detection performance. Therefore, this application can improve the speed and accuracy of fish detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the following will briefly introduce the drawings required for use in the embodiments. Obviously, the drawings in the following description are only some embodiments of this application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0033] Figure 1 Schematic diagram of the functional modules of the portable fish detection device provided by an embodiment of this application;

[0034] Figure 2 Schematic diagram of the working process of the fish attracting module;

[0035] Figure 3 Schematic diagram of the working process of the data acquisition module;

[0036] Figure 4 Schematic diagram of the working process of the data processing module;

[0037] Figure 5 Schematic diagram of the structure of the portable fish detection device equipped with a telescopic rod. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0038] The following will clearly and completely describe the technical solutions in the embodiments of this application with reference to the drawings in the embodiments of this application. Obviously, the described embodiments are only a part of the embodiments of this application, rather than all of them. Based on the embodiments in this application, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the scope of protection of this application.

[0039] To make the above objects, features, and advantages of this application more obvious and understandable, the following will further describe this application in detail with reference to the drawings and specific embodiments.

[0040] In an exemplary embodiment, as Figure 1 shown, a portable fish detection device is provided, including: a fish attracting module, a data acquisition module, and a data processing module. The fish attracting module is used to periodically emit blue light and green light to attract fish; the data acquisition module is used to acquire fish image data after attracting fish; the data processing module, connected to the data acquisition module, is used to identify the species of fish according to the fish image data by using a built-in fish image recognition model; the fish image recognition model is obtained by training the improved YOLOv5 model with training samples; the improved YOLOv5 model includes a coordinate attention mechanism and an adaptive spatial feature fusion module.

[0041] In a specific embodiment, as Figure 2 shown, the fish attracting module includes a color-changing LED lamp, a white light illuminating lamp, and a central control system. The color-changing LED lamp is used to periodically emit blue light and green light; the white light illuminating lamp is used to irradiate white light for auxiliary illumination; the central control system, connected to the color-changing LED lamp and the white light illuminating lamp respectively, is used to control the color-changing LED lamp and the white light illuminating lamp.

[0042] According to relevant research, blue light (450 - 495nm) and green light (495 - 570nm) have strong attraction to fish in deep water, and the phototaxis of different fish is different. Therefore, the fish attracting module will control the color-changing LED lamp to periodically emit blue light and green light through the central control system, so as to attract fish more efficiently. In addition, for different water environments, it is necessary to adjust the illumination intensity and incident angle of the light. For example, when blue light is used at night or under low light conditions, the light intensity is usually controlled at 1500 - 2000lux to avoid fish escaping due to too strong light, so as to ensure the best fish attracting effect. For areas with high water transparency, a larger incident angle (more than 60 degrees) is used to attract fish to provide a wider coverage range; while for areas with turbid water quality, a smaller incident angle (about 30 degrees) is used to attract fish, which helps to concentrate the light, enhance the penetration in deep water, and improve the effective coverage range of the light. The white light illuminating lamp provides white light for auxiliary illumination, and the switch is controlled by the central control system, which can effectively improve the quality of image acquisition in an environment with relatively dark underwater light. The intensity of the white light is controlled within 2000lux, which can not only ensure the image quality but also reduce the interference to fish, thus improving the fish perception accuracy.

[0043] In a specific embodiment, as Figure 3As shown in the figure, the data acquisition module includes: an underwater high-definition camera, a multi-beam sonar sensor, a current meter, a water quality sensor, and a GPS. The underwater high-definition camera is used to collect fish image data after attracting fish; the multi-beam sonar sensor is used to determine the distance between the fish and the image acquisition point and the number of fish by using frequency-modulated continuous wave technology and echo analysis technology; the current meter is used to measure the water flow velocity through the Doppler effect; the water quality sensor is used to measure the water environment conditions where the fish are located; the GPS is used to record the position of the image acquisition point and the time of image acquisition.

[0044] The underwater high-definition camera adopts advanced optical separation technology and low-light imaging technology, which can effectively reduce the interference of water impurities on the image in turbid waters, thereby improving the imaging effect. The high-resolution images taken can assist in determining the species of fish.

[0045] The multi-beam sonar sensor adopts frequency-modulated continuous wave (FMCW) technology and echo analysis technology, and can accurately detect the distance between the fish and the image acquisition point and the number of fish. where N is the number of fish, P i is the echo intensity of the i-th fish, σ is the sonar cross-section, and thus the number of the fish school can be calculated. Specifically, the intensity of the echo signal received by the sonar is closely related to the number and density of the fish. By statistically analyzing the echo signal, the number of fish per unit volume, that is, the fish school density, can be estimated. In addition, combined with the spatial range detected by the sonar, the overall scale and distribution of the fish school can be deduced. This method for evaluating fish school density based on echo statistics can effectively evaluate the fish school density under the conditions of a small fish school density and sufficient statistical samples.

[0046] The current meter measures the water flow velocity through the Doppler effect. where V is the water flow velocity, Δf is the frequency shift, c is the speed of sound, and f 0 is the transmission frequency. The water quality sensor is used to measure the water environment conditions where the fish are located, including pH value, dissolved oxygen, conductivity, turbidity, and temperature. The GPS is used to record the position of the image acquisition point and the time of image acquisition. The data collected by the high-definition camera, multi-beam sonar sensor, current meter, water quality sensor, and GPS will be transmitted to the data processing module through a data transmission line.

[0047] In a specific embodiment, such as Figure 4As shown in the figure, the data processing module is responsible for analyzing and processing the data collected by the data acquisition module to achieve the perception and recognition of fish species characteristics. The data processing module has a fish image recognition model trained based on deep learning. The data processing module has pre-collected images of different fish in water as training samples to train the improved YOLOv5 model, so as to obtain the fish image recognition model. The YOLOv5 model is the fifth-generation model in the YOLO (You Only Look Once) series, continuing the lightweight and high efficiency of the YOLO family. It can simultaneously achieve target localization and classification in one detection, with extremely high real-time performance and accuracy. However, when dealing with complex environments such as similar underwater fish appearances and easily deformed fish bodies, the recognition accuracy of the traditional YOLOv5 model still has certain limitations. To solve these problems, this application has made structural improvements to the YOLOv5 model, especially introducing the CA (Coordinate Attention) mechanism and the ASFF (Adaptive Spatial Feature Fusion) module. The introduction of the CA mechanism solves the problems of feature extraction and recognition of fish under complex backgrounds and occlusion conditions. Traditional convolutional neural networks do not effectively consider the dependence relationship between spatial positions and channels when extracting features. The CA mechanism effectively enhances the feature representation ability by simultaneously paying attention to space and channels. Its basic operation is as follows:

[0048] X' = X·σConv 1×1 ((F spatial )·F channel )

[0049] Among them, X is the input feature map, and X' is the feature map after the CA mechanism; F spatial is to capture spatial features through convolution. F channel is channel attention, which re-weights features through the dependence relationship between channels; σ is the Sigmoid activation function, used to map the weights to the [0,1] interval.

[0050] The ASFF module solves the problem of feature information loss during multi-scale detection. The multi-scale detection of the traditional YOLOv5 model relies on simple feature splicing, and the information transmission is not sufficient. The ASFF module can adaptively fuse features of different scales to improve the overall detection performance. The basic operation of the ASFF module is as follows:

[0051] F out = α 1 ·F 1 + α 2 ·F 2 + α 3 ·F 3

[0052] Among them: F 1, F 2 , F 3 are feature maps of different scales. α 1 , α 2 , α 3 are adaptive weights generated through convolution and are used to control the contributions of features at different scales. These improvements enable the YOLOv5 model to have stronger adaptability when dealing with fish with similar appearances and large morphological variations. Especially in the environment of high-density fish schools, the improved YOLOv5 model can maintain fast and accurate recognition, and the overall detection performance has been significantly improved. To better achieve fish recognition, this application adopts the DeepSORT algorithm (DeepSimple Online and Realtime Tracking) to enhance the multi-object tracking ability of the device in complex underwater environments. The core of the DeepSORT algorithm lies in combining object detection, trajectory prediction, and the re-identification model (ReID). The ReID model is responsible for extracting the appearance features of fish to solve the problem of identity consistency of objects under perspective changes, occlusion, or complex environmental interference. To further improve the stability of the algorithm, the ReID model is optimized, specifically including increasing the number of layers of the residual network from 6 to 9 layers, significantly enhancing the ability to extract the appearance features of fish under complex water environment conditions. The optimized ReID model can generate more accurate feature vectors of fish and work in coordination with the trajectory matching and object detection modules in DeepSORT, enabling the device to show higher accuracy in complex underwater multi-object tracking tasks.

[0053] In addition, this application also introduces the dense connection mechanism of DenseNet in the YOLOv5 model, connecting the output of each layer with the outputs of all previous layers, enhancing the reuse ability of feature maps, and effectively improving the ability to capture fish details in complex underwater environments. At the same time, a skip connection mechanism is added to the Neck structure of YOLOv5, and the multi-scale detection performance is optimized through cross-layer feature fusion, ensuring the accuracy and robustness of fish target detection at different scales.

[0054] Through the combined application of the improved YOLOv5 model and the DeepSORT algorithm, this application realizes the efficient detection and accurate tracking of underwater fish. The improved YOLOv5 model is responsible for real-time detecting and locating fish, and the DeepSORT algorithm performs multi-object association and tracking of fish based on the detection results. Combining sonar and camera data, this application can not only identify fish species but also track the movement states and positions of fish, ensuring high-precision recognition and tracking effects can still be maintained under complex water conditions.

[0055] The working process of the above-mentioned portable fish detection device provided by this application is as follows:

[0056] After the fish attracting module realizes the fish attracting function, fish image data is collected by the underwater high-definition camera, and the water environment information of the measurement point is obtained through the water quality sensor. The fish image data is transmitted to the fish image recognition model in the data processing module of the system, and information such as the category and confidence level of the fish is recognized, and a bounding box is generated to mark the position of the fish. Then, the DeepSORT algorithm is used to locate and track the fish. The detection results generated by the entire detection and tracking process include information such as the species of underwater fish, the position of the bounding box, the confidence level, and the movement trajectory. These data are sent to the user through the wireless transmission module after analysis for the user to remotely call. The user can, according to the needs, display and analyze the results of fish recognition through the visualization interface, and view key information such as the number, position, and species of the fish school in real time.

[0057] In an optional embodiment, the portable fish detection device provided by the present application further includes: a display control module and a power supply module.

[0058] The display control module is respectively connected to the fish attracting module and the data processing module, and is used to control the working mode of the fish attracting module (freely switch between intelligent acquisition and manual acquisition), and to visualize the processing results of the data processing module, so as to facilitate the user to observe and analyze the situation of the fish school.

[0059] The power supply module is used to provide electrical energy for the fish attracting module, the data acquisition module, the data processing module, and the display control module.

[0060] In an optional embodiment, as Figure 5 shown, the carrier of the above-mentioned portable fish detection device is a telescopic rod, and the interior is designed to be hollow. The telescopic rod includes a retractable first rod body, a second rod body, and a third rod body. The power supply module and the data processing module are arranged on the first rod body, and a GPS is also arranged on the first rod body. A spiral data transmission line is arranged inside the second rod body. A protective cover is arranged at one end of the third rod body away from the second rod body, and the fish attracting module and the data acquisition module are arranged inside the protective cover.

[0061] The advantages of the portable fish detection device provided by the present application are as follows:

[0062] 1. It improves the efficiency and accuracy of fish detection and provides convenience for fish scientific investigations.

[0063] 2. It overcomes the existing destructive fish investigation method of electrofishing and provides a new way for fish scientific investigation methods.

[0064] 3. It solves the problem of inconvenient hull transportation in traditional fish scientific investigation methods. Its portability can save a lot of manpower and material resources, and is fast and efficient.

[0065] 4. Achieving the perfect combination of the multi-beam sonar detection system and the underwater imaging system can not only investigate the density and species of fish, but also intuitively display the real-time images of fish, providing important information for the perception, recognition and species identification of fish schools.

[0066] Based on the same inventive concept, in an exemplary embodiment, a portable fish detection method is provided, including:

[0067] Luring fish by periodically emitting blue light and green light;

[0068] Collecting fish image data after luring fish;

[0069] According to the fish image data, using a fish image recognition model to identify the species of fish; the fish image recognition model is obtained by training the improved YOLOv5 model with training samples; the improved YOLOv5 model includes a coordinate attention mechanism and an adaptive spatial feature fusion module.

[0070] As an optional implementation manner, the portable fish detection method further includes: according to the recognition result of the fish image recognition model, using the DeepSORT algorithm to locate and track fish.

[0071] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0072] The databases involved in the embodiments provided in this application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., without limitation. The processors involved in the embodiments provided in this application can be general-purpose processors, central processors, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., without limitation.

[0073] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0074] In this article, specific examples are used to elaborate on the principles and implementation manners of this application. The descriptions of the above embodiments are only used to help understand the method and its core idea of this application; at the same time, for those of ordinary skill in the art, according to the idea of this application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to this application.

Claims

1. A portable fish detection device, characterized in that: include: A fish attracting module, used to periodically emit blue light and green light to attract fish; A data acquisition module, used for collecting fish image data after luring fish; A data processing module is connected to the data acquisition module and is used to identify the type of fish according to the fish image data using a built-in fish image recognition model; the fish image recognition model is obtained by training an improved YOLOv5 model with training samples; the improved YOLOv5 model includes a coordinate attention mechanism and an adaptive spatial feature fusion module.

2. The portable fish detection device according to claim 1, characterized in that: The fish attracting module comprises: A color-changing LED light for periodically emitting blue light and green light; White light illuminator, used for irradiating white light auxiliary lighting; The central control system is connected to the color-changing LED lamp and the white light illumination lamp respectively, and is used for controlling the color-changing LED lamp and the white light illumination lamp.

3. The portable fish detection device according to claim 1, characterized in that: The data acquisition module includes: an underwater high-definition camera, which is used to collect fish image data after luring fish.

4. The portable fish detection device according to claim 3, characterized in that: The data acquisition module also includes: Multi-beam sonar sensors, used to determine the distance of fish from the image acquisition point and the number of fish using frequency modulated continuous wave technology and echo analysis technology; Current meters, used to measure water velocity using the Doppler effect; Water quality sensors, used to measure the water conditions in which fish are exposed; GPS, used to record the location of the image acquisition point and the time of image acquisition.

5. The portable fish detection device according to claim 1, characterized in that: The data processing module is also used to locate and track the fish using the DeepSORT algorithm according to the recognition result of the fish image recognition model.

6. The portable fish detection device according to claim 1, characterized in that: The portable fish detection device also includes: The display control module is connected to the fish attracting module and the data processing module respectively, and is used to control the working mode of the fish attracting module and to visualize the processing results of the data processing module.

7. The portable fish detection device according to claim 6, characterized in that: The portable fish detection device also includes: A power supply module is used to provide electrical energy to the fish attracting module, the data acquisition module, the data processing module and the display control module.

8. The portable fish detection device according to claim 7, characterized in that: The carrying body of the portable fish detection device is a telescopic rod, which includes a telescopic first rod body, a second rod body and a third rod body. The power supply module and the data processing module are arranged on the first rod body. A protective cover is arranged at one end of the third rod body away from the second rod body. The fish attracting module and the data acquisition module are arranged in the protective cover.

9. A portable fish detection method, characterized in that: include: Fish are attracted by periodically emitting blue and green lights; Collecting fish image data after attracting fish; According to the fish image data, using a fish image recognition model to identify the type of fish; The fish image recognition model is obtained by training an improved YOLOv5 model with training samples; the improved YOLOv5 model includes a coordinate attention mechanism and an adaptive spatial feature fusion module.

10. The portable fish detection method according to claim 9, characterized in that: Portable fish detection methods also include: According to the recognition results of the fish image recognition model, the DeepSORT algorithm is used to locate and track the fish.

Images