Intelligent fishing system and method

Abstract

The invention belongs to the field of intelligent fishing, and discloses an intelligent fishing system which is provided with an advanced sensor, a camera and an AI recognition module and is used for automatically recognizing and processing fishes. When a fish enters the fishing box, the sensor detects the existence of the fish, the camera records an image of the fish and uploads the image to the AI system for feature recognition such as type, size and gender, and after recognition is completed, the system judges whether the fish is a target object or not, guides the fish to a collection net if the fish is the target fish, and releases the fish if the fish is not the target fish. A passive fishing tool, a fish recognition box and artificial intelligence control equipment are combined into a whole, and the deep influence on scientific research, monitoring and commercial fishing is achieved; different from an active fishing method which is intensive in labor and high in cost, the device is a tool which is low in manpower, high in efficiency and controllable in cost, and is suitable for various river and lake water bodies in the Yangtze River basin and the Misxiche River basin; by integrating AI identification and a fishery management model, huge advantages can be brought to researchers, resource managers and fishery practitioners.

Description

Technical Field

[0001] The present invention belongs to but is not limited to the field of smart fishing technology, and in particular relates to a smart fishing system and method. Background Art

[0002] In recent years, fish stocks in the Yangtze River have steadily declined, with many rare species even facing extinction. This decline is the result of a long-term accumulation of multiple human factors. Overfishing is the most direct cause, followed by habitat destruction and deteriorating water quality. The widespread use of simple, non-selective traditional fishing gear leads to the indiscriminate capture of target and non-target fish, disrupting the ecological balance. While the government is strengthening resource protection, the key to restoring and rebuilding sustainable fishing capacity remains the intelligent transformation of fishing gear.

[0004] Currently, the most effective removal method is the modified unified method (MUM), an improved version of China's traditional method of chasing, blocking, stabbing, and spreading fish for the four major carps. While this method significantly reduces time consumption, it still consumes significant human, financial, and time resources. Therefore, this basin urgently needs a new passive intelligent fishing method that integrates perception, decision-making, and intelligent control to improve operational efficiency, reduce human involvement, and achieve both economic and ecological resource utilization.

[0005] The closest prior art to the present application is patent CN111513013A, which proposes a "live fish sorting and conveying device based on artificial intelligence recognition." This device uses an onboard camera to capture fish images and a convolutional neural network to determine health and length. The fish are then transported to different aquaculture ponds via mechanical pipes. An earlier patent, US4963035A, uses monocular video to measure fish geometry and roughly classify them by species based on area ratio. A flap mechanism is then used within the processing line for size grading. EP2777390A1 installs size detection sensors and controllable windows within the trawl net to enable active sorting within the net. The recently reported Smartrawl system combines an AI camera with an adjustable net port, enabling immediate discarding of non-target fish during bottom trawls. Norway's Deep Vision stereo camera-acoustic composite sensor identifies species and measures size in real time at the trawl net port. Domestically, CN107691612B also relies on weight thresholds and touch switches for mechanical grading, but lacks visual recognition capabilities.

[0006] Smartrawl and Deep Vision are suitable for marine environments and are deployed in active nets. This invention will be applied to inland rivers and lakes and deployed in passive nets. Summary of the Invention

[0007] In order to solve the problems existing in the prior art, the present invention provides an intelligent fishing system. The system consists of three parts: an AI recognition system module, a fish recognition box and fishing gear. Figure 1 When fish are intercepted by the net, they swim along the guide net in search of an exit, eventually being guided through a narrow opening into the guide section of the net. The guide section typically has multiple tapered openings, allowing one-way entry for fish. A fish identification box is connected after the last tapered opening. In the fish identification box, species identification, length estimation, and sex determination are performed. If the fish is the target individual, it is introduced into the catch collection net. This process is accomplished through the operation of an AI recognition system module, which can be integrated with the fish identification box or used separately from the fish identification box.

[0008] The present invention is implemented as follows: a smart fishing system, characterized by being equipped with advanced sensors, cameras, and an AI recognition module to automatically identify and process fish. When a fish enters the fish identification box, the first sensor (Sensor1) detects its presence, and the camera records its image and uploads it to the AI ​​system for identification of species, size, gender, and other characteristics.

[0009] Furthermore, after the intelligent fishing system completes the identification, the system determines whether the fish is a target object. If it is a target fish, the first exit is opened and the fish is guided to the collection net physically or optically; the first exit is closed after the second sensor detects that the fish has left; if it is not a target fish, the second exit is opened and the fish is released; the second exit is closed after the second sensor detects that the fish has left; the system then reopens the entrance and repeats the process.

[0010] Furthermore, the smart fishing system uses a binocular camera to obtain 3D images of fish, extracts depth information and geometric relationships through stereo triangulation, and calculates the length of the fish.

[0011] Furthermore, the intelligent fishing system includes an AI species recognition system that adapts to different ecosystem environments. The recognition steps include:

[0012] (1) Use convolutional neural networks (CNNs) for image recognition to identify fish shapes, colors, and patterns;

[0013] (2) Develop a text recognition model based on taxonomic descriptions;

[0014] (3) Use transfer learning and reinforcement learning to continuously optimize model performance;

[0015] (4) Use advanced algorithms such as YOLO (You Only Look Once) or Faster R-CNN to achieve real-time positioning and classification of fish in video streams.

[0016] In combination with the above technical solutions and the technical problems solved, the advantages and positive effects of the technical solutions to be protected by the present invention are as follows:

[0017] The intelligent fishing system constructed by the present invention is based on an embedded multimodal sensor array and an edge AI computing module. It can simultaneously complete underwater behavior monitoring, target identification and data storage with a single person on duty, breaking through the pain points of previous aquatic ecological research such as "complex deployment, difficult maintenance, and slow analysis". The system can stream video, sonar and environmental parameters to the local GPU unit in real time, complete the determination of the target fish's body length, species and group density in milliseconds, and transmit it back to the cloud database via 5G or Starlink, providing scientific researchers with "large sample, long time, cross-water" continuous observation data. This end-to-end closed loop not only significantly shortens the experimental cycle and labor costs, but also allows for seamless connection between field surveys and laboratory analysis, greatly improving research efficiency and data credibility in fields such as fish behavior, population ecology and invasive biology.

[0018] In terms of monitoring and resource protection, the system's deep learning model can perform multi-label identification of invasive species, rare and endangered species, and key functional species, with a minimum detection size of up to 3 cm. Combined with a programmable diversion mechanism, the equipment can make a release or retention decision within 0.2 seconds after the individual enters the cage, achieving three-dimensional precise control of "species-specification-time period" to avoid overfishing and damage to non-target species. In continuous operation, the equipment can collect and annotate hundreds of thousands of frames of high-definition images every day, providing real-time warnings and post-event traceability for government fishery and environmental protection departments, and accumulating reusable training samples for scientific research databases, forming an ecological closed loop of monitoring-warning-management.

[0019] Technically, this invention pioneered the introduction of a Transformer-based multi-scale visual network into the passive fishing gear scenario, and combined it with an adaptive threshold-triggered pneumatic-hydraulic hybrid diversion mechanism, completely resolving the historical problem that the traditional "mesh adjustment" solution was difficult to balance resource protection and fishing efficiency. Its core algorithm performs dynamic enhancement and feature compensation training for uneven underwater lighting, turbidity changes, and dense occlusions, so that the system's Top-1 recognition accuracy in yellow and turbid water bodies remains above 92%. At the same time, the device supports OTA firmware upgrades and online model distillation, which not only ensures technological leadership but also reserves interfaces for seamless integration with emerging application scenarios such as big data platforms, blockchain traceability, and carbon sink assessments in the future. This fills the gap in passive fishing gear intelligent fishing technology at home and abroad, and lays the foundation for the intelligent infrastructure of modern sustainable fisheries. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a schematic diagram of a smart fishing system provided by an embodiment of the present invention.

[0021] Figure 2This is a structural diagram of a fish identification box provided by an embodiment of the present invention.

[0022] Figure 3 This is a workflow diagram of the smart fishing system provided by an embodiment of the present invention.

[0023] Figure 4 This is a block diagram of the structural composition of the intelligent fishing system provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0025] like Figure 1 、 2 As shown, an embodiment of the present invention provides a smart fishing system equipped with advanced sensors, cameras and AI recognition modules to automatically identify and process fish; when a fish enters a fishing box, the sensor detects its presence, the camera records its image and uploads it to the AI ​​system for species, size, gender and other feature identification. After the identification is completed, the system determines whether the fish is a target object. If it is a target fish, it will be guided to a collection net. If it is not a target fish, it will be released.

[0026] The intelligent fishing system opens the first exit if it is a target fish and guides it to the collection net physically or optically; closes the first exit after the second sensor detects that the fish has left; opens the second exit if it is not a target fish and releases it; closes the second exit after the second sensor detects that the fish has left; and then reopens the exit and repeats the process.

[0027] Figure 3 The following is a workflow diagram of the intelligent fishing system of the present invention. The specific steps are as follows:

[0028] Step S101: Fish swim into the guide net

[0029] Step S102: Fish enter the fish identification box through the entrance.

[0030] Step S103: The first sensor detects a fish entry signal and starts image acquisition.

[0031] Step S104: The camera collects fish body images, and the binocular system obtains 3D structure data.

[0032] Step S105: The AI ​​recognition system processes the image, extracts features and identifies information such as fish species, gender, and body length.

[0033] Step S106: Determine whether it is the target fish:

[0034] If it is the target fish, go to step S107;

[0035] If it is not the target fish, go to step S108.

[0036] Step S107: Open the first outlet to guide the fish into the collection device, and close the first outlet after the second sensor confirms that the fish have left.

[0037] Step S108: Open the second outlet and release non-target fish. After the second sensor confirms the fish have left, close the second outlet.

[0038] Step S109: The system resets the entry state and prepares to process the next fish.

[0039] This process realizes automatic identification, intelligent diversion and continuous fishing control of fish, improving efficiency and ecological protection.

[0040] Figure 4 This is a block diagram of the structure of the intelligent fishing system of the present invention, which includes the following modules:

[0041] Fish identification box 1: It is the main channel structure for fish to enter the system, and it is equipped with:

[0042] Entrance control device 11;

[0043] A first outlet 12 and a second outlet 13;

[0044] The entrance and exit are controlled by electronic control mechanism;

[0045] Equipped with a first sensor 14 and a second sensor 15 to monitor the entry and exit status of fish.

[0046] Image acquisition unit 2: includes a binocular camera 21, an auxiliary lighting unit 22 and an image signal processing module 23, which realizes high-precision, multi-angle shooting and 3D modeling of the fish body.

[0047] AI Recognition System Module 3: This is the core analysis unit of the system and mainly includes:

[0048] Image recognition submodule 31: fish body shape recognition based on CNN algorithm;

[0049] Real-time object detection submodule 32: uses YOLO or Faster R-CNN to locate and classify fish;

[0050] Data comparison and decision module 33: compares the recognition results with the target database and outputs control instructions;

[0051] Learning Optimization Module 34: Supports transfer learning and continuous training.

[0052] Control execution module 4: receives the control signal output by the recognition system, controls the opening and closing of the first outlet, the second outlet and the entrance, and realizes intelligent guidance of the fish path.

[0053] Data recording and communication module 5: used to store recognition results, images and statistical information, and realize data synchronization and remote monitoring through wireless communication module.

[0054] The fish identification box 1 adopts a streamlined rectangular shell and is arranged coaxially with the direction of water flow. The inlet control device 11 is fixed to the front end of the shell and adopts an electric gate + elastic sealing ring to cooperate with the "convex and concave lock" on the inner wall of the box to achieve reversible locking; the first outlet 12 and the second outlet 13 are respectively located on opposite sides of the rear section of the box, with an angle of about 30° between them, so as to form a diversion without changing the main water flow axis. The first sensor 14 is embedded in the inner edge of the inlet gate to capture the instantaneous impedance change of the fish body breaking the surface flow; the second sensor 15 is coaxially arranged with the center line of the first outlet, and records in real time whether the fish body has completely left the box through a two-way light curtain. All sensors are fixed in the "dovetail groove" of the box wall through flexible flat cables, and finally converge into the waterproof junction box on the top to achieve modular and quick assembly and disassembly.

[0055] The image acquisition unit 2 is directly "hanged" on the V-shaped guide rail above the identification box via a three-claw quick-release bracket. Two binocular cameras 21 are aligned with the center of the identification area in a "symmetrical convergence" manner with a 15° downward tilt and a baseline of 120mm, ensuring that the fish is within the optimal parallax range at both ends of the longitudinal axis. Auxiliary lighting units 22 are distributed in a ring around the outer edge of the camera, and the inner side of the lampshade is covered with a diffuse reflection coating. Combined with the real-time exposure adjustment algorithm of the image signal processing module 23, they can maintain a signal-to-noise ratio of ≥70dB in turbid, low-light, or backlit environments, providing reliable raw data for subsequent 3D modeling.

[0056] The AI ​​recognition system module 3 is directly connected to the image acquisition unit via Gigabit Ethernet. The image recognition submodule 31 performs primary feature extraction on the GPU. The real-time target detection submodule 32 then outputs the fish bounding box and species probability vector at 30fps. The decision module 33 compares this vector with the local SQLite target database using a hash index. Based on the "species-length-fishing policy" ternary rule, it generates the "exit ID, opening and closing sequence" instruction, keeping hard real-time latency under 200ms. The learning optimization module 34 caches high-confidence misclassified samples and automatically triggers transfer learning during offline periods, flashing incremental weight files to the recognition submodule over-the-air (OTA) for continuous self-evolution.

[0057] The control execution module 4 is connected to the AI ​​module via an RS-485 industrial bus. Its internal three-degree-of-freedom stepper motor driver board precisely controls the gate angles at the entrance, first exit, and second exit. When a decision is issued, the module first checks the status of the second sensor to prevent concurrent network congestion. It then completes the diversion process in the following order: "Entrance closed → Target exit opened → Non-target exit maintained → Entrance reopened." The data recording and communication module 5 is connected to the local router via a CAN-to-Wi-Fi gateway. In addition to storing images, metadata, and gate logs on industrial-grade SD storage, it also pushes them to a cloud-based monitoring platform in real time using the MQTT protocol. This allows remote users to simultaneously view catch statistics, equipment health, and environmental parameters. It also supports one-click policy and firmware updates, creating a closed-loop, unmanned intelligent fishing operation process.

[0058] The intelligent fishing system uses binocular camera technology to estimate the length of fish, specifically including:

[0059] Use dual cameras to obtain 3D images of fish;

[0060] The depth information and geometric relationship are extracted through stereo triangulation to calculate the length of the fish body;

[0061] The system can adapt to different fish species and morphologies to ensure measurement accuracy;

[0062] It provides a user-friendly interface to facilitate field application by fishery managers, researchers and ecologists, allowing them to obtain data in real time and support fish population management.

[0063] The intelligent fishing system includes an efficient and accurate AI species recognition system that adapts to different ecosystems (rivers, lakes, and ocean environments). It is based on a fish image dataset and trained with a deep learning model, and provides a user interface for researchers, environmental volunteers, and citizen scientists. The recognition steps include:

[0064] (1) Use convolutional neural networks (CNNs) for image recognition to identify fish shapes, colors, and patterns;

[0065] (2) Develop a text recognition model based on taxonomic descriptions;

[0066] (3) Use transfer learning and reinforcement learning to continuously optimize model performance;

[0067] (4) Advanced algorithms such as YOLO or Faster R-CNN are used to achieve real-time positioning and classification of fish in video streams, and accurate identification can be achieved even in complex fish environments, which is suitable for scientific research and commercial fishing scenarios.

[0068] The system will be field-tested in multiple waters of the Yangtze River or Mississippi River basins, and key indicators such as identification accuracy, invasive species detection rate, and response speed will be verified through comparison with experts to ensure its reliability and adaptability in fisheries management.

[0069] The Smart Fishing System (SFS) is a revolutionary fisheries solution that combines passive fishing gear, fish identification boxes, and artificial intelligence (AI) control equipment. It has profound implications for research, monitoring, and commercial fishing. Unlike labor-intensive and costly active fishing methods, the system is a low-labor, highly efficient, and cost-effective tool suitable for various rivers and lakes in the Yangtze River and Mississippi River basins. By integrating AI identification with fishery management models, the SFS system offers significant advantages for researchers, resource managers, and fishery practitioners.

[0070] The intelligent fishing system provided by this invention is based on artificial intelligence recognition, sensing technology and automatic control technology, and has broad application prospects and product transformation space. Its specific application areas and related products are as follows: 1. Application areas:

[0071] (1) Freshwater fishery resource management: Applicable to inland natural water bodies such as rivers and lakes, used for scientific fishing, ecological monitoring and commercial fishing activities, especially suitable for deployment in the Yangtze River Basin and other ecologically sensitive areas.

[0072] (2) Control of invasive alien species: It can be deployed in waters such as the Mississippi River in North America to efficiently identify and remove invasive species such as China's four major carps, thereby protecting local ecosystems.

[0073] (3) Fishery law enforcement and resource supervision: As a technical tool for fishery departments to monitor the compliance of fishing activities, it assists in real-time recording and automatic identification of non-target fish and illegal fishing activities.

[0074] (4) Aquatic ecological research and species surveys: used by scientific research institutions to conduct long-term, systematic data collection on fish population structure, species diversity, reproductive status, etc., to improve the efficiency of ecological research.

[0075] (5) Smart fishery industry upgrade: As part of modern intelligent fishery equipment, it can work in conjunction with the smart fishery platform, data center, and supervision system to promote the digital transformation of the industry.

[0076] 2. Related product forms:

[0077] Embedded AI recognition module: can be integrated into existing fishing gear, such as nets, trap boxes, gillnets, etc., for intelligent upgrade.

[0078] Movable intelligent identification and fishing device: equipped with cameras, sensors, AI chips and automatic switch structure.

[0079] Smart fishery terminal equipment: Integrates monitoring, fishing, identification, and data collection, suitable for use by scientific research institutions and government regulatory departments.

[0080] Ecological monitoring and analysis system platform: upload front-end fishing data to the platform to realize the construction and visual monitoring of fish resource database.

[0081] The intelligent fishing system of the present invention has demonstrated significant technical advantages and practical effects during experiments and applications. The relevant evidence includes the following aspects:

[0082] 1. Technical evidence of efficiency improvement:

[0083] Compared with traditional fishing methods, this system can be operated by a single person to complete identification, screening and data recording, which significantly reduces labor costs and operation time;

[0084] The intelligent identification and control system enables instant fish classification and diversion, avoiding manual repetitive operations and misjudgments.

[0085] 2. Technical evidence of accuracy and scientificity:

[0086] Using binocular stereo vision technology, combined with CNN and YOLO algorithms, the system can accurately identify fish species, size, shape, color, and other characteristics. Smartrawl has a false catch rate of 30-50%, while Deep Vision has a false catch rate of only 2%. These are applied to active fishing gear. This invention, when applied to passive fishing gear, will have an even lower false catch rate.

[0087] The AI ​​system has learning and adaptive capabilities, and is continuously optimized through reinforcement learning and transfer learning, maintaining high recognition consistency and stability even when deployed across ecosystems.

[0088] 3. Evidence of ecological and resource protection effects:

[0089] By precisely releasing non-target species and substandard individuals, we can achieve the goal of "zero bycatch," providing effective technical support for the protection and recovery of endangered fish populations.

[0090] The system can continuously monitor species activities in specific areas, accumulate and analyze long-term data on the ecological status of water bodies, and help to warn of ecological changes;

[0091] 4. Evidence of transformational potential and commercial value:

[0092] It can be adapted to existing passive fishing gear (such as nets), eliminating the need to rebuild the fishing gear system, and has a low threshold for deployment and promotion;

[0093] Its data value and ease of application are widely recognized by fishermen and scientific research institutions, and it has a foundation for large-scale promotion;

[0094] It has good docking capabilities with existing domestic monitoring systems (such as the Yangtze River Fishery Resources Protection Monitoring Network), forming an industrial cascade integration platform.

[0095] It should be noted that the embodiments of the present invention can be implemented by hardware, software, or a combination of software and hardware. The hardware portion can be implemented using dedicated logic; the software portion can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or dedicated design hardware. Those skilled in the art will appreciate that the above-mentioned devices and methods can be implemented using computer-executable instructions and / or contained in processor control code, for example, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The devices and modules of the present invention can be implemented by hardware circuits such as very large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., can also be implemented by software executed by various types of processors, or can be implemented by a combination of the above-mentioned hardware circuits and software, such as firmware.

[0096] The above description is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by any technician familiar with this technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered by the scope of protection of the present invention.

Claims

1. An intelligent fishing system, characterized in that: It includes a fishing box, an entrance door, a first exit door, a second exit door, a sensor assembly, an imaging assembly, a control processor, and a guiding mechanism; The sensor assembly detects the entry of a fish into the catch box and sends an entry signal to the control processor; The imaging component images the fish body and sends the image data to the control processor; The control processor performs fish body feature recognition on the image data. If the recognition result is a target fish, the control processor controls the entrance door to close, opens the first exit door, and drives the guiding mechanism to guide the target fish to the collection net. If the recognition result is a non-target fish, the control processor controls the entrance door to close, opens the second exit door, and releases the non-target fish. When the sensor assembly detects that the fish has left the corresponding exit, the control processor closes the exit door and reopens the entrance door.

2. The intelligent fishing system according to claim 1, characterized in that: The guiding mechanism includes a diversion channel and a light stimulation device; the control processor starts the light stimulation device when the first exit door is opened and cooperates with the diversion channel to guide the target fish into the collection net, and stops the light stimulation device when the second exit door is opened to reduce the stress of non-target fish.

3. The intelligent fishing system according to claim 1, characterized in that: The sensor assembly includes a first distance sensor and a second distance sensor; the first distance sensor is used to detect the fish entering the fishing box and trigger the entrance door to close, and the second distance sensor is used to detect the fish passing through the exit and trigger the corresponding exit door to close.

4. The intelligent fishing system according to claim 1, characterized in that: The imaging component is a binocular imaging unit. The control processor calculates the depth information of the fish body based on the stereo triangulation algorithm and combines it with the convolutional neural network model to identify the fish species, gender and growth stage.

5. The intelligent fishing system according to claim 1, characterized in that: The control processor sets a continuous learning module, which adopts a transfer learning strategy to use new sample data collected on site to update the recognition model to improve the recognition accuracy.

6. The intelligent fishing system according to claim 1, characterized in that: The system is equipped with an environmental adaptation module, which automatically adjusts the exposure parameters of the imaging component and the output power of the light stimulation device according to the turbidity of the water and the light intensity to maintain image quality and reduce fish stress.

7. A smart fishing method, characterized in that: The following steps are involved: Step 1: Detect fish entering the fishing box and close the entrance door; Step 2: Collect fish body images and perform feature recognition; Step 3: judge the identification result. If it is a target fish, open the first exit door and start the guiding mechanism to guide the collection net. If it is a non-target fish, open the second exit door to release it. Step 4: When the exit sensor detects that the fish has left, it closes the corresponding exit door and reopens the entrance door.

8. The intelligent fishing method according to claim 7, characterized in that: In the second step, a convolutional neural network is used to classify the fish shape and texture features, and the binocular depth information is combined to calculate the estimated length and weight of the fish.

9. A computer-readable storage medium, characterized in that Instructions are stored thereon, and when the instructions are executed on a processor, the processor is caused to execute the smart fishing method according to any one of claim 7 or claim 8.

10. A fish sorting mechanism, characterized in that: It includes an openable and closable entrance door, a first exit door, a second exit door, a first distance sensor, a second distance sensor and a control unit; the control unit controls the closing of the entrance door according to the detection signal of the first distance sensor, and drives the opening and closing of the first exit door or the second exit door according to the fish body recognition result; the first distance sensor and the second distance sensor are used to cooperate to judge the passing status of the fish body to ensure that the exit door is closed in time.

Images