Underwater fish bait intake detection system based on intelligent multi-modal data fusion

Abstract

The invention provides an underwater fish bait taking detection system based on intelligent multi-modal data fusion. The technical problems that an existing fish bait taking behavior monitoring method is single in data collection, insufficient in detection accuracy, poor in real-time performance and low in sensitivity are solved. The device comprises a power supply module, and further comprises a sensing module, a data processing module and a display module which are connected with the power supply module, and the sensing module is connected with the data processing module; the sensing module is used for acquiring multi-modal data including a biological digital signal, an acceleration digital signal, a tension digital signal and image data in real time; the data processing module is used for preprocessing the multi-modal data, performing weighted fusion on the time sequence features of the biological digital signal, the acceleration digital signal, the tension digital signal and the image data through a multi-modal data fusion technology to obtain a feature fusion result, and outputting the feature fusion result; and comprehensively analyzing the feature fusion result by using a deep learning algorithm to obtain a decision result. The method can be widely applied to the technical field of underwater monitoring.

Description

Technical Field

[0001] This application belongs to the technical field of underwater monitoring. More specifically, it relates to a multi-modal data fusion underwater fish feeding detection system based on embodied intelligence. Background Art

[0002] In recent years, with the rapid development of sensor technology, image processing technology, and artificial intelligence algorithms, multi-modal data fusion technology has provided new solutions for fish behavior analysis. Currently, the research on fish behavior analysis mainly focuses on the analysis of single-modal data, such as visual-based fish feeding intensity recognition, sound-based fish feeding intensity recognition, and water-quality-based fish feeding intensity recognition. However, the single-modal analysis method has limitations and cannot comprehensively capture the multi-dimensional information of fish behavior.

[0003] Zhu Wentao from Yangzhou University proposed a multi-modal fusion method for monitoring fish feeding intensity in the paper "Research and Implementation of Fish Feeding Intensity Classification Based on Multi-modal Fusion". By fusing the information of three modalities: water quality, image, and sound, high-precision recognition of fish feeding intensity was achieved. However, this method mainly focuses on the fusion of water quality, image, and sound modalities, and ignores the importance of acceleration and force signals in fish behavior analysis, resulting in single data collection and insufficient detection accuracy in practical applications. Moreover, since this method requires separate preprocessing and extraction of each multi-modal data, the real-time performance of the monitoring method is poor and the sensitivity is low. Therefore, there is an urgent need to provide a multi-modal data fusion underwater fish feeding detection system based on embodied intelligence to solve the above problems. Summary of the Invention

[0004] The purpose of the embodiments of this application is to provide a multi-modal data fusion underwater fish feeding detection system based on embodied intelligence to solve the technical problems of single data collection, insufficient detection accuracy, poor real-time performance, and low sensitivity in the existing fish feeding behavior monitoring methods.

[0005] To achieve the above purpose, the embodiments of this application provide a multi-modal data fusion underwater fish feeding detection system based on embodied intelligence, including a power supply module, and further including a sensing module, a data processing module, and a display module connected to the power supply module, where the sensing module is connected to the data processing module; The sensing module is used to obtain multi-modal data in real time, including biological digital signals, acceleration digital signals, tensile digital signals, and image data; The data processing module is used to preprocess the multi-modal data, weight and fuse the time series features of biological digital signals, acceleration digital signals, tensile digital signals, and image data to obtain a feature fusion result, and comprehensively analyze the feature fusion result using a deep learning algorithm to obtain a decision result of fish feeding; Formula for weighted fusion: ; In the formula, is the result of feature fusion, is the weight of the biological digital signal, is the time series feature of the biological digital signal, is the weight of the acceleration digital signal, is the time series feature of the acceleration digital signal, is the weight of the tensile digital signal, is the time series feature of the tensile digital signal, is the weight of the image data, is the time series feature of the image data.

[0006] Preferably, the data processing module includes an understanding module and a decision-making module; The understanding module adopts a dual-channel deep learning neural network model, including a first channel and a second channel; The first channel is used to denoise, enhance the image and detect the target for the image data during the fish feeding process, obtain the features of the image data, and after unifying the dimensions, perform standardization to extract the time series features of the two-dimensional image data during the fish feeding process; The second channel is used to preprocess and extract features from the biological digital signal, acceleration digital signal and tensile digital signal. After unifying the dimensions of the features, perform standardization processing to obtain the time series features of the one-dimensional biological digital signal, acceleration digital signal and tensile digital signal.

[0007] Preferably, the decision-making module is used to perform weighted fusion on the time series features of the image data during the fish feeding process with the time series features of the biological digital signal, acceleration digital signal and tensile digital signal to obtain the feature fusion result, and comprehensively analyze the feature fusion result using a deep learning algorithm to obtain the decision result of fish feeding.

[0008] Preferably, the perception module includes a bioelectric sensor module, an acceleration sensor module, a tensile sensor module and an image sensor module; The bioelectric sensor module is used to detect the biological digital signal generated by the fish body; The acceleration sensor module is used to monitor the acceleration change of the fishing bait in real time during the fish feeding process to obtain the acceleration digital signal; The tensile sensor module is used to detect the minute tensile change generated by the fish biting the hook in real time to obtain the tensile digital signal; The image sensor module is used to synchronously acquire the image data during the fish feeding process.

[0009] Preferably, the decision-making module can also continuously adjust the weights of weighted fusion according to historical decision results and the actual detection success rate.

[0010] Preferably, the decision result includes whether the fish is feeding, and the species and weight of the fish.

[0011] Preferably, the display module is connected to the data processing module for real-time feedback of the decision result.

[0012] Preferably, the bioelectric sensor module includes an electrode array and a signal acquisition circuit; The electrode array is used to detect the bioelectric signals generated by the fish body during feeding; The signal acquisition circuit is used to amplify and filter the bioelectric signals and convert them in the first analog-to-digital converter to obtain bio-digital signals; The bio-digital signals can reflect the bioelectric activities of the fish body.

[0013] Preferably, the acceleration sensor module includes an accelerometer, a conditioning circuit, and a second analog-to-digital converter; The accelerometer is used to obtain the acceleration analog signal of the fishing bait in three-dimensional space; The conditioning circuit is used to amplify and filter the acceleration analog signal; The second analog-to-digital converter is used to convert the amplified and filtered acceleration analog signal into an acceleration digital signal.

[0014] Preferably, the tension sensor module includes a strain gauge, a Wheatstone bridge, and a third analog-to-digital converter; The strain gauge is used to detect the tiny tension change of the fishing line when the fish feeds and convert it into a resistance change; The Wheatstone bridge is used to convert the resistance change into a tension analog signal and cooperate with an amplifier circuit to amplify the tension analog signal; The third analog-to-digital converter is used to convert the tension analog signal into a tension digital signal.

[0015] The beneficial effects of this application are as follows: This application provides a multi-modal data fusion underwater fish feeding detection system based on embodied intelligence. Based on the concept of embodied intelligence, the perception module can obtain multi-modal data in real time, thus overcoming the limitations of single-modal data. By using multi-modal data fusion technology, the sensitivity and accuracy of fish feeding behavior detection are significantly improved. Among them, the data processing module integrates multi-modal data fusion technology and deep learning algorithms to analyze multi-modal data in real time and achieve real-time detection of fish feeding behavior. Specifically, the time series features of acceleration digital signals, tensile digital signals, and the time series features of image data are weighted and fused. During the fusion process, a weighting mechanism is used to assign corresponding weights to different modal data, thereby improving the recognition accuracy of fish feeding behavior and achieving accurate detection of fish behavior in complex underwater environments. In this process, relying on deep learning algorithms to achieve accurate discrimination and comprehensive analysis of fish information, and finally obtaining the decision result of fish feeding.

[0016] In summary, this application not only overcomes the limitations of single-modal data, significantly improves the sensitivity and accuracy of fish feeding behavior detection, but also provides real-time feedback on fish feeding information through the data processing module, with better real-time performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of this application, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the following drawings 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.

[0018] Figure 1 It is a schematic structural diagram of a multi-modal data fusion underwater fish feeding detection system based on embodied intelligence provided by an embodiment of this application; Figure 2 It is a schematic diagram of the connection of each module provided by an embodiment of this application; Figure 3 It is a flowchart of data processing by the data processing module provided by an embodiment of this application.

[0019] Drawings: 1. Bioelectric sensor module; 2. Acceleration sensor module; 3. Tensile sensor module; 4. Image acquisition module; 5. Fishing line; 6. Fishing rod; 7. Data processing module; 8. Display module; 9. Power supply module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] In order to make the technical problems, technical solutions, and beneficial effects to be solved by this application more clearly understood, the following further details this application in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not used to limit this application.

[0021] The present invention proposes a multi-modal data fusion underwater fish bait-taking detection system based on embodied intelligence. Through the multi-modal perception, understanding, decision-making, and control capabilities of embodied intelligence, using a bioelectric sensor module 1, an acceleration sensor module 2, a tensile force sensor module 3, an image acquisition module 4, and a deep learning algorithm, it realizes accurate detection, species identification, and weight estimation of fish bait-taking behaviors, and provides real-time feedback through a data processing module, providing a brand-new technical solution for aquaculture, ecological monitoring, and intelligent fishery management.

[0022] Please refer to Figure 1 , Figure 2 , which is a multi-modal data fusion underwater fish bait-taking detection system based on embodied intelligence provided by an embodiment of this application. It includes a power supply module 9, and also includes a perception module, a data processing module 7, and a display module 8 connected to the power supply module 9. The perception module is connected to the digital processing module 7.

[0023] The perception module is used to obtain multi-modal data in real time, including bio-digital signals, acceleration digital signals, tensile force digital signals, and image data, aiming to comprehensively reflect various characteristics and states of fish during the bait-taking process.

[0024] Specifically, the perception module includes a hook-shaped bioelectric sensor module 1, an acceleration sensor module 2 installed on the upper side of the hook-shaped bioelectric sensor module 1, a tensile force sensor module 3, and an image acquisition module 4. The tensile force sensor module 3 and the image acquisition module 4 are encapsulated in the same waterproof housing and are connected to the data processing module 7 installed at the end of the fishing rod 6 through a fishing line 5 with wires clamped.

[0025] Among them, the bioelectric sensor module 1 is used to detect the bioelectric signals generated by the fish body. The bioelectric sensor module 1 adopts a high-sensitivity electrode array and a high-precision signal acquisition circuit. Based on the principle of capacitive coupling, the electrode array can detect the weak bioelectric signals generated by the fish body during bait-taking. These bioelectric signals are generated during the muscle activity and nerve conduction of the fish. The bioelectric sensor module 1 is installed at the fishhook and is connected to the data processing module 7 through the wires clamped in the fishing line 5. The signal acquisition circuit includes a preamplifier, a filter, and a first analog-to-digital converter, which are used to amplify and filter the bioelectric signals and convert them in the first analog-to-digital converter to obtain bio-digital signals. Among them, the amplifier can amplify the weak signals to an appropriate amplitude, and at the same time remove environmental noise through the filter, so that the finally output bio-digital signals can clearly reflect the bioelectric activities of the fish body.

[0026] The acceleration sensor module 2 is used to monitor the acceleration change of the bait in real time during the process of fish taking bait and obtain the acceleration digital signal. The acceleration sensor module 2 adopts an accelerometer based on micro-electromechanical system (MEMS) technology, which can accurately measure the acceleration change of the bait in three-dimensional space during the process of fish taking bait, and is composed of a signal conditioning circuit and a second analog-to-digital converter. The module is connected to the data processing module 7 through the wire clamped in the fishing line 5, and is used to detect the acceleration change of the bait in real time during the process of fish taking bait in the water, and convert the acceleration analog signal into an acceleration digital signal after conditioning, and transmit it to the data processing module 7 to assist in analyzing the fish's bait-taking behavior.

[0027] The tension sensor module 3 is used to detect the tiny force changes generated by the fish biting the hook in real time and obtain the tension digital signal. The tension sensor module 3 is composed of a strain sensor, a Wheatstone bridge and a third analog-to-digital converter. It is equipped with a temperature compensation circuit, which adjusts the parameters of the Wheatstone bridge according to the feedback of the water temperature sensor to offset the temperature dependence of the strain sensor, thereby reducing the impact of temperature changes on the measurement results. The module is connected to the data processing module 7 through the wire clamped in the fishing line 5, and is used to detect the tiny force signal generated by the fish biting the hook in real time, and convert the tension analog signal into a tension digital signal and transmit it to the data processing module 7. Specifically, the strain sensor is connected to the fishhook-shaped bioelectric sensor module 1 through a shorter fishing line. Compared with being installed on the fishing rod 6, the distance is closer and the measurement is more accurate. It can detect the tiny tension changes generated by the fishing line when the fish takes the bait and convert it into resistance changes. The Wheatstone bridge converts the resistance change output by the strain sensor into a tension analog signal, and cooperates with the amplifier circuit to perform preliminary amplification processing on the tension analog signal to improve the stability and sensitivity of the tension analog signal. The third analog-to-digital converter converts the amplified tension analog signal into a tension digital signal, and transmits it to the digital processing module 7 via the fishing line 5 clamped with the conductor.

[0028] The image acquisition module 4 synchronously acquires the image data of the process of fish taking bait through the underwater camera. The optical system of the camera in the image acquisition module 4 is optimized, and the underwater optical lens is used to reduce the refraction and scattering of underwater light, improve the clarity of the image data, and is packaged in a waterproof housing together with the tension sensor module 3, and is used to synchronously acquire the image data of the process of fish taking bait, and transmit the image data to the data processing module 7 through the wire clamped in the fishing line 5.

[0029] The data processing module 7 includes an understanding module and a decision-making module, which are used to receive the multimodal data collected by the perception module, preprocess the multimodal data including noise reduction, feature extraction, and feature enhancement, fuse the time series features of the bio-digital signal, acceleration digital signal, and tensile digital signal with the spatial features of the image data through multimodal data fusion technology, and comprehensively analyze the data using deep learning algorithms to accurately judge whether the fish is feeding and obtain the decision result of the fish feeding.

[0030] Please refer to Figure 3 , which is the flowchart of the data processing module 7 of this application for processing multimodal data. Specifically, the data processing module 7 processes and analyzes the data transmitted by the bioelectric sensor module 1, acceleration sensor module 2, tensile sensor module 3, and image acquisition module 4, denoises and filters the bio-digital signal, acceleration digital signal, and tensile digital signal converted by the first analog-to-digital converter, second analog-to-digital converter, and third analog-to-digital converter, then extracts features through a time series encoder, and then standardizes the features after unifying the dimensions to obtain the time series features of the bio-digital signal, acceleration digital signal, and tensile digital signal. Preprocess the image data collected by the camera, including noise reduction, image enhancement, and target detection, etc., extract the features of the image data during the fish feeding process, and standardize and extract the time series features of the image data during the fish feeding process after unifying the dimensions. Weightedly fuse the time series features of the bio-digital signal, acceleration digital signal, tensile digital signal, and image data, and adopt a feature weighting mechanism to assign different weights to the two-modal data to obtain the feature fusion result of the fish, so as to improve the detection accuracy; use an object detection model based on the YOLO-Nano network to identify the type of fish, and at the same time use a recurrent neural network (GRU) to analyze the dynamic features of the fish feeding behavior, and comprehensively judge the fish feeding behavior, the type and weight of the fish through a two-channel deep neural network to obtain the decision result of the fish feeding.

[0031] Among them, the understanding module uses multimodal data fusion technology and deep learning algorithms to deeply analyze and intelligently understand the multimodal data collected by the perception module, and obtain the time series features of the image data and the time series features of the bio-digital signal, acceleration digital signal, and tensile digital signal. This method can not only extract the independent features of different-modal data, but also mine the correlation information between different-modal data.

[0032] Specifically, the understanding module adopts a dual-channel deep neural network structure. Among them, the first channel is used to process the image data during the fish feeding process, detect and classify the target fish through the object detection model, and extract the time series features of the two-dimensional fish feeding behavior image data; the second channel is used to process the biological digital signal, acceleration digital signal, and pulling force digital signal, and extract the dynamic features of the fish feeding behavior through the gated recurrent unit (GRU). Among them, the dynamic features of the fish feeding behavior refer to the time series features of the one-dimensional biological digital signal, acceleration digital signal, and pulling force digital signal.

[0033] In an optional embodiment, an object detection model based on the YOLO-Nano network and a dual-channel deep neural network model combined with time series feature analysis are adopted. The first channel is used to process the image data during the fish feeding process, and the YOLO-Nano network is used to detect and classify the target fish to obtain the time series features of the image data. . Among them, the YOLO-Nano network can quickly and accurately locate the fish in the image. The second channel is used to process the biological digital signal, acceleration digital signal, and pulling force digital signal, and the gated recurrent unit (GRU) is used to extract the dynamic features of the fish feeding behavior, that is, the time series features of the biological digital signal, acceleration digital signal, and pulling force digital signal.

[0034] Specifically, after the biological digital signal, acceleration digital signal, and pulling force digital signal are preprocessed, features are extracted through the time series encoder respectively to obtain the features of the biological digital signal , the features of the acceleration digital signal and the features of the pulling force digital signal , where is the feature dimension, . The gated recurrent unit can well process time series data and capture the variation law of the signal over time.

[0035] Input the features of the biological digital signal, acceleration digital signal, pulling force digital signal, and image data into the fully connected layer of the gated recurrent unit to map the features of different dimensions to a unified dimension. The formula is: ; In the formula, h i is the new feature of the unified dimension after transformation, h i raw are the features of each signal, including the biological digital signal , the acceleration digital signal , the pulling force digital signal and the image data , is the projection matrix, is the bias term.

[0036] The formula for standardizing the new features with unified dimensions after transformation to eliminate the dimension difference is as follows: ; In the formula, are the mean and standard deviation of each signal feature respectively, is the feature of each signal after standardization, including the time series features of biological digital signals, acceleration digital signals, tensile digital signals and image data, is a very small constant used for numerical stability to prevent the denominator from being zero.

[0037] Based on the features of each signal after standardization obtained by the understanding module, the decision-making module makes accurate decisions using intelligent algorithms. This decision-making process comprehensively considers various factors and the collaborative information of different modal data to ensure the reliability and accuracy of the decision results.

[0038] Based on the features of each signal after standardization obtained from the understanding result and the weights of the feature weighting mechanism, the final decision is achieved through the feature fusion module between channels. The time series features of the image data during the fish feeding process obtained from the first channel and the time series features of biological digital signals, acceleration digital signals and tensile digital signals obtained from the second channel are weighted and fused to obtain the feature fusion result, and whether the fish is feeding is comprehensively judged according to the feature fusion result to improve the accuracy and reliability of detection.

[0039] Specifically, according to the features of each signal after standardization , weights are generated through the feature weighting mechanism, and the formulas include: ; ; In the formula, performs a linear transformation on the features of each signal after standardization, () performs a non-linear activation, maps the transformed features to scalar scores , is the bias vector, is the weight matrix, is the weight vector, are the weights of biological digital signals, acceleration digital signals, tensile digital signals and image data, satisfying , is the weight of biological digital signals, is the weight of acceleration digital signals, is the weight of tensile digital signals, is the weight of image data, is the scalar score of biological digital signals, is the scalar score of the acceleration digital signal, is the scalar score of the tensile digital signal, is the scalar score of the image data.

[0040] The formula for fusing the features of the biological digital signal, acceleration digital signal, tensile digital signal, and image data by weight to obtain the weighted fusion result includes: ; In the formula, is the result of feature fusion, where, integrates the information of multi-source signals, is the weight of the biological digital signal, is the time series feature of the biological digital signal, is the weight of the acceleration digital signal, is the time series feature of the acceleration digital signal, is the weight of the tensile digital signal, is the time series feature of the tensile digital signal, is the weight of the image data, is the time series feature of the image data.

[0041] The weighted decision-making process comprehensively considers the contributions of different modal data under different environmental conditions, combines the current environmental water turbidity, water flow velocity, light intensity, etc., and automatically adjusts the weights of each modal data through an intelligent algorithm to significantly improve the recognition accuracy of the bait-taking behavior and achieve accurate detection of fish behavior in various complex underwater environments. And it has the functions of self-learning and self-adaptation. According to the historical decision results of the detection system and the actual detection success rate, the reinforcement learning algorithm is used to continuously adjust the weights of the weighted fusion, so that the performance of the detection system gradually improves during long-term use.

[0042] The display module 8 includes a liquid crystal display screen and a buzzer, is connected to the data processing module 7, and is encapsulated in the same waterproof housing with the data processing module 7. This module can emit a prompt sound when the fish takes the bait, and at the same time display the type and weight of the fish, providing real-time information feedback for the user.

[0043] The power supply module 9 includes a lithium battery and a lithium battery charge and discharge management circuit, providing power support for the entire system. In order to adapt to the underwater environment, the power supply module 9 is waterproofed and wireless charging is adopted to ensure the safe and stable operation of the system underwater.

[0044] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. A professional technician can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of this application.

[0045] The above-described embodiments are only used to illustrate the technical solutions of this application, rather than to limit them; although this application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A multimodal data fusion underwater fish feeding detection system based on embodied intelligence, including a power supply module, characterized in that, It also includes a sensing module, a data processing module, and a display module connected to the power supply module, and the sensing module is connected to the data processing module; The sensing module is used to obtain multi-modal data in real time, including bio-digital signals, acceleration digital signals, tensile digital signals, and image data; The data processing module is used to preprocess the multi-modal data, perform weighted fusion on the time series features of the bio-digital signal, the acceleration digital signal, the tensile digital signal, and the image data to obtain a feature fusion result, and comprehensively analyze the feature fusion result using a deep learning algorithm to obtain a decision result of fish feeding; The formula for the weighted fusion: ； Wherein, is the feature fusion result, is the weight of the biological digital signal, is the time series feature of the biological digital signal, is the weight of the acceleration digital signal, is the time series feature of the acceleration digital signal, is the weight of the tensile digital signal, is the time series feature of the tensile digital signal, is the weight of the image data, is the time series feature of the image data.

2. The multimodal data fusion underwater fish feeding detection system based on embodied intelligence according to claim 1, characterized in that, The data processing module includes an understanding module and a decision module; The understanding module adopts a dual-channel deep learning neural network model, including a first channel and a second channel; The first channel is used to perform noise reduction, image enhancement, and target detection on the image data during the fish feeding process to obtain the features of the image data, and after unifying the dimensions, perform standardization to extract the time series features of the two-dimensional image data during the fish feeding process; The second channel is used to preprocess and extract features from the bio-digital signal, the acceleration digital signal, and the tensile digital signal, and after unifying the dimensions of the features, perform standardization processing to obtain the time series features of the one-dimensional bio-digital signal, acceleration digital signal, and tensile digital signal.

3. The multi-modal data fusion underwater fish feeding detection system based on embodied intelligence according to claim 2, wherein The decision module is used to perform weighted fusion on the time series features of the image data during the fish feeding process with the time series features of the bio-digital signal, acceleration digital signal, and tensile digital signal to obtain the feature fusion result, and comprehensively analyze the feature fusion result using a deep learning algorithm to obtain a decision result of fish feeding.

4. The multimodal data fusion underwater fish feeding detection system based on embodied intelligence according to claim 3, characterized in that, The sensing module includes a bioelectric sensor module, an acceleration sensor module, a tensile sensor module, and an image sensor module; The bioelectric sensor module is used to detect the bio-digital signal generated by the fish body; The acceleration sensor module is used to monitor the acceleration change of the bait in real time during the fish feeding process to obtain the acceleration digital signal; The tensile sensor module is used to detect the minute tensile change generated by the fish biting the hook in real time to obtain the tensile digital signal; The image sensor module is used to synchronously obtain the image data during the fish feeding process.

5. The multimodal data fusion underwater fish feeding detection system based on embodied intelligence according to claim 4, wherein, The decision module can also continuously adjust the weight of the weighted fusion according to the historical decision result and the actual detection success rate.

6. The multimodal data fusion underwater fish bait-taking detection system based on embodied intelligence according to claim 5, characterized in that, The decision result includes whether the fish is feeding and the species and weight of the fish.

7. The multimodal data fusion underwater fish feeding detection system based on embodied intelligence according to claim 6, characterized in that The display module is connected to the data processing module and is used to feedback the decision result in real time.

8. The multimodal data fusion underwater fish bait intake detection system based on embodied intelligence according to claim 7, wherein, The bioelectric sensor module includes an electrode array and a signal acquisition circuit; The electrode array is used to detect the bioelectric signal generated by the fish body during feeding; The signal acquisition circuit is used to amplify and filter the bioelectric signal and perform conversion in a first analog-to-digital converter to obtain a bio-digital signal; The biological digital signal can reflect the bioelectrical activity of the fish body.

9. The multi-modal data fusion underwater fish bait-taking detection system based on embodied intelligence according to claim 8, characterized in that, The acceleration sensor module includes an accelerometer, a conditioning circuit, and a second analog-to-digital converter; The accelerometer is used to obtain the acceleration analog signal of the fishing bait in three-dimensional space; The conditioning circuit is used to amplify and filter the acceleration analog signal; The second analog-to-digital converter is used to convert the amplified and filtered acceleration analog signal into an acceleration digital signal.

10. A multimodal data fusion underwater fish feeding detection system based on embodied intelligence as claimed in claim 9, characterized in that, The tensile force sensor module includes a strain sensor, a Wheatstone bridge, and a third analog-to-digital converter; The strain sensor is used to detect the minute tensile force change generated by the fishing line when the fish takes the bait and convert it into a resistance change; The Wheatstone bridge is used to convert the resistance change into a tensile force analog signal and amplify the tensile force analog signal in cooperation with an amplifier circuit; The third analog-to-digital converter is used to convert the tensile force analog signal into the tensile force digital signal.

Images