A fish school detection method, device, equipment and medium for fishing

Abstract

The application discloses a fish school detection method, device and equipment for fishing and a medium, and relates to the technical field of fishing auxiliary equipment. The method comprises the following steps: acquiring sensor signals to be detected by fish school through multi-dimensional processing of underwater original data; acquiring image data to be detected by fish school through a signal conditioning method based on the sensor signals to be detected by fish school; and acquiring a fish school detection result through fish school detection based on the image data to be detected by fish school through an image processing algorithm. The application not only realizes efficient integration and accurate conditioning of multi-source data, greatly improves the quality and effectiveness of image data, but also significantly improves the accuracy of fish school detection, the effective target recognition rate and the trajectory continuity through multi-stage screening and an accurate tracking mechanism, reduces the false detection rate and the missed detection rate, has real-time and adaptability of data processing, and improves the fishing experience and fishing efficiency.

Description

Technical Field

[0001] This invention relates to the field of fishing auxiliary equipment technology, and in particular to a method, device, equipment and medium for detecting fish schools in fishing. Background Technology

[0002] Fishing, as a recreational activity, is enjoyed by many enthusiasts. During fishing, the location of fish is a key factor affecting the success rate. Traditional fishing methods rely mainly on the angler's experience to determine the location of fish, which is less accurate and easily leads to missing the best fishing opportunity.

[0003] While some existing fish-finding devices attempt to acquire information about underwater fish schools, they suffer from several shortcomings: some devices employ a single-dimensional detection method, resulting in inaccurate fish behavior assessments due to insufficient data dimensions; others lack effective data processing mechanisms, leading to inaccurate fish detection results and failing to meet the actual needs of anglers; data transmission is not optimized for fishing scenarios, resulting in transmission delays, data loss, or excessive redundancy; and fish school recognition algorithms have poor adaptability, struggling to cope with complex environments such as insufficient underwater light and turbid water, resulting in low recognition accuracy and failing to provide reliable fish school information support for anglers.

[0004] Therefore, there is an urgent need for a method, device, equipment, and medium for detecting fish schools in fishing to address the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to propose a method, device, equipment, and medium for detecting fish schools in fishing, in order to solve the problems of existing fish school detection technologies for fishing, such as single data dimension, inaccurate detection results, low transmission efficiency, and insufficient recognition accuracy, and to achieve comprehensive, efficient, and accurate detection of underwater fish schools.

[0006] In a first aspect, to achieve the above objective, the present invention provides a method for detecting fish schools in fishing, comprising:

[0007] S1. Utilize raw underwater data for multi-dimensional processing to obtain sensor signals for fish detection;

[0008] S2. Based on the sensor signal of the fish school to be detected, the image data of the fish school to be detected is obtained using a signal conditioning method;

[0009] S3. Based on the image data to be detected, perform fish detection using an image processing algorithm to obtain the fish detection result.

[0010] Optionally, S1, perform multi-dimensional processing on raw underwater data to obtain sensor signals for fish detection, including:

[0011] The collected raw underwater data is integrated using a multi-dimensional data synchronization and alignment algorithm to obtain multi-dimensional underwater data;

[0012] Preliminary noise reduction is performed on the underwater multi-dimensional data to obtain the noise-reduced underwater multi-dimensional data;

[0013] The noise-reduced underwater multi-dimensional data is calibrated to obtain calibrated underwater multi-dimensional data.

[0014] The calibrated underwater multidimensional data is correlated and integrated to obtain a single-timestamp multidimensional dataset as the sensor signal for fish swarm detection.

[0015] Optionally, S2, based on the sensor signal of the fish school to be detected, an image data for the fish school to be detected is obtained using a signal conditioning method, including:

[0016] The sensor signals to be detected from the fish school are classified and analyzed to obtain multi-source signal correlation data, which includes optical sub-signals, sonar sub-signals, water depth sub-signals and water quality sub-signals.

[0017] Based on the multi-source signal correlation data, classify signal conditioning is performed to obtain the conditioned multi-source signal;

[0018] The conditioned multi-source signal correlation data is fused to obtain a composite image signal fused from the multi-signal data.

[0019] Secondary noise reduction processing is performed on the composite image signal fused from the multi-signal fusion to obtain image data for fish detection.

[0020] Optionally, based on the multi-source signal correlation data, categorized signal conditioning is performed to obtain the conditioned multi-source signal, including:

[0021] Using the multi-source signal correlation data, optical sub-signals, sonar sub-signals, water depth sub-signals, and water quality sub-signals are obtained;

[0022] Based on the optical sub-signal, a combined algorithm of Gaussian filtering and bilateral filtering is used to obtain the conditioned optical sub-signal;

[0023] Based on the conditioned optical sub-signal and water quality data, the gain is adaptively adjusted, and the gray world method and adaptive exposure compensation correction are used to obtain the optical image signal.

[0024] The sonar sub-signal is enhanced and interference frequency filtered to obtain a conditioned sonar sub-signal.

[0025] Based on the conditioned sonar sub-signal, a grayscale and pseudocolor mapping algorithm is used to obtain the sonar pseudocolor image signal;

[0026] Based on the water depth sub-signal and the water quality sub-signal, a moving average filter is used to obtain the conditioned water depth sub-signal and water quality sub-signal;

[0027] The calibration parameters are extracted from the conditioned water depth sub-signal and water quality sub-signal to obtain a fused calibration parameter set;

[0028] The optical image signal, the sonar pseudo-color image signal, and the fusion calibration parameter set are acquired as a conditioned multi-source signal.

[0029] Optionally, S3, based on the image data to be detected, fish detection is performed using an image processing algorithm to obtain fish detection results, including:

[0030] Using the image data of the fish school to be detected, obtain candidate images of fish to be detected;

[0031] The connected component labeling algorithm is used to extract features from the candidate images of fish to be detected, and the images of fish to be detected are obtained.

[0032] Fish detection is performed on the image containing fish to be detected, and the fish detection results are obtained.

[0033] Optionally, using the image data of the fish school to be detected, candidate images of fish to be detected are obtained, including:

[0034] The image data to be detected for fish swarms is subjected to noise reduction, enhancement, and adaptive parameter adjustment to obtain preprocessed image data for fish swarm detection.

[0035] Based on the preprocessed image data of the fish swarm to be detected, the sonar energy ratio and image grayscale variance are calculated to obtain preliminary screened image data;

[0036] Determine whether the initially screened image data is a fishless image. If so, obtain an adjustment instruction and execute the first operation; otherwise, obtain the initially screened image data as an initially screened fish-containing image.

[0037] Based on the pre-selected images containing fish, a pre-trained semantic segmentation model is used to obtain the mask image of the images containing fish.

[0038] Verify whether there is a valid fish area in the mask image of the fish image. If there is, obtain the preliminarily screened fish image as a fish candidate image to be detected. Otherwise, obtain the adjustment instruction and execute the first operation.

[0039] The first operation is: based on the adjustment instruction, acquire new raw underwater data and return to S1.

[0040] Optionally, fish detection is performed based on the image containing fish to be detected, and the fish detection result is obtained, including:

[0041] Based on the image of fish to be detected, a connected component labeling algorithm is used to obtain the suspected fish area;

[0042] Based on the suspected fish school region, feature extraction is performed to obtain the feature vector of the suspected fish school region.

[0043] The feature vectors of the suspected fish schools are matched with the feature template library of fishing species using normalized cross-correlation to filter out effective fish schools and then obtain a time-series sequence of images with fish.

[0044] Kalman filtering and Hungarian algorithm are applied to the time-series fish image sequence to obtain a continuous trajectory dataset of the fish target;

[0045] Based on the continuous trajectory dataset of the fish target, the fish detection results are obtained.

[0046] Secondly, to achieve the above objectives, the present invention provides a fish detection device for fishing, comprising: a fish detection module, a WIFI control module, and an intelligent analysis module;

[0047] The fish detection module is used to perform multi-dimensional processing on raw underwater data to obtain sensor signals for the fish school to be detected.

[0048] The WIFI control module is used to acquire image data of the fish school to be detected using a signal conditioning method based on the sensor signal of the fish school to be detected.

[0049] The intelligent analysis module is used to perform fish detection using image processing algorithms based on the image data of the fish to be detected, and to obtain the fish detection results.

[0050] Thirdly, to achieve the above objectives, the present invention provides an electronic device, comprising: one or more processors; and a storage device having stored one or more programs thereon, which, when executed by the one or more processors, cause the one or more processors to implement the method described in any implementation of the first aspect.

[0051] Fourthly, to achieve the above objectives, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by one or more processors, implements the method as described in any implementation of the first aspect.

[0052] Compared with the closest existing technology, the present invention has the following advantages:

[0053] This invention employs a comprehensive design for continuous multi-target tracking, encompassing multi-dimensional data synchronization and alignment, categorized signal conditioning, semantic segmentation to filter candidate images containing fish, connected component labeling and normalized cross-correlation matching to filter effective fish school regions, and the fusion of Kalman filtering and the Hungarian algorithm. Combined with the collaborative work of fish detection, WIFI control, and intelligent analysis modules, it effectively addresses the pain points of fish school detection in complex underwater environments. It not only achieves efficient integration and precise conditioning of multi-source data, significantly improving the quality and effectiveness of image data, but also, through a multi-stage filtering and precise tracking mechanism, significantly enhances the accuracy of fish school detection, the effective target recognition rate, and trajectory continuity, reducing false positives and false negatives. Furthermore, it possesses real-time and adaptable data processing capabilities, providing accurate, reliable, and efficient support for detecting the location, quantity, type, and movement characteristics of fish schools in fishing scenarios, thereby improving the fishing experience and fishing efficiency. Attached Figure Description

[0054] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0055] Figure 1 This is a flowchart of a fish school detection method for angling according to an embodiment of the present invention;

[0056] Figure 2 This is a schematic diagram of a fish detection device for angling according to an embodiment of the present invention;

[0057] Figure 3 This is a schematic diagram of the structure of the electronic device proposed in an embodiment of the present invention. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions in the embodiments of this invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this invention, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0059] The terminology used in the embodiments section of this invention is for the purpose of explaining specific embodiments of the invention only, and is not intended to limit the invention.

[0060] like Figure 1 As shown, this embodiment of the invention provides a method for detecting fish schools during fishing, including:

[0061] S1. Utilize raw underwater data for multi-dimensional processing to obtain sensor signals for fish detection;

[0062] The core of this step is to perform multi-dimensional preprocessing on raw underwater data to obtain the signals from the fish school detection sensors, thereby addressing issues such as temporal misalignment, heterogeneous formats, and noise interference in the raw underwater data from multiple sources. Using raw data collected by various types of underwater sensors as the processing object, a series of multi-dimensional processing operations are performed, including synchronous alignment and integration of multi-source data, targeted preliminary noise reduction to eliminate explicit interference, and consistency calibration to achieve format and range unification. This eliminates interference information in the raw data, unifies data standards, and establishes data associations, forming a structured single-timestamp multi-dimensional dataset. The output is a standardized fish school detection sensor signal, providing high-quality data support for subsequent signal conditioning.

[0063] S2. Based on the sensor signal of the fish school to be detected, the image data of the fish school to be detected is obtained using a signal conditioning method;

[0064] This step aims to generate image data for fish school detection based on the sensor signals from the target fish school using signal conditioning methods. This step first receives structured sensor signals, then performs signal conditioning processes such as classification and parsing, type-specific conditioning, multi-source fusion, and standardization. Targeted optimization is performed based on the characteristics of different types of sensor signals, while multi-source sensor signals are fused and integrated. Finally, the conditioned signal is transformed into image data for fish school detection that meets the subsequent detection requirements, achieving the conversion and optimization from sensor signals to image data.

[0065] S3. Based on the image data to be detected, perform fish detection using an image processing algorithm to obtain the fish detection result;

[0066] This step utilizes image processing algorithms to analyze the image data of the fish school to be detected, complete the fish school detection, and obtain the final detection result. First, image quality is optimized through image processing operations such as filtering and noise reduction, and adaptive enhancement. Then, a specific fish school detection algorithm (such as a deep learning-based semantic segmentation model) is used to separate the fish school from the background and locate suspected fish school regions. Combined with feature matching, effective fish school targets are selected. Furthermore, tracking algorithms can be used to track the fish school in time. Finally, a structured fish school detection result containing information such as the location, quantity, species, movement trajectory, and behavioral characteristics of the fish school is output, providing accurate support for fishing decisions.

[0067] In summary, steps S1 to S3 form a progressive chain of data processing and detection: multi-dimensional preprocessing of raw underwater data to generate structured sensor signals, signal conditioning to convert into image data to be detected, and image processing algorithms to detect fish schools. This process eliminates data interference through multi-dimensional preprocessing, optimizes data quality through targeted signal conditioning, and locates and filters fish schools using precise image processing algorithms. It not only achieves efficient transformation from raw data to structured detection results but also ensures the accuracy and reliability of fish school detection, accurately outputting core information related to fish schools and providing strong technical support for fishing decisions.

[0068] As one possible implementation, in the above embodiments, step S1 may specifically include the following steps:

[0069] S1-1. Use a multi-dimensional data synchronization and alignment algorithm to integrate the collected raw underwater data to obtain multi-dimensional underwater data;

[0070] The core of this step is to resolve the time misalignment issue in data collected by different sensors, and to achieve spatiotemporal matching of multi-source data. Specifically:

[0071] The fish detection module is equipped with multiple sensor components, including high-definition image sensors, sonar sensors, water depth sensors, and water quality sensors. A unified clock source triggering mechanism is used to ensure that all sensors start collecting data synchronously (time error ≤1ms).

[0072] Image sensor: Acquires underwater optical signals according to preset parameters (such as 1080P resolution, 25 frames / second frame rate), with each frame image accompanied by a millisecond-level timestamp;

[0073] Sonar sensor: emits sound waves at a fixed frequency (e.g., 30kHz), synchronously collects sonar signals, and binds a timestamp to each transmit-receive cycle;

[0074] Water depth sensor: Real-time acquisition of water depth signals, outputting a set of depth data every 50ms, with timestamps;

[0075] Water quality sensor: Real-time acquisition of raw water turbidity signals (i.e., water quality signals, in NTU), outputting one set of data every 100ms, with a timestamp.

[0076] A multi-dimensional data synchronization alignment algorithm is adopted. Based on the timestamp of the underwater optical signal, the timestamps of the sonar signal, water depth signal, and water quality signal are matched with the timestamp of the original optical signal to filter out four types of original data within the same time window (time error ≤ 1ms). Finally, the matched underwater optical signal, sonar signal, water depth signal, and water quality signal are combined according to the time sequence to form preliminary underwater multi-dimensional data, ensuring that each set of data corresponds to the environmental information at the same underwater detection time.

[0077] S1-2. Perform preliminary noise reduction on the underwater multi-dimensional data to obtain the noise-reduced underwater multi-dimensional data;

[0078] The core of this step is to specifically remove explicit noise from different types of data to avoid interfering with subsequent calibration and integration, while improving data quality while retaining valid information. Specifically:

[0079] Based on the noise characteristics of raw data from different sensors, a targeted filtering algorithm is used to remove obvious interference signals (without changing the essence of the effective signal):

[0080] Underwater optical signal noise reduction: A median filtering algorithm (window size 3×3) is used to remove suspended particles in the water and salt-and-pepper noise caused by camera light-sensing noise, while retaining effective optical information such as fish outlines and underwater background;

[0081] Sonar signal noise reduction: An adaptive moving average filtering algorithm (sliding window length of 5 sampling points) is adopted to smooth the signal jitter caused by water ripples during sound wave propagation. At the same time, weak and invalid signals caused by environmental noise (such as water flow sound and external sound wave interference) are eliminated through energy threshold screening (setting a minimum reflection energy threshold).

[0082] Water depth / water quality signal noise reduction: A first-order low-pass filter algorithm (cutoff frequency 1Hz) is adopted to filter high-frequency fluctuations caused by interference from the sensor's own circuitry or water flow impact, making the depth and turbidity data more stable.

[0083] S1-3. The noise-reduced underwater multi-dimensional data is calibrated to obtain calibrated underwater multi-dimensional data.

[0084] The core of this step is to achieve format unification, range calibration, and error correction for multi-source data, ensuring data comparability and accuracy. The specific steps are as follows:

[0085] Signal format standardization: Converting raw data from different sensors into a unified digital signal format (32-bit floating-point) to avoid data integration failures due to format differences.

[0086] Underwater optical signals: converted from analog pixel signals to digital grayscale value signals (0-255 range);

[0087] Sonar signal: Converted from analog voltage signal to digital energy value signal (0-1000 range);

[0088] Water depth signal: Converted from analog resistance signal to digital depth value (0-65 meters range);

[0089] Water quality signal: Converted from analog current signal to digital turbidity value (0-50 NTU range).

[0090] Signal range calibration: Based on sensor factory calibration parameters and field environment correction, ensuring data accuracy.

[0091] Sonar signal calibration: Adjust the signal gain coefficient according to the current water depth data. For example, the gain is increased by 1.2 times for every 5 meters increase in water depth to compensate for the energy attenuation of sound waves in deep water areas.

[0092] Water depth signal calibration: Using the water surface as the zero point, the depth value is verified by the propagation time of the underwater reflected signal fed back by the sonar sensor, and the sensor drift error is corrected. The accuracy after calibration is ±0.1 meters.

[0093] Water quality signal calibration: The initial gain is preset based on the water turbidity data. For example, the gain is increased by 1.1 times for every 10 NTU increase in turbidity, providing basic parameters for subsequent signal conditioning.

[0094] S1-4. Based on the calibrated underwater multi-dimensional data, perform correlation and integration to obtain a single timestamp multi-dimensional dataset as the sensor signal to be detected for fish schools;

[0095] The core of this step is to structurally encapsulate the calibrated multi-source data into a standardized dataset that can be directly used for subsequent signal conditioning. Specifically:

[0096] Based on the timestamp matching results, the underwater image data after single-frame calibration is bound to the calibrated sonar signal data, calibrated water depth data, and calibrated water quality data of the corresponding time window into a single timestamp multi-dimensional data group.

[0097] Each data set is structured and labeled, including: data set number, acquisition timestamp, sensor operating status (normal / abnormal), and calibration parameters (such as optical gain and sonar gain).

[0098] Perform validity checks: Check the completeness (no missing fields) and rationality (e.g., water depth data does not exceed the sensor range, turbidity data is within a reasonable range) of each signal in the data group one by one, and remove invalid data groups, such as missing data caused by sensor malfunction or abnormal values ​​that exceed the range.

[0099] All valid single-timestamp multi-dimensional data sets are arranged in chronological order to form a structured sensor signal for fish school detection. This signal contains complete multi-dimensional information about the underwater environment and can be directly transmitted to the WIFI control module for signal conditioning.

[0100] In summary, the entire process from steps S1-1 to S1-4 revolves around the standardized processing of underwater multi-source raw data, effectively solving the pain points of time misalignment, heterogeneous format, and insufficient accuracy of underwater multi-source data. This enables the output sensor signals to have technical advantages such as strong spatiotemporal consistency, high data purity, and standardized format, providing high-quality and highly reliable data support for subsequent signal conditioning and image analysis for fish swarm detection.

[0101] As one possible implementation, in the above embodiments, step S2 may specifically include the following steps:

[0102] S2-1. Classify and analyze the sensor signals of the fish to be detected to obtain multi-source signal correlation data;

[0103] The core of this step is to separate signal types, verify validity, and establish correlation indexes to avoid interference from abnormal data, laying the foundation for subsequent categorized conditioning. Specifically:

[0104] The system receives a single-timestamp multi-dimensional dataset from the fish detection module, which consists of sensor signals from the fish swarm to be detected. The data is then automatically separated into four sub-signals based on the signal type: optical sub-signal (digital grayscale image), sonar sub-signal (digital energy value sequence), water depth sub-signal (digital depth value), and water quality sub-signal (digital NTU value).

[0105] Perform signal validity verification: Remove outliers from each sub-signal that are outside the reasonable range, such as optical sub-signal grayscale value >255 or <0, water depth >65 meters, turbidity >50 NTU, and fill in missing data (using adjacent valid data interpolation method) to ensure the reliability of the treatment object.

[0106] Establish a signal association index: taking a single frame image of the optical signal as the core, bind the sonar signal, water depth data, and water quality data within the corresponding time window to generate a signal association group of single frame core and multi-source auxiliary, providing environmental parameter support for categorized conditioning.

[0107] S2-2. Based on the multi-source signal correlation data, perform categorized signal conditioning to obtain the conditioned multi-source signal;

[0108] This step employs differentiated conditioning strategies for the four types of sub-signals in the multi-source signal correlation data: optical, sonar, water depth, and water quality. Finally, it integrates the various optimized sub-signals and outputs the conditioned multi-source signal, laying a high-quality signal foundation for subsequent signal fusion.

[0109] S2-3. Perform fusion processing on the conditioned multi-source signal correlation data to obtain a composite image signal fused from the multi-signal data;

[0110] The core of this step is to integrate the detail advantages of optical signals with the anti-hazing advantages of sonar signals. By optimizing image adaptability through calibration parameters, the advantages of multiple signal sources are complemented to generate a composite image signal that combines image detail with sonar anti-interference capabilities. Specifically:

[0111] Spatial coordinate alignment: Using the conditioned optical image signal as the reference coordinate system, the target area (such as fish schools or obstacles) of the sonar pseudo-color image is accurately mapped to the corresponding position of the optical image through time stamp synchronization and spatial mapping algorithm, ensuring that the target coordinates of the two images are consistent (error ≤ 2 pixels).

[0112] Weighted fusion calculation: A primary and secondary signal weighted fusion algorithm is adopted, setting the weight of the optical sub-signal to 0.7 (core detection basis), the weight of the sonar sub-signal to 0.2, and the weight of the water depth / water quality sub-signal to 0.1, and performing pixel-level weighted calculation. Among them, the calibration parameter correction value is adjusted according to the turbidity, and the contribution of the sonar signal is increased in the case of severe turbidity.

[0113] Edge enhancement and background suppression: The Laplacian operator edge enhancement algorithm is applied to the fused image to strengthen the outline edges of the fish school; at the same time, adaptive threshold background suppression is used to reduce the brightness and contrast of the water background, highlight the fish school and obstacle targets, and output a composite image signal fused from multiple signals.

[0114] S2-4. Perform secondary noise reduction processing on the composite image signal fused from the multi-signal fusion to obtain image data for the fish school to be detected;

[0115] The core of this step is to eliminate the superimposed noise generated during the fusion process and output standardized detection image data. Specifically:

[0116] Secondary noise reduction optimization: Median filtering (3×3 window) is used to perform secondary noise reduction on the fused image to remove a small number of noise points and artifacts introduced during signal fusion;

[0117] Format and resolution standardization: Convert composite images to RGB 8-bit format, 1080P resolution, and 25 frames per second frame rate to ensure compatibility with the input requirements of subsequent image processing algorithms (such as semantic segmentation and feature extraction).

[0118] Structured annotation: Add metadata annotations to each frame of image, including: acquisition timestamp, current water depth, turbidity level, signal conditioning parameters (optical gain, sonar amplification factor), and image credibility score, to facilitate subsequent algorithm traceability and adaptation adjustments.

[0119] In summary, steps S2-1 to S2-4 not only solve the problem that single sensor signals are easily interfered with by noise and turbidity in complex underwater environments, significantly improving the distinction between fish targets and the background, but also enhance the adaptability of images to different water depths and turbidity scenes through the complementary advantages of multi-source signals. This provides high-quality and reliable input data for subsequent accurate detection of fish, effectively reducing the false positive and false negative rates in subsequent detection stages.

[0120] As one possible implementation, in the above embodiments, step S2-2 may specifically include the following steps:

[0121] S2-2-1. Using the multi-source signal correlation data, obtain optical sub-signals, sonar sub-signals, and water depth and water quality sub-signals;

[0122] The multi-source signal correlation data is a previously established time-series, structured correlation dataset. This step, based on differences in signal attributes such as physical characteristics, data format, and functional positioning, precisely separates four core sub-signals from the correlation data: optical sub-signals (raw digital grayscale images / pixel sequences), sonar sub-signals (time-series digital energy values), water depth sub-signals (digital depth values), and water quality sub-signals (NTU turbidity values). The aim is to achieve signal classification and management, avoid mutual interference between different types of signals, and provide independent processing objects for subsequent precise classification and adjustment.

[0123] S2-2-2. Based on the optical sub-signal, a combined algorithm of Gaussian filtering and bilateral filtering is used to obtain the conditioned optical sub-signal;

[0124] Optical signals are susceptible to interference from the underwater environment, such as scattering from suspended particles in the water and noise in the camera circuitry, leading to blurred images and loss of detail at the edges of fish schools. This step employs a combination algorithm of Gaussian filtering (5×5 window) and bilateral filtering:

[0125] First, Gaussian filtering is used to smooth the optical sub-signal, quickly attenuating high-frequency noise such as specks from tiny particles and high-frequency interference from circuits. Then, bilateral filtering is used to further reduce noise while accurately preserving key details such as the edges of the fish school and the texture of fish scales, thus solving the problem of excessive blurring of the target outline caused by single Gaussian filtering. Finally, the denoised and detailed adjusted optical sub-signal is output, laying a clear signal foundation for subsequent image correction.

[0126] S2-2-3. Based on the conditioned optical sub-signal and water quality data, adaptively adjust the gain and use the gray world method and adaptive exposure compensation correction to obtain the optical image signal;

[0127] Underwater turbidity and changes in water depth can cause problems such as insufficient brightness and color distortion in optical signals. This step involves two parts:

[0128] First, the gain is adaptively adjusted based on water quality data. Specifically, the gain coefficient is dynamically set according to the NTU turbidity level: 1.0 for clear water ≤10 NTU, 1.2-1.5 for moderate turbidity 10-30 NTU, and 1.5-1.8 for heavy turbidity >30 NTU, ensuring clear fish signals in low-light / turbid environments. Second, the grayscale world method is used to correct color casts caused by water refraction, making the image colors closer to the real environment. Simultaneously, adaptive exposure compensation adjusts the brightness histogram distribution to address the issue of overly dark images caused by increased water depth. The final output is an optical image signal (RGB format, directly usable for subsequent fusion or detection) with balanced brightness, true color, and clear details.

[0129] S2-2-4. Perform signal enhancement and interference frequency filtering on the sonar sub-signal to obtain the conditioned sonar sub-signal;

[0130] Sonar sub-signals are prone to attenuation (e.g., signals reflected by small fish) and interference (e.g., water noise, low-frequency external sound waves) during propagation, affecting target identification. The core of this step is weak signal enhancement and interference filtering.

[0131] On one hand, a programmable gain amplifier (PGA) is used to specifically enhance the sonar sub-signal, amplifying weak reflection signals such as fish schools by 3-8 times, while limiting the gain of strong signals such as those from the bottom of the water and large obstacles to avoid signal saturation distortion. On the other hand, a bandpass filtering algorithm (center frequency 30kHz, bandwidth ±5kHz) is used to accurately eliminate interference signals of non-target frequencies (such as water flow noise and irrelevant external sound waves), while retaining the effective reflection signals of fish schools and obstacles. The final output is a conditioned sonar sub-signal with a high signal-to-noise ratio and prominent target signal.

[0132] S2-2-5. Based on the conditioned sonar sub-signal, a grayscale and pseudo-color mapping algorithm is used to obtain the sonar pseudo-color image signal;

[0133] The conditioned sonar sub-signals are digital energy value sequences, which are abstract in form, making it difficult to intuitively observe target distribution and impossible to directly fuse with optical images. This step uses a grayscale and pseudo-color mapping algorithm to convert the digital energy values ​​of the sonar sub-signals into intuitive pseudo-color images. Specifically:

[0134] Color mapping rules are set according to energy levels: high energy, such as large obstacles, is mapped to red; medium energy, such as schools of fish, is mapped to yellow; and low energy, such as water backgrounds, is mapped to blue. Simultaneously, the image resolution is unified to match the optical image, ensuring spatial coordinate matching during subsequent fusion. The final output is a sonar pseudo-color image signal that intuitively reflects the energy distribution of the target, enabling visualization of the sonar signal and providing visual input for multi-source signal fusion.

[0135] S2-2-6. Based on the water depth and water proton signals, a moving average filter is used to obtain the conditioned water depth and water proton signals;

[0136] The water depth and raw water quality sub-signals are susceptible to fluctuations caused by water flow and minor vibrations in the detection equipment, such as instantaneous drops in water depth and small fluctuations in turbidity, which can affect the reliability of subsequent calibration parameters. This step employs a moving average filtering algorithm (window length of 3 sampling points) to smooth the water depth and water quality sub-signals. Specifically, by averaging the effective data from three adjacent sampling points, instantaneous fluctuations are filtered out, while data trends are preserved, preventing volatile data from interfering with calibration parameter calculations. The final output is a stable and continuous conditioned water depth and water quality sub-signals, providing reliable data support for the accurate extraction and fusion calibration parameters.

[0137] S2-2-7. Extract calibration parameters from the conditioned water depth sub-signal and water quality sub-signal to obtain a fused calibration parameter set;

[0138] The conditioned water depth and water proton signals are raw digital values ​​and cannot be directly used for calibration of multi-source signal fusion. The core of this step is to transform the raw data into standardized fusion calibration parameters: first, extracting basic calibration parameters, including the stable depth value of the current detected water layer and the water turbidity level (clear / moderately turbid / severely turbid); second, calculating derived calibration parameters, such as calculating the sonar signal reliability score (0-10 points) based on the signal-to-noise ratio of the sonar sub-signals. Finally, these parameters are integrated into a fusion calibration parameter set, providing a quantitative calibration basis for the subsequent weighted fusion of optical image signals and sonar pseudo-color image signals, such as adjusting the fusion weights of the two types of images according to the turbidity level.

[0139] S2-2-8. Acquire the optical image signal, the sonar pseudo-color image signal, and the fusion calibration parameter set as a conditioned multi-source signal;

[0140] By aggregating the acquired optical image signals (main image source), sonar pseudo-color image signals (auxiliary image source), and fusion calibration parameter set (fusion calibration basis), a complete conditioned multi-source signal is formed. This signal features clear main image, complementary auxiliary images, and controllable parameters. It retains the detail advantages of optical images and the anti-interference advantages of sonar images, and adapts to different underwater environments through calibration parameters. This provides high-quality, standardized input data for subsequent multi-source signal fusion and accurate fish detection.

[0141] In summary, steps S2-2-1 to S2-2-8, through the design of targeted conditioning by type, environmental adaptive optimization, and parameterized calibration, not only solve the problem of single sub-signals being susceptible to noise and environmental interference, but also significantly improve the clarity, stability, and anti-interference ability of various signals. Furthermore, by deeply binding water quality and depth data with signal conditioning, the adaptability to different underwater environments is enhanced. At the same time, the complementary advantages of optical signal detail and sonar signal anti-interference are achieved, providing standardized and highly reliable input data for subsequent multi-source signal fusion and accurate fish detection, effectively reducing the false positive rate, false negative rate, and computational complexity of subsequent detection processes.

[0142] As one possible implementation, in the above embodiments, step S3 may specifically include the following steps:

[0143] S3-1. Using the image data of the fish school to be detected, obtain candidate images of fish to be detected;

[0144] Taking mixed image data containing / without fish as input, the system first optimizes image quality through targeted noise reduction and adaptive contrast enhancement. Then, it combines sonar energy ratio and image grayscale variance prediction to remove obviously fishless images. Finally, it uses a semantic segmentation model to extract effective suspected fish areas and screen out candidate images with potential fish targets. Its core function is to quickly eliminate interference from fishless images, identify key processing targets, and reduce the computational complexity of subsequent processes.

[0145] S3-2. The connected component labeling algorithm is used to extract features from the candidate images of fish to be detected to obtain images of fish to be detected.

[0146] Candidate images containing fish may still contain non-fish interference areas such as aquatic plants and long, narrow obstacles that resemble the shape of fish. Further verification is needed. Specifically:

[0147] Extraction of a single fish group region: For the segmentation mask of the candidate image with fish, a connected component labeling algorithm is used to label each independent fish group target (single fish or cluster) in the mask of the candidate image with fish, generate the minimum bounding rectangle and record the pixel coordinates.

[0148] Dual feature extraction of morphology and texture: For each marked region, morphological features such as Hu moment invariants (7 feature values), region area, aspect ratio, and contour complexity are extracted, as well as LBP texture feature histogram (256 dimensions), forming a 263-dimensional feature vector.

[0149] Feature validity verification, excluding non-fish interference:

[0150] Feature threshold filtering: Set the range of morphological features of fishing species, such as aspect ratio 1.5-5.0 and area ≥50 pixels, and remove feature vectors that do not meet the range, such as flat obstacles and long strips of aquatic plants;

[0151] Fishless feature comparison: The filtered feature vectors are compared with the fishless image background feature library. If the similarity is ≥0.8, it is determined to be a non-fish interference area and is removed. The fishless image background feature library contains feature templates for blank water bodies, bubbles, and obstacles.

[0152] Results are categorized as follows: If all marked regions are removed, the image is moved from the fish-containing candidate image queue to the fish-free queue; if at least one valid feature vector is retained, it is determined to be a fish-containing image to be detected and proceeds to the next step of recognition. The aim is to accurately eliminate non-fish interference, ensuring that the images entering subsequent detection stages are genuine fish-containing images, thereby improving the accuracy of the detection results.

[0153] S3-3. Detect fish based on the image of fish to be detected and obtain the fish detection result.

[0154] Based on confirmed images of fish, a feature matching algorithm is used to identify specific fish species and suspected targets. Then, Kalman filtering and the Hungarian algorithm are used to track the inter-frame correlation of fish schools in time-series images, avoiding double counting or omissions. Finally, the pixel information is converted into actual body size and water location by combining image metadata, and the number of fish species and movement patterns are counted. The output is a structured detection result containing core information. The core value is to accurately extract key information about fish schools, providing a comprehensive and quantitative basis for subsequent decision-making.

[0155] In summary, the process from steps S3-1 to S3-3, through a closed-loop design of layer-by-layer screening, precise elimination, and quantitative detection, effectively filters out fishless images and non-fish interference (aquatic plants, bubbles, etc.), significantly improving the accuracy and anti-interference ability of fish detection. Furthermore, the pre-screening reduces invalid calculations and improves overall processing efficiency, ultimately providing comprehensive and reliable quantitative data support for subsequent decisions such as fishing guidance and fisheries monitoring.

[0156] As one possible implementation, in the above embodiments, step S3-1 may specifically include the following steps:

[0157] S3-1-1. Perform noise reduction and enhancement and adaptive parameter adjustment on the image data to be detected to obtain preprocessed image data for fish detection.

[0158] The image data to be detected is a mixed dataset containing images of fish and images without fish, and it suffers from problems such as turbid water, noise interference, and complex backgrounds. This step is the basic preprocessing stage of the entire screening process. The input is the image data to be detected, and the core operation includes two parts:

[0159] Noise reduction and enhancement processing: A combination algorithm of median filtering (3×3 window) and guided filtering is used to remove random noise in the image (such as water particle reflection noise and equipment electronic noise). At the same time, the CLAHE (contrast-limited adaptive histogram equalization) algorithm is used to improve the grayscale contrast between the fish area and the background and enhance the outline details of the fish.

[0160] Adaptive parameter adjustment: Based on environmental metadata, such as appropriately increasing the filter kernel size for higher turbidity and adjusting the enhancement intensity coefficient for deeper water, the key parameters for noise reduction and enhancement are dynamically optimized to avoid insufficient adaptability caused by fixed parameters. The output is preprocessed image data for fish detection, aiming to unify image quality standards, eliminate the impact of environmental interference and equipment noise on subsequent screening, and lay the foundation for accurate extraction of screening indicators.

[0161] S3-1-2. Based on the preprocessed image data of the fish swarm to be detected, the sonar energy ratio and image grayscale variance are calculated to obtain the preliminary screened image data;

[0162] Using preprocessed image data as input, images clearly devoid of fish are quickly removed through two quantification metrics, reducing computational redundancy in subsequent processing, as detailed below:

[0163] Calculate the proportion of sonar energy: Using sonar auxiliary data, extract the energy value of the corresponding sonar reflection signal in the image, and calculate the proportion of this energy value to the total sonar energy of the whole image. As a physical target, fish will produce specific sonar reflections, and their energy proportion is usually higher than the preset low threshold.

[0164] Calculate the grayscale variance of the image: Statistically analyze the dispersion of grayscale values ​​in the preprocessed image. In images with fish, the grayscale difference between the fish area and the background is significant, and the grayscale variance is usually higher than the preset low threshold. Images that meet both indicators and are higher than the corresponding preset low threshold are retained to form preliminary screened image data. The purpose is to quickly narrow down the candidate range by quantifying the indicators, filter out low-energy, low-grayscale difference images caused by the absence of fish, and improve screening efficiency.

[0165] S3-1-3. Determine whether the image data of the preliminary screening is a fishless image. If so, obtain the adjustment instruction and execute step S3-1-6. Otherwise, obtain the image data of the preliminary screening as a fish-containing image of the preliminary screening.

[0166] Using the initially screened image data as input, the core is to perform a secondary verification of the initial screening results to avoid false screenings caused by unreasonable threshold settings. Specific operations are as follows:

[0167] Based on the preset rules for determining fishless images, such as the sonar energy ratio being higher than but close to the low threshold and the grayscale variance having no obvious peak, or combined with auxiliary judgment conditions such as whether there is a large area of ​​uniform background in the image and no suspected target contours, the preliminary screened image data is judged to be fishless images.

[0168] If an image is determined to be fishless, the system generates an adjustment command, which includes the dimensions to be adjusted, such as the detection position of the acquisition device, and executes step S3-1-6. If a suspected fish school is determined to exist, the image data initially screened is directly used as an image with fish in the initial screening and proceeds to the next stage. The purpose is to build a feedback verification mechanism for the initial screening, preventing fishless images from entering subsequent complex processing flows, and optimizing the original data acquisition conditions through adjustment commands to improve the effectiveness of subsequent data.

[0169] S3-1-4. Based on the pre-selected images containing fish, a pre-trained semantic segmentation model is used to obtain the mask image of the images containing fish.

[0170] Using pre-selected images containing fish as input, the core objective is to achieve pixel-level precise segmentation of the fish school region through a deep learning model. Specific operations include:

[0171] Model foundation: The pre-trained semantic segmentation model used is based on an underwater fish swarm dataset, which includes images of fish swarms of different species and in different environments, along with corresponding pixel-level annotations. The annotation categories are divided into fish swarms, underwater background, and non-fish interference objects, such as aquatic plants, bubbles, and obstacles, and are trained accordingly.

[0172] Inference Process: Initially selected images containing fish are input into a pre-trained model. The model performs feature extraction, upsampling, and pixel classification to output a mask image of the fish-containing images. In the mask image, each pixel corresponds to a category label, retaining only the pixel labels for the fish group category, while other categories are labeled as background. The goal is to accurately separate fish-containing regions from non-fish-containing regions from the initial images containing fish, providing pixel-level evidence for subsequent verification of valid fish-containing regions.

[0173] S3-1-5. Verify whether there is a valid fish area in the mask image of the fish image. If there is, obtain the preliminarily screened fish image as a fish candidate image to be detected. Otherwise, obtain the adjustment instruction and execute step S3-1-6.

[0174] Using a masked image of fish as input, the core objective is to eliminate false positive areas representing fish, such as small areas of noise or debris in the masked image that are mistakenly identified as fish, and to confirm the true and valid fish population. The specific steps are as follows:

[0175] Define the criteria for an effective fish school area: two core conditions are preset, including the area threshold, such as the pixel area of ​​a single fish school area is ≥50 pixels, to avoid misjudging small bubbles and particles as fish schools; and the area shape threshold, such as the area aspect ratio is between 1.5 and 5, which conforms to the shape proportions of common fish schools or single fish.

[0176] Verification process: Traverse the fish category regions in the mask image and determine if there are regions that meet the above criteria. If they exist, the initially screened images containing fish are confirmed as candidate images for fish detection, and the process proceeds to the subsequent fish detection stage; if they do not exist, the system generates an adjustment instruction and executes step S3-1-6. The purpose is to filter out false positive interference through quantification criteria, ensure the authenticity of candidate images containing fish, and maintain the closed-loop feedback characteristic of the process to improve the overall reliability of the screening.

[0177] S3-1-6. Based on the adjustment command, acquire new underwater raw data and return to S1;

[0178] This step is a closed-loop control operation of the process. The triggering condition is either step S3-1-3 determining that there is no fish in the image or step S3-1-5 verifying that there is no effective fish population in the area. Specific operations are as follows:

[0179] Instruction parsing: Based on the adjustment instructions generated in the preceding steps, the original data acquisition parameters that need to be optimized are clearly defined;

[0180] Data Acquisition: Based on the adjusted parameters, new raw underwater data are reacquired;

[0181] Process restart: Return the newly acquired underwater raw data to step S1 to restart. The purpose is to avoid the failure of screening due to insufficient raw data quality or unreasonable acquisition parameters through a closed-loop mechanism of filtering invalid data, parameter adjustment, re-acquisition, and process restart, thereby improving the process's adaptability to complex underwater environments and ensuring that effective fish-containing candidate images can be obtained in the end.

[0182] In summary, steps S3-1-1 to S3-1-6 employ a multi-layered design—preprocessing to optimize data quality, rapid redundancy reduction using dual indicators, precise semantic segmentation for localization, quantization verification to remove false positives, and a closed-loop mechanism for adaptive adjustment—to effectively filter out equipment noise, water interference, and false positive targets, significantly improving the accuracy and reliability of fish-containing image screening. Furthermore, pre-screening reduces unnecessary computation, and adaptive parameters adapt to complex underwater environments, significantly enhancing screening efficiency and environmental adaptability. Simultaneously, the closed-loop feedback mechanism ensures the effectiveness of data collection, providing a high-quality, highly reliable input foundation for subsequent accurate fish detection (fish species identification, quantity statistics, etc.), effectively reducing the false positive rate and computational complexity of subsequent detection.

[0183] As one possible implementation, in the above embodiments, step S3-3 may specifically include the following steps:

[0184] S3-3-1. Based on the image of fish to be detected, a connected component labeling algorithm is used to obtain the suspected fish area.

[0185] Taking an image containing fish to be detected as input, the core operation is to separate all potential fish targets from the image using a connected component labeling algorithm. The specific steps are as follows:

[0186] First, the image containing fish is preprocessed, such as by grayscale conversion and binarization, transforming the image into black and white to highlight the difference between the target area and the background. Then, a connected component labeling algorithm is used to scan the binarized image, marking adjacent pixels with the same pixel value (target pixel value) as the same connected component. Each connected component represents a suspected fish school region. The output is the suspected fish school region, the purpose of which is to quickly locate all possible fish targets in a complex underwater background, laying the foundation for subsequent feature extraction and matching.

[0187] S3-3-2. Based on the suspected fish area, perform feature extraction to obtain the feature vector of the suspected fish area;

[0188] Using the suspected fish population area as input, representative features of each suspected area are extracted to form a feature vector. Specific operations include:

[0189] For each suspected fish school area, multiple features are extracted, including shape features (such as area, perimeter, roundness, rectangularity, aspect ratio, etc., used to describe the geometry of the area), texture features (such as energy, entropy, contrast, etc. of the gray-level co-occurrence matrix, used to describe the arrangement of pixels within the area), and gray-level features (such as average gray level, gray-level variance, maximum gray level, minimum gray level, etc., used to describe the gray-level distribution of the area). All extracted feature values ​​are combined into a vector, namely the feature vector of the suspected fish school area. The purpose is to transform the information of the suspected fish school area into a mathematical form, facilitating matching and comparison with a feature template library of fishing species.

[0190] S3-3-3: Perform normalized cross-correlation matching between the feature vector of the suspected fish school area and the feature template library of fishing species to filter out the effective fish school area, and then obtain a time-series image sequence of fish.

[0191] The effective fish population regions with high matching degree with the feature template of the fishing species were selected by normalized cross-correlation matching algorithm, and a time-series sequence of fish images was constructed according to the time order. The specific operation is as follows:

[0192] The fishing species feature template library stores typical feature vector templates for various fishing species. For each suspected fish school region's feature vector, its normalized cross-correlation coefficient with each feature template in the library is calculated. This coefficient reflects the similarity between the two vectors. A matching threshold is set; when the cross-correlation coefficient is greater than the threshold, the suspected fish school region is considered to have successfully matched the corresponding fish species template, thus becoming a valid fish school region. Based on the temporal order of image acquisition, images containing valid fish school regions are combined into a time-series sequence of fish-containing images. The output is a time-series sequence of fish-containing images, the purpose of which is to filter out the true fishing school regions, exclude non-target fish species or interference regions, and provide continuous image data for subsequent fish school target tracking.

[0193] S3-3-4. Apply Kalman filtering and Hungarian algorithm to the time-series fish image sequence to obtain a continuous trajectory dataset of the fish target;

[0194] The fish swarm target was tracked using Kalman filtering and the Hungarian algorithm to obtain its continuous trajectory dataset, as detailed below:

[0195] Kalman filtering is used to predict and update the position and velocity of fish targets. In a time-series of fish images, the initial position of the fish target is first detected in the first frame image. Then, the possible position of the fish target in the next frame image is predicted based on Kalman filtering. Next, the observed position of the fish target is found in the actual next frame image through a detection algorithm. The predicted position is then corrected using the update equation of Kalman filtering to obtain a more accurate target position.

[0196] The Hungarian algorithm is used to solve the data association problem in multi-target tracking. When there are multiple fish targets, the Hungarian algorithm is used to find the optimal matching relationship by calculating the distance or similarity between the predicted position and the observed position, and then assigning each observed position to the corresponding predicted target, thereby achieving simultaneous tracking of multiple fish targets.

[0197] By processing the entire time-series of images of fish, a continuous trajectory dataset of the fish targets is obtained, which includes information such as the position and velocity of each fish target at different time points. The output is a continuous trajectory dataset, the purpose of which is to achieve continuous tracking of the fish targets and obtain their motion trajectory and behavioral characteristics.

[0198] S3-3-5. Based on the continuous trajectory dataset of the fish target, obtain the fish detection results;

[0199] Using a continuous trajectory dataset of fish targets as input, the core task is to analyze and process the trajectory data to obtain the final fish detection result. Specifically:

[0200] The trajectory of each fish target in a continuous trajectory dataset is analyzed, including calculating parameters such as trajectory length, velocity changes, and direction changes, to understand the movement patterns and behavioral characteristics of the fish school. Simultaneously, by combining image information from a time-series sequence of fish images, the number, size, and species of the fish school can be further estimated and confirmed. The final fish school detection result is generated by combining the above analysis results, including information such as the location, number, species, movement trajectory, and behavioral characteristics of the fish school. The output is the fish school detection result, aiming to provide users with comprehensive and accurate fish school-related information to meet their needs for fish school monitoring and research.

[0201] In summary, steps S3-3-1 to S3-3-5 employ a connected component labeling algorithm to achieve rapid and accurate localization of suspected fish school areas, effectively eliminating interference from complex backgrounds. Feature extraction and normalized cross-correlation matching, combined with a fishing species feature template library, significantly improve the screening accuracy of effective fish school areas and reduce interference from non-target fish species and false positives. The Kalman filter's prediction-update mechanism, in conjunction with the Hungarian algorithm's optimal data association capability, successfully solves the problems of target occlusion and position drift in multi-target tracking, ensuring the continuity and accuracy of fish school trajectories. Finally, through trajectory data analysis, comprehensive and accurate output of information such as fish school location, quantity, species, and movement characteristics is achieved. The overall process balances detection efficiency and reliability, significantly reducing the false positive and false negative rates of fish school detection in complex underwater environments, providing high-quality technical support for scenarios such as fish school monitoring and fishery resource assessment.

[0202] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides an embodiment of a fish school detection device for fishing, which is similar to... Figure 1 Corresponding to the method embodiments shown, this device can be specifically applied to various electronic devices.

[0203] like Figure 2 As shown, a fish detection device for fishing in this embodiment includes: a fish detection module, a WIFI control module, and an intelligent analysis module;

[0204] The fish detection module is used to perform multi-dimensional processing on raw underwater data to obtain sensor signals for the fish school to be detected.

[0205] This module serves as the core of the front-end data acquisition and preprocessing system for fish school detection. It integrates a high-definition image sensor, sonar sensor, depth sensor, and water quality sensor using a fish finder, and is equipped with a movable float structure as the carrier module. The movement of the float is driven by a WIFI control module, which can adjust its floating position and other attitudes according to instructions to switch detection areas. This module achieves synchronous acquisition of multiple types of sensors (time error ≤1ms) through a unified clock source triggering mechanism, accurately capturing multiple types of raw underwater signals. Differentiated filtering algorithms are used to remove invalid interference based on the noise characteristics of different signals. It also completes the format standardization, range calibration, and timestamp association integration of multi-source data, and finally outputs structured sensor signals for fish school detection that include core signals, auxiliary signals, and metadata, providing comprehensive and effective raw data support for subsequent processing.

[0206] The WIFI control module is used to acquire image data of the fish school to be detected using a signal conditioning method based on the sensor signal of the fish school to be detected.

[0207] This module utilizes a fish finder WIFI box, combining data transmission, signal conditioning, and command execution as its core functions. It serves as a crucial hub connecting the fish detection module and the intelligent analysis module. On one hand, it receives sensor signals from the fish detection module via a two-way wireless communication link. It first verifies the signal validity, then performs targeted conditioning based on the signal type: optical signal noise reduction, gain adjustment, and color shift correction; sonar signal amplification, filtering, and pseudo-color mapping; and water depth and water quality signal processing for fusion calibration parameters. Subsequently, using the optical signal as a reference, it aligns the spatial coordinates of the sonar signal and performs pixel-level weighted fusion. After further optimization, it generates image data for fish detection, which is then encapsulated via TCP / IP protocol and transmitted to the intelligent analysis module with low latency and high stability via the WIFI link. On the other hand, it receives control commands from the intelligent analysis module, such as commands to adjust parameters in the detection area, and parses them into executable control signals. This precisely drives the float mechanism of the fish detection module, enabling float depth adjustment, horizontal displacement, and other actions, optimizing the subsequent data acquisition range and quality.

[0208] The intelligent analysis module is used to perform fish detection using image processing algorithms based on the image data of the fish to be detected, and to obtain the fish detection results.

[0209] This module is the core module for generating fish detection results and the initiator of commands for system closed-loop control. After receiving the image data of the fish to be detected transmitted by the WIFI control module, it first optimizes the image quality through operations such as noise reduction and adaptive enhancement, and then initially filters images based on sonar energy ratio and image grayscale variance. For candidate images with fish, a pre-trained simplified semantic segmentation model is loaded to separate the fish from the background. After morphological optimization and small region removal, the suspected fish area is located, feature vectors are extracted and matched with a fishing species template library to filter effective targets. Kalman filtering and Hungarian algorithm are used to achieve inter-frame correlation tracking of the fish to avoid counting errors. Finally, by combining image metadata, the pixel information of the fish school is converted into actual parameters, the species and number of fish are counted, and their swimming trajectories are recorded. A structured detection result containing the judgment result, core information of the fish school, and environmental parameters is generated and visualized for the user. When it is determined that there are no fish in the current detection area or the detection signal quality is poor, a targeted control command containing float movement adjustment parameters is generated and fed back to the WIFI control module through the WIFI link. This drives the fish detection module to switch to the detection area where fish are more likely to exist, improving the effectiveness of data collection and the success rate of fish detection. This ensures that the system adapts to changes in the underwater environment and matches the detection needs of fishing scenarios.

[0210] In summary, the fish detection module, WIFI control module, and intelligent analysis module of the fish school detection system work together to construct a closed-loop system of "collection-conditioning-analysis-feedback-adjustment". The fish detection module collects and outputs structured signals through simultaneous acquisition and precise preprocessing by multiple sensors. The WIFI control module realizes signal conditioning, low-latency transmission, and command parsing and execution. The intelligent analysis module relies on optimization algorithms to complete accurate detection, tracking, and parameter conversion of fish schools and generate adaptive adjustment commands. Through the efficient linkage of each module, the system achieves accurate identification of underwater fish schools, statistics on fish species and numbers, and recording of swimming trajectories. It can adapt to changes in the underwater environment to optimize the detection area, effectively improving the effectiveness of data collection and the success rate of fish school detection, and providing accurate and comprehensive technical support for fishing decisions.

[0211] In this embodiment, the specific processing of a fish school detection device for fishing and its resulting technical effects can be referred to separately. Figure 1 The relevant descriptions of steps S1, S2 and S3 in the corresponding embodiments will not be repeated here.

[0212] It should be noted that the implementation details and technical effects of each module and unit in the device provided in the embodiments of this disclosure can be referred to the description of other embodiments in this disclosure, and will not be repeated here.

[0213] The following is for reference. Figure 3 It shows a schematic diagram of the structure of a computer system 500 suitable for implementing the electronic device of the present disclosure. Figure 3The computer system 500 shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0214] like Figure 3 As shown, the computer system 500 may include a processing device 501 (e.g., a central processing unit, a graphics processing unit, etc.), which can perform various appropriate actions and processes according to a program stored in ROM 502 or a program loaded from storage device 508 into random access RAM 503. RAM 503 also stores various programs and data required for the operation of the computer system 500. The processing device 501, ROM 502, and RAM 503 are interconnected via bus 504. I / O interface 505 is also connected to bus 504.

[0215] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 508 including, for example, magnetic tapes, hard disks, etc.; and communication devices 509. Communication device 509 allows computer system 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 A computer system 500 with various electronic devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively.

[0216] In particular, according to embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, it performs the functions defined in the methods of embodiments of this disclosure.

[0217] It should be noted that the computer-readable medium described in this disclosure can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this disclosure, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this disclosure, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0218] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device.

[0219] The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the following functions: Figure 1 The illustrated embodiments and their alternative implementations demonstrate a method for detecting fish schools used in fishing.

[0220] Computer program code for performing the operations of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0221] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0222] The units or modules described in the embodiments of this disclosure can be implemented in software or hardware. The names of the units or modules do not necessarily limit the unit itself; for example, an acquisition module can also be described as "acquiring preset prompts, including modality fusion prompts, attention mechanism prompts, and / or time-related prompts."

[0223] The above description is merely a preferred embodiment of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features disclosed in this disclosure that have similar functions.

Claims

1. A method for detecting fish schools in angling, characterized in that, include: S1. Utilize raw underwater data for multi-dimensional processing to obtain sensor signals for fish detection; S2. Based on the sensor signal of the fish school to be detected, the image data of the fish school to be detected is obtained using a signal conditioning method; S3. Based on the image data to be detected, perform fish detection using an image processing algorithm to obtain the fish detection result.

2. The method for detecting fish schools for angling according to claim 1, characterized in that, S1. Utilize raw underwater data for multi-dimensional processing to obtain sensor signals for fish detection, including: The collected raw underwater data is integrated using a multi-dimensional data synchronization and alignment algorithm to obtain multi-dimensional underwater data; Preliminary noise reduction is performed on the underwater multi-dimensional data to obtain the noise-reduced underwater multi-dimensional data; The noise-reduced underwater multi-dimensional data is calibrated to obtain calibrated underwater multi-dimensional data. The calibrated underwater multidimensional data is correlated and integrated to obtain a single-timestamp multidimensional dataset as the sensor signal for fish swarm detection.

3. The method for detecting fish schools for angling according to claim 1, characterized in that, S2. Based on the sensor signal of the fish school to be detected, image data for fish school detection is obtained using a signal conditioning method, including: The sensor signals to be detected from the fish school are classified and analyzed to obtain multi-source signal correlation data, which includes optical sub-signals, sonar sub-signals, water depth sub-signals and water quality sub-signals. Based on the multi-source signal correlation data, classify signal conditioning is performed to obtain the conditioned multi-source signal; The conditioned multi-source signal correlation data is fused to obtain a composite image signal fused from the multi-signal data. Secondary noise reduction processing is performed on the composite image signal fused from the multi-signal fusion to obtain image data for fish detection.

4. The method for detecting fish schools for angling according to claim 3, characterized in that, Based on the multi-source signal correlation data, categorized signal conditioning is performed to obtain the conditioned multi-source signal, including: Using the multi-source signal correlation data, optical sub-signals, sonar sub-signals, water depth sub-signals, and water quality sub-signals are obtained; Based on the optical sub-signal, a combined algorithm of Gaussian filtering and bilateral filtering is used to obtain the conditioned optical sub-signal; Based on the conditioned optical sub-signal and water quality data, the gain is adaptively adjusted, and the gray world method and adaptive exposure compensation correction are used to obtain the optical image signal. The sonar sub-signal is enhanced and interference frequency filtered to obtain a conditioned sonar sub-signal. Based on the conditioned sonar sub-signal, a grayscale and pseudocolor mapping algorithm is used to obtain the sonar pseudocolor image signal; Based on the water depth sub-signal and the water quality sub-signal, a moving average filter is used to obtain the conditioned water depth sub-signal and water quality sub-signal; The calibration parameters are extracted from the conditioned water depth sub-signal and water quality sub-signal to obtain a fused calibration parameter set; The optical image signal, the sonar pseudo-color image signal, and the fusion calibration parameter set are acquired as a conditioned multi-source signal.

5. The method for detecting fish schools for angling according to claim 1, characterized in that, S3. Based on the image data to be detected, perform fish detection using an image processing algorithm to obtain the fish detection result, including: Using the image data of the fish school to be detected, obtain candidate images of fish to be detected; The connected component labeling algorithm is used to extract features from the candidate images of fish to be detected, and the images of fish to be detected are obtained. Fish detection is performed on the image containing fish to be detected, and the fish detection results are obtained.

6. The method for detecting fish schools for angling according to claim 5, characterized in that, Using the image data of the fish school to be detected, candidate images of fish to be detected are obtained, including: The image data to be detected for fish swarms is subjected to noise reduction, enhancement, and adaptive parameter adjustment to obtain preprocessed image data for fish swarm detection. Based on the preprocessed image data of the fish swarm to be detected, the sonar energy ratio and image grayscale variance are calculated to obtain preliminary screened image data; Determine whether the initially screened image data is a fishless image. If so, obtain an adjustment instruction and execute the first operation; otherwise, obtain the initially screened image data as an initially screened fish-containing image. Based on the pre-selected images containing fish, a pre-trained semantic segmentation model is used to obtain the mask image of the images containing fish. Verify whether there is a valid fish area in the mask image of the fish image. If there is, obtain the preliminarily screened fish image as a fish candidate image to be detected. Otherwise, obtain the adjustment instruction and execute the first operation. The first operation is: based on the adjustment instruction, acquire new raw underwater data and return to S1.

7. The method for detecting fish schools for angling according to claim 6, characterized in that, Based on the image containing fish to be detected, fish detection is performed to obtain the fish detection results, including: Based on the image of fish to be detected, a connected component labeling algorithm is used to obtain the suspected fish area; Based on the suspected fish school region, feature extraction is performed to obtain the feature vector of the suspected fish school region. The feature vectors of the suspected fish schools are matched with the feature template library of fishing species using normalized cross-correlation to filter out effective fish schools and then obtain a time-series sequence of images with fish. Kalman filtering and Hungarian algorithm are applied to the time-series fish image sequence to obtain a continuous trajectory dataset of the fish target; Based on the continuous trajectory dataset of the fish target, the fish detection results are obtained.

8. A fish school detection device for angling, employing the method as described in any one of claims 1-7, characterized in that, include: Fish detection module, WIFI control module and intelligent analysis module; The fish detection module is used to perform multi-dimensional processing on raw underwater data to obtain sensor signals for the fish school to be detected. The WIFI control module is used to acquire image data of the fish school to be detected using a signal conditioning method based on the sensor signal of the fish school to be detected. The intelligent analysis module is used to perform fish detection using image processing algorithms based on the image data of the fish to be detected, and to obtain the fish detection results.

9. An electronic device, characterized in that, include: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, It stores a computer program thereon, wherein the computer program, when executed by one or more processors, implements the method as described in any one of claims 1-7.

Images