Automated system for identifying fish gender

Abstract

This invention discloses an automated system for identifying the sex of fish, belonging to the fields of aquaculture automation and machine vision technology. The automated system for identifying the sex of fish includes: a front-end section comprising a host computer, probes, and an automatic gonad identification system; and a back-end section comprising a sorting conveyor belt. The probes are used to collect fish sex characteristics and transmit the collected images or signals to the host computer. The automatic gonad identification system uses a deep learning model to perform sex identification on the collected images and outputs a sex determination result. The host computer controls the sorting conveyor belt to perform sorting operations based on the determination result. This invention aims to solve the problem of the lack of an integrated automated system in the prior art that can integrate multimodal information acquisition (such as vision and ultrasound), automatic intelligent identification, and real-time linkage with physical sorting actuators.

Description

Technical Field

[0001] This invention relates to the fields of aquaculture automation and machine vision technology, and in particular to an automated system for identifying the sex of fish. Background Technology

[0002] In aquaculture, seedling breeding, fishery resource management, and genetic breeding research, quickly and accurately distinguishing the sex of fish is a crucial and frequently performed fundamental task. For example, in the monosex farming of economically important fish such as golden pomfret and sturgeon, it is necessary to separate males and females to improve growth efficiency and economic benefits; in the conservation of rare fish species and selective breeding, it is necessary to perform non-destructive sex identification and pairing of parent fish.

[0003] Currently, fish sex determination mainly relies on the following methods:

[0004] Artificial external morphological observation method: Operators rely on experience to identify fish by observing secondary sexual characteristics such as body shape, color, fin shape, or genital protrusions. This method is highly subjective, has low accuracy, and is almost unusable for juvenile fish or fish with indistinct sexual characteristics. It is also inefficient and cannot meet the needs of large-scale aquaculture.

[0005] Dissection method: This method involves dissecting the fish to directly observe the gonads. Although accurate, it is a destructive test that causes the live fish to die. It cannot be used in scenarios where it is necessary to preserve the parent fish or precious live fish, resulting in economic losses and waste of resources.

[0006] Endoscopic and other instrument-assisted methods: Although they can achieve a certain degree of live identification, they are complex to operate, require high technical skills, are slow to detect, and may cause stress or physical damage to the fish. They are also not suitable for large-scale, assembly-line operation environments.

[0007] In recent years, with the development of machine vision and deep learning technologies, some studies have attempted to apply image analysis to fish sex identification. For example, cameras are used to capture images of the fish's side or belly, and algorithms are used to analyze texture, color, or shape features. However, these existing technologies mostly remain at the "recognition" stage, usually existing as an independent detection step, failing to form an efficient, closed-loop automated process with subsequent physical sorting. Furthermore, for fish species that rely on surface features and are easily affected by the environment, or for which internal gonadal information is required for accurate identification, the universality and accuracy of a single visual recognition solution are limited.

[0008] Therefore, there is a lack of an integrated automation system in the current technology that can integrate multimodal information acquisition (such as vision and ultrasound), automatic intelligent identification, and real-time linkage with physical sorting execution mechanisms. Such a system aims to achieve unmanned operation of the entire process from fish feeding, feature acquisition, sex determination to automatic sorting, in order to solve the problems of low efficiency, large damage, high subjective error in traditional methods, and the disconnect between the "identification" and "sorting" links in existing automation technologies, thereby providing key technical equipment for the precision and intelligent management of the aquaculture industry. Summary of the Invention

[0009] The purpose of this invention is to address the lack of an integrated automated system in the prior art that can integrate multimodal information acquisition (such as vision and ultrasound), automatic intelligent identification, and real-time linkage with physical sorting execution mechanisms, and proposes an automated system for identifying the sex of fish.

[0010] To achieve the above objectives, the present invention adopts the following technical solution:

[0011] An automated system for identifying the sex of fish includes: a front-end section comprising a host computer, a probe, and an automatic gonad identification system; and a back-end section comprising a sorting conveyor belt. The probe is used to collect fish sex characteristics and transmit the collected images or signals to the host computer. The automatic gonad identification system uses a deep learning model to identify the sex of the collected images and outputs the sex determination result. The host computer controls the sorting conveyor belt to perform sorting operations based on the determination result.

[0012] Preferably, the probe includes a visual probe and / or an ultrasonic probe; the visual probe is used to acquire images of fish body surface features, and the ultrasonic probe is used to acquire images of fish gonads.

[0013] Preferably, the automatic gonad identification system constructs a sex recognition model by training and annotating fish gonad images, which can identify sex characteristics on the body surface or gonad morphological characteristics.

[0014] Preferably, the sex identification model identifies surface features including the distinctive markings or anal fin shape of male fish; and identifies gonadal features including the cystic structure of the ovary or the substantial echogenic features of the testes.

[0015] Preferably, the sorting conveyor belt includes a conveyor belt, a diversion baffle, and multiple collection troughs; the diversion baffle operates according to the control signal of the host computer to sort the fish into the collection troughs corresponding to their sex.

[0016] Preferably, the sorting conveyor belt further includes a re-inspection channel for accommodating unidentified individuals for manual processing.

[0017] Preferably, the host computer is also used to count the number of males and females, the accuracy of identification, and to generate data reports.

[0018] Preferably, the method includes the following steps: S1: placing the fish to be identified on the conveyor belt and stabilizing its posture; S2: acquiring images or signals of the fish's sex characteristics through a probe; S3: the automatic gonad identification system processes and identifies the acquired information and outputs the sex determination result; S4: controlling the sorting conveyor belt to sort the fish to the corresponding collection tank according to the determination result; S5: recording the identification data and generating a statistical report.

[0019] Preferably, in step S2, the acquisition method includes visual acquisition or ultrasonic acquisition; in step S3, the recognition process includes image preprocessing and feature extraction and determination based on a deep learning model.

[0020] Preferably, in step S4, individuals that fail to be identified are guided to the re-inspection channel for manual processing.

[0021] Compared with the prior art, the present invention provides an automated system for identifying the sex of fish, which has the following advantages:

[0022] 1. This invention achieves non-destructive, fully automated assembly line operation: By integrating image acquisition, intelligent recognition, and automatic sorting devices, this invention constructs a complete "detection-judgment-execution" closed-loop system, completing the entire process from fish loading to sex separation without human intervention. This completely overcomes the destructive drawbacks of traditional dissection methods, perfectly protecting live fish, and is especially suitable for sex selection of rare broodstock and high-value fry, while freeing manpower from repetitive and arduous identification work.

[0023] 2. This invention significantly improves identification efficiency and accuracy: The system uses a deep learning-based automatic gonad identification model for determination, enabling the sex determination of a single fish within 1-2 seconds. This processing speed is far superior to manual observation relying on experience, meeting the high-throughput requirements of large-scale aquaculture. The deep learning model is trained on massive amounts of labeled data, ensuring stable and objective identification accuracy, effectively avoiding subjective errors and misjudgments caused by differences in experience and fatigue in manual identification.

[0024] 3. This invention possesses multimodal information fusion and strong generalization capabilities: The system can be configured with both visual and ultrasonic probes, enabling rapid initial screening through surface features and precise acquisition of internal gonadal morphology information through ultrasonic images. This multimodal acquisition strategy greatly expands the range of fish species applicable to the system. Whether it's fish species that rely on secondary sexual characteristics, or juvenile fish or fish species with indistinct sexual characteristics that can only be identified through gonadal structure, this system can effectively handle them all, demonstrating strong generalization capabilities.

[0025] 4. This invention ensures accurate and efficient linkage of sorting operations: The system synchronizes the identification results to the control system of the sorting conveyor belt in real time through the host, realizing seamless connection between information flow and execution flow. The sorting device accurately triggers the baffle action according to the instructions, ensuring that each fish is sorted into the correct collection tank, avoiding the delay and operation error of manual sorting, and truly realizing the "one-click conversion" from identification results to physical separation.

[0026] 5. This invention provides data-driven management and decision support: the system automatically records and statistically analyzes key data such as the number of males and females identified in each batch and the identification accuracy rate, and generates structured reports. This provides an objective and quantitative data foundation for population management, breeding plan formulation, and production efficiency assessment in aquaculture farms, promoting the upgrade of aquaculture from experience-based management to data-driven and precision management. Attached Figure Description

[0027] Figure 1 This is a schematic diagram of the overall structure of the automated system for identifying the sex of fish according to the present invention.

[0028] Figure 2 This is a flowchart illustrating the workflow of the automated system for identifying the sex of fish according to the present invention.

[0029] Figure 3 The flowchart of the automated system for identifying the sex of fish according to the present invention is shown below;

[0030] Figure 4 This is a schematic diagram of the sorting conveyor belt workflow of the automated system for identifying the sex of fish according to the present invention.

[0031] Figure 5 This is a schematic diagram of the system data flow and control signals of the automated system for identifying the sex of fish according to the present invention. Detailed Implementation

[0032] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0033] Example 1: An automated system for identifying the sex of fish.

[0034] refer to Figure 1 (A schematic diagram of the overall system structure, which should clearly show the front-end host, probe, back-end sorting conveyor belt and their connection relationship.) This embodiment provides an automated system for identifying the sex of fish.

[0035] The system mainly consists of a front-end unit 10 and a back-end unit 20.

[0036] The front-end unit 10 includes: a host 11, which serves as the central processing and control unit of the system and can be an industrial control computer or a high-performance embedded system. It is internally equipped with automatic gonad identification system software and a touchscreen display for human-machine interaction. The host 11 is connected to the probe 12 and the control box 21 of the sorting conveyor belt 20 via data cables.

[0037] Probe 12: In this embodiment, to facilitate non-destructive and accurate sex determination of juvenile fish (such as tilapia fry with a body length of 3-5 cm), probe 12 is a high-resolution linear array ultrasonic probe. This probe is suspended directly above the conveyor belt 22 via an adjustable bracket, with its scanning direction perpendicular to the running direction of the conveyor belt 22, ensuring continuous scanning of the fish's abdomen. Probe 12 is connected to the host 11 via a USB 3.0 or Gigabit Ethernet interface, transmitting the acquired ultrasonic image sequence in real time.

[0038] Automatic Gonad Identification System (Software Module): Integrated into host 11. The core of this system is a deep learning model based on a convolutional neural network (CNN). This model is trained using thousands of expert-annotated ultrasound images of fish abdomens containing gonads (ovaries / testes). It can automatically segment the gonad region from the input ultrasound images and extract its morphological features (such as area, perimeter, and echo uniformity) and texture features, ultimately outputting a "female" or "male" determination and its confidence level.

[0039] The backend unit 20 includes:

[0040] The sorting conveyor system consists of a conveyor belt 22, a diversion baffle 23, a female fish collection tank 24, a male fish collection tank 25, and a re-inspection collection tank 26. The conveyor belt 22 is driven by a servo motor, and its speed is adjustable. At the diversion point at the end of the conveyor belt 22, a diversion baffle 23 that can rotate around an axis is provided. Its default position guides the fish into the re-inspection collection tank 26. The rotation of the baffle 23 is driven by a solenoid valve or a small servo motor and controlled by commands issued by the host 11 through the control box 21. The control box 21 contains a PLC or microcontroller, used to receive commands from the host 11 and precisely control the timing of the baffle 23's actions.

[0041] The workflow of this system is as follows (in combination with...) Figure 1 and Figure 2 , Figure 2 (Workflow diagram)

[0042] Feeding and positioning: The operator places the calmed golden pomfret fry to be identified evenly at the beginning of the conveyor belt 22. The conveyor belt 22 runs at a constant speed (e.g., 0.2 m / s), and the fry move stably with their bellies facing upwards within the limiting guide groove.

[0043] Feature acquisition: When the fish fry move to the detection area directly below the ultrasound probe 12, the probe 12 is triggered to quickly scan its abdomen, obtain a clear ultrasound image of the gonad cross section, and upload it to the host 11 in real time.

[0044] Automatic Identification: Host 11 invokes the automatic gonad identification system. The system first preprocesses the received ultrasound image, including Gaussian filtering for noise reduction and contrast enhancement. The preprocessed image is then input into a trained CNN model. The model completes inference within one second and outputs a sex determination result (e.g., "male," with 98% confidence). Simultaneously, host 11 records the fry's unique ID (generated from its position on the conveyor belt and timestamp).

[0045] Sorting and Transfer: The host 11 sends the judgment result and the corresponding fry ID to the control box 21. The control box 21 accurately calculates the moment the fry arrives at the diversion baffle 23 based on the conveyor belt speed and the distance from the detection point to the sorting point. When the fry arrives, the control box 21 drives the baffle 23 to rotate to the corresponding position.

[0046] If the result is "female", turn the baffle to the left and guide the fry into the female fish collection tank 24.

[0047] If the result is "male", turn the baffle to the right and guide the fry into the male fish collection tank 25.

[0048] If the confidence level of the model output is lower than the preset threshold (e.g., 85%), or if the system fails to determine the failure due to image blurring, the host 11 will not send a rotation command, the baffle 23 will remain in the default position, and the fish fry will be introduced into the re-inspection collection tank 26 to await subsequent manual processing.

[0049] Data Recording: Throughout the process, the database of host 11 records the ID, image snapshot, judgment result, confidence level, and sorting destination of each fish in real time. After a batch is completed, the system can automatically generate statistical reports containing the number and ratio of males and females, and the overall recognition accuracy (by comparing with the results of manual verification in the re-inspection tank).

[0050] Example 2: Another specific implementation method

[0051] In another implementation, for larger adult fish with distinct secondary sexual characteristics (e.g., a certain ornamental fish whose sex can be distinguished by the shape of their anal fin), the system can be adjusted as follows:

[0052] Probe 12 configuration: A high-resolution industrial color camera (visual probe) is used instead of or mounted alongside the ultrasonic probe. The camera is equipped with a ring LED light source to eliminate shadows and provide uniform illumination, so as to clearly capture the morphological and color features of the fish's side or anal fin area.

[0053] The model for automatically identifying the gonadal system was developed by replacing the training data with a large number of close-up images of fish bodies from the side or anal fins labeled with male and female tags. The target features learned by the model became the edge contour, aspect ratio, and specific markings of the anal fin.

[0054] Sorting conveyor belt: Due to the large weight of the adult fish, the structure of conveyor belt 22 and baffle 23 needs to be strengthened accordingly, and collection tanks 24, 25 and 26 are also replaced with larger capacity water tanks.

[0055] The workflow of this system is similar to that of Example 1, the difference being that the feature acquisition object is changed to a visual image of the body surface, and the recognition model is a visual recognition model targeting body surface features. This demonstrates the modularity and configurability of the system of the present invention, which can adapt to the identification needs of different fish species by changing different probes and corresponding recognition models.

[0056] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

[0057] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0058] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. An automated system for identifying the sex of fish, characterized in that, include: The front-end includes a host, a probe, and an automatic gonad identification system; the back-end includes a sorting conveyor belt; the probe is used to collect fish sex characteristics information and transmit the collected images or signals to the host; the automatic gonad identification system uses a deep learning model to identify the sex of the collected images and outputs the sex determination result. The host computer controls the sorting conveyor belt to perform sorting operations based on the judgment result.

2. The automated system for identifying the sex of fish according to claim 1, characterized in that, The probe includes a visual probe and / or an ultrasonic probe; the visual probe is used to acquire images of fish body surface features, and the ultrasonic probe is used to acquire images of fish gonads.

3. The automated system for identifying the sex of fish according to claim 2, characterized in that, The automatic gonad recognition system constructs a sex recognition model by training and annotating fish gonad images, and is able to identify sex characteristics on the body surface or gonad morphological characteristics.

4. The automated system for identifying the sex of fish according to claim 3, characterized in that, The sex identification model identifies surface features, including the distinctive markings or anal fin shape of male fish; and identifies gonadal features, including the cystic structure of the ovary or the substantial echogenic features of the testes.

5. The automated system for identifying the sex of fish according to claim 1, characterized in that, The sorting conveyor belt includes a conveyor belt, a diversion baffle, and multiple collection tanks; the diversion baffle operates according to the control signal of the host computer to sort the fish into the collection tanks corresponding to their sex.

6. The automated system for identifying the sex of fish according to claim 5, characterized in that, The sorting conveyor belt also includes a re-inspection channel for accommodating unidentified individuals for manual processing.

7. The automated system for identifying the sex of fish according to claim 1, characterized in that, The host computer is also used to count the number of males and females, identify accuracy, and generate data reports.

8. A method for sexing fish using an automated system as described in any one of claims 1-7, characterized in that, Includes the following steps: S1: Place the fish to be identified on the conveyor belt and stabilize its posture; S2: Collect images or signals of the fish's sex characteristics through the probe; S3: The automatic gonad identification system processes and identifies the collected information and outputs the sex determination result; S4: Control the sorting conveyor belt to sort the fish to the corresponding collection tank according to the determination result. S5: Record identification data and generate statistical reports.

9. The method according to claim 8, characterized in that, In step S2, the acquisition method includes visual acquisition or ultrasonic acquisition; in step S3, the recognition process includes image preprocessing and feature extraction and determination based on a deep learning model.

10. The method according to claim 8, characterized in that, In step S4, individuals that fail to be identified are guided to the re-inspection channel for manual processing.

Images