Fish phenotype intelligent identification device

Abstract

A fish phenotype intelligent identification device disclosed by the present invention comprises a support frame, a clamping frame, a module expansion mechanism, a clamping mechanism, a polarization filter mechanism and a handheld stabilizing mechanism, the clamping frame is fixedly arranged in the middle of the front side of the support frame, cold shoe interfaces are symmetrically formed in the upper and lower parts of the support frame, and chucks are slidably clamped in the cold shoe interfaces. A first screw rod is fixedly arranged in the middle of the chuck, a first threaded sleeve sleeves the first screw rod in a threaded fit mode, a tightening disc is fixedly arranged outside the first threaded sleeve, a light supplementing lamp is fixedly arranged at the upper end of the first screw rod, a clamping mechanism is arranged on the clamping frame, a polarization filter mechanism is arranged in the middle of the other side of the clamping frame, and handheld stabilizing mechanisms are arranged on the two sides of the support frame. After the chuck is clamped into the cold shoe connector through the cold shoe connector, the tightening disc is rotated to drive the first threaded sleeve and the first threaded rod to be in threaded fit to rotate and descend for tightening and fixing, the polarized lens and the mounting ring are inserted into the limiting ring along the limiting groove, and the polarized lens is used for eliminating reflected light on the body surface of an identified fish.

Description

Technical Field

[0001] The present invention relates to the technical field of phenotypic information recognition, and in particular to an intelligent fish phenotypic recognition device. Background Art

[0002] Fish phenotypes refer to observable morphological, physiological, and behavioral characteristics of fish. Intelligent fish phenotype recognition is to use intelligent algorithms and advanced imaging technologies to automatically identify and classify these phenotypic characteristics. In daily life, the true species of fish can be quickly judged by recognizing fish phenotypes; in aquaculture, excellent varieties with fast growth rates and strong disease resistance can be screened by recognizing fish phenotypes; in market supervision, it can be preliminarily judged whether fish belong to natural fishery resources prohibited from fishing by recognizing fish phenotypes. In the utility model with the application number 202322826338.X, a non-contact fish phenotypic information measurement device is disclosed. The image acquisition device and the industrial control computer are integrated into one through a connecting device, which can be carried around by the operator and operated by hand. It has a simple structure and low cost, can meet the usage requirements of different scenarios, and can non-contact, high-throughput, and accurately obtain the phenotypic information of fish;

[0003] However, the above non-contact fish phenotypic information measurement device is integrated, resulting in a low modularity of the device, unable to be freely expanded as needed, and the function of the device is relatively single. When the above non-contact fish phenotypic information measurement device takes pictures, the camera is affected by the mucus, scales on the fish body surface, and water layer reflection, and it is impossible to clearly photograph and identify the fish in the water. At the same time, the above non-contact fish phenotypic information measurement device only measures the phenotypic data of fish and cannot identify the species of fish. Therefore, the present invention proposes an intelligent fish phenotypic recognition device to solve the problems existing in the prior art. Summary of the Invention

[0004] In view of the above problems, the object of the present invention is to propose an intelligent fish phenotypic recognition device. The intelligent fish phenotypic recognition device is symmetrically provided with multiple groups of cold shoe interfaces on the upper and lower sides of the support frame. After the chuck is inserted into the cold shoe interface, the tightening disk is rotated to drive the first threaded sleeve to rotate and descend in threaded cooperation with the first screw rod for tightening and fixing, which is convenient for modular expansion of the fill light as needed to improve the compatibility effect in low-light environments. By inserting the polarizing lens and the mounting ring into the limiting ring along the limiting groove, the reflected light from the fish body surface is eliminated by the polarizing lens, which is convenient for photographing and identifying the fish at the bottom of the water.

[0005] To achieve the object of the present invention, the present invention is realized through the following technical solutions: A fish phenotype intelligent recognition device, comprising a support frame, a clamping frame, a module expansion mechanism, a clamping mechanism, a polarization filter mechanism and a handheld stabilization mechanism. A clamping frame is fixedly provided in the middle of the front side of the support frame. The module expansion mechanism includes a cold shoe interface, a chuck, a first screw, a first threaded sleeve, a tightening disc and a fill light. Cold shoe interfaces are symmetrically opened on the upper and lower sides of the support frame. A chuck is slidably clamped on the cold shoe interface. A first screw is fixedly provided in the middle of the chuck. A first threaded sleeve is threadedly sleeved on the first screw. A tightening disc is fixedly provided outside the first threaded sleeve. A fill light is fixedly provided at the upper end of the first screw. A clamping mechanism is provided on the clamping frame. A polarization filter mechanism is provided in the middle of the other side of the clamping frame. Handheld stabilization mechanisms are provided on both sides of the support frame.

[0006] Further improvement lies in that: There are four groups of cold shoe interfaces, and an internal battery is provided in the fill light.

[0007] Further improvement lies in that: The clamping mechanism includes a chute, clamping blocks, a bidirectional screw, second threaded sleeves and an intelligent recognition main body. A chute is opened in the middle of the clamping frame. Clamping blocks are symmetrically slidably provided in the upper and lower parts of the chute. A bidirectional screw is rotatably provided in the middle of the front side of the clamping blocks. Second threaded sleeves are symmetrically threadedly engaged with the upper and lower parts of the bidirectional screw. The second threaded sleeves are fixedly matched with the front sides of the clamping blocks. An intelligent recognition main body is clamped between the rear sides of the clamping blocks. The intelligent recognition main body is Bluetooth-connected to the handheld stabilization mechanism.

[0008] Further improvement lies in that: The polarization filter mechanism includes a filter holder, a limiting ring, a limiting groove, a mounting ring and a polarization lens. A filter holder is fixedly provided in the middle of the other side of the clamping frame. A limiting ring is fixedly provided at the end of the filter holder. A limiting groove is opened on the limiting ring. A mounting ring is inserted into the limiting groove. A polarization lens is fixedly provided in the mounting ring.

[0009] Further improvement lies in that: The handheld stabilization mechanism includes threaded holes, connecting frames, fixing bolts, quick-release handles and a power supply control mechanism. Threaded holes are symmetrically opened in the middle of both sides of the support frame. Connecting frames are symmetrically provided on both sides of the support frame. Fixing bolts are threadedly engaged between the connecting frames and the support frame. Quick-release handles are fixedly provided on the outer sides of the connecting frames. A power supply control mechanism is provided in the quick-release handles.

[0010] Further improvement lies in that: The power supply control mechanism includes a Bluetooth remote control button, a battery compartment and a battery cover. A Bluetooth remote control button is provided above the quick-release handle. A battery compartment is opened at the bottom inside the quick-release handle. A battery cover is threadedly screwed into the bottom of the battery compartment.

[0011] A further improvement lies in that the Bluetooth remote control button is connected to the intelligent recognition main body via Bluetooth for remote control of photographing, and a battery is provided in the battery compartment for power supply.

[0012] A further improvement lies in that the intelligent recognition main body is provided with a photographing module, an algorithm module, an image segmentation module and a recognition module. The photographing module is used to take fish pictures. The algorithm module is used to extract feature points such as the head, fins and tail of the fish in the pictures. The image segmentation module is used to segment the image into grids according to the feature points and predict the bounding boxes and categories of objects in each grid unit. The recognition module is used to recognize the segmented grid units and give the species results of the target fish in real time.

[0013] A further improvement lies in that the specific steps to obtain the species results of the target fish through the recognition algorithm are as follows:

[0014] S1: Obtain fish picture data from the intelligent recognition main body and then process it;

[0015] S2: Use a convolutional neural network to automatically extract the features of the image through the convolutional layer and add labels to the image;

[0016] S3: Divide the collected image data set into a training set, a validation set and a test set. The training set is used to train the model so that the model learns the relationship between the picture features and the labels. The validation set is used to adjust the parameters of the model during the training process;

[0017] S4: Input the preprocessed image data and the corresponding labels into the selected model for training. After training is completed, integrate the model into the software to recognize the photographed fish pictures.

[0018] The beneficial effects of the present invention are as follows: The present invention symmetrically opens multiple sets of cold shoe interfaces on the upper and lower parts of the bracket frame. After the chuck is inserted into the cold shoe interface, the tightening disc is rotated to drive the first threaded sleeve to rotate downward in threaded cooperation with the first screw rod for tightening and fixing, which is convenient for module expansion of fill lights according to needs to improve the compatibility effect in low-light environments. By inserting the polarizing lens and the mounting ring into the limiting ring along the limiting groove, the polarizing lens is used to eliminate the reflected light on the fish body surface, which is convenient for photographing and identifying fish at the bottom of the water. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is the overall front three-dimensional schematic diagram of the present invention;

[0020] Figure 2 is the front three-dimensional schematic diagram of the bracket frame of the present invention;

[0021] Figure 3 is the overall bottom three-dimensional schematic diagram of the present invention;

[0022] Figure 4Three-dimensional schematic diagram of the quick-release handle of the present invention;

[0023] Figure 5 Three-dimensional schematic diagram of the mounting ring of the present invention;

[0024] Figure 6 Three-dimensional schematic diagram of the limit ring of the present invention;

[0025] Figure 7 System architecture diagram of the intelligent recognition main body of the present invention.

[0026] Wherein: 1. Bracket frame; 2. Clamping frame; 3. Cold shoe interface; 4. Chuck; 5. First screw; 6. First threaded sleeve; 7. Tightening disc; 8. Fill light; 9. Slide groove; 10. Clamping block; 11. Bi-directional screw; 12. Second threaded sleeve; 13. Intelligent recognition main body; 14. Filter holder; 15. Limit ring; 16. Limit groove; 17. Mounting ring; 18. Polarizing lens; 19. Threaded hole; 20. Connecting frame; 21. Fixed bolt; 22. Quick-release handle; 23. Bluetooth remote control button; 24. Battery compartment; 25. Battery cover. Detailed implementation method

[0027] In order to deepen the understanding of the present invention, the present invention will be further described in detail below in conjunction with embodiments. These embodiments are only used to explain the present invention and do not constitute a limitation to the protection scope of the present invention.

[0028] Fish phenotypes refer to various observable characteristics and properties of fish. Fish body shapes include spindle-shaped, such as carp, with a fusiform body, pointed at both ends and wide in the middle, which can reduce the resistance during swimming in water; laterally compressed, like pomfret, with a laterally compressed body, usually suitable for quickly swimming in the middle and upper water layers; flattened, such as stingray, with a flat body, mostly inhabiting the bottom; club-shaped, for example eel, with a slender club-shaped body, good at shuttling in caves or reef crevices.

[0029] Based on this, according to Figure 1 、 Figure 2 、 Figure 3 、 Figure 4 、 Figure 5 、 Figure 6 、 Figure 7As shown in the figure, this embodiment provides a fish phenotype intelligent recognition device, including a support frame 1, a clamping frame 2, a module expansion mechanism, a clamping mechanism, a polarization filter mechanism, and a handheld stabilization mechanism. A clamping frame 2 is fixedly arranged in the middle of the front side of the support frame 1. The module expansion mechanism includes a cold shoe interface 3, a chuck 4, a first screw 5, a first threaded sleeve 6, a tightening disc 7, and a fill light 8. Cold shoe interfaces 3 are symmetrically arranged up and down on the support frame 1. A chuck 4 is slidably clamped into the cold shoe interface 3. A first screw 5 is fixedly arranged in the middle of the chuck 4. A first threaded sleeve 6 is threadedly sleeved on the first screw 5. A tightening disc 7 is fixedly arranged outside the first threaded sleeve 6. A fill light 8 is fixedly arranged at the upper end of the first screw 5. There are four groups of cold shoe interfaces 3. An internal battery is arranged in the fill light 8. When taking pictures and sampling in a low-light environment, the chuck 4 is clamped into the cold shoe interface 3 as needed. The tightening disc 7 is rotated to drive the first threaded sleeve 6 to move downward in threaded cooperation with the first screw 5. The fill light 8 is tightened and fixed on the support frame 1 by the cooperation of the tightening disc 7 and the chuck 4. It is convenient to expand and fix the device through the cold shoe interface 3 and facilitate free expansion, solving the problem of the relatively single function of the device. A clamping mechanism is arranged on the clamping frame 2. A polarization filter mechanism is arranged in the middle of the other side of the clamping frame 2. Handheld stabilization mechanisms are arranged on both sides of the support frame 1.

[0030] The clamping mechanism includes a chute 9, clamping blocks 10, a bidirectional screw 11, second threaded sleeves 12, and an intelligent recognition main body 13. A chute 9 is opened in the middle of the clamping frame 2. Clamping blocks 10 are symmetrically and slidably arranged up and down in the chute 9. A bidirectional screw 11 is rotatably arranged in the middle of the front side of the clamping blocks 10. Second threaded sleeves 12 are symmetrically and threadedly matched with the bidirectional screw 11 up and down. The second threaded sleeves 12 are fixedly matched with the front side of the clamping blocks 10. An intelligent recognition main body 13 is clamped between the rear sides of the clamping blocks 10. The intelligent recognition main body 13 is connected to the handheld stabilization mechanism via Bluetooth. Before recognition, the intelligent recognition main body 13 is placed between the clamping blocks 10. The bidirectional screw 11 is rotated to drive the second threaded sleeves 12 to move towards each other by threaded cooperation, so that the clamping blocks 10 move towards each other to clamp the intelligent recognition main body 13. The specific data of the intelligent recognition main body are: a 6.8-inch OLED multi-touch screen, a resolution of 3840x1644px, a touch sampling rate of 240Hz, a 12.2-megapixel camera, an aperture of F / 2.2, and a battery capacity of 4600mAh.

[0031] The polarization filter mechanism includes a filter holder 14, a limit ring 15, a limit groove 16, a mounting ring 17, and a polarization lens 18. In the middle of the other side of the clamping holder 2, a filter holder 14 is fixedly provided. At the end of the filter holder 14, a limit ring 15 is fixedly provided. A limit groove 16 is opened on the limit ring 15. An mounting ring 17 is inserted into the limit groove 16. A polarization lens 18 is fixedly provided inside the mounting ring 17. Before taking pictures of fish in water, the polarization lens 18 is inserted into the limit ring 15 and fixed by the mounting ring 17 inserted into the limit groove 16. The polarization lens 18 effectively weakens and eliminates the reflected light on the fish body surface, making the field of view clear and natural, effectively eliminating the interference of polarized light on the recognition result. At the same time, it is convenient to rotate the outer ring of the polarizer according to the shooting angle to adjust the angle between the polarizer and the light source to obtain the maximum polarization effect.

[0032] The handheld stabilization mechanism includes threaded holes 19, connecting brackets 20, fixing bolts 21, quick-release handles 22, and a power supply control mechanism. Threaded holes 19 are symmetrically opened in the middle of both sides of the bracket frame 1. Connecting brackets 20 are symmetrically provided on both sides of the bracket frame 1. Fixing bolts 21 are threadedly engaged between the connecting brackets 20 and the bracket frame 1. Quick-release handles 22 are fixedly provided on the outside of the connecting brackets 20. A power supply control mechanism is provided inside the quick-release handles 22. Before using the device, the connecting brackets 20 and the quick-release handles 22 are fixed on both sides of the bracket frame 1 by the fixing bolts 21. When taking pictures and sampling, hold the quick-release handles 22 with both hands to keep the bracket frame 1 stable.

[0033] The power supply control mechanism includes a Bluetooth remote control button 23, a battery compartment 24, and a battery cover 25. A Bluetooth remote control button 23 is provided above the quick-release handle 22. A battery compartment 24 is opened at the bottom inside the quick-release handle 22. A battery cover 25 is screwed into the bottom of the battery compartment 24. The Bluetooth remote control button 23 is Bluetooth-connected to the intelligent recognition main body 13 for remote control of taking pictures. The battery compartment 24 is provided with battery power supply. When in use, after the Bluetooth remote control button 23 is Bluetooth-connected to the intelligent recognition main body 13, press the Bluetooth remote control button 23 to control the intelligent recognition main body 13 to take pictures. Bluetooth connection is realized by the Bluetooth chip inside the intelligent recognition main body 13, and battery power supply is provided by putting a battery into the battery compartment 24.

[0034] The intelligent recognition main body 13 is provided with a photographing module, an algorithm module, an image segmentation module, and an identification module. The photographing module is used to take pictures of fish to obtain pictures. The algorithm module is used to extract feature points such as the head, fins, and tail of the fish pictures. The image segmentation module is used to segment the image into networks based on the feature points and predict the bounding boxes and categories of objects in each network unit. The identification module is used to identify the segmented grid units and give the species results of the target fish in real time. The intelligent main body and the power supply mechanism can be independently updated and iterated. After iteration, the original clamping mechanism, filter mechanism, and stabilization mechanism can be directly applied without re-opening the mold to develop new hardware, which can meet the requirements of rapid and flexible equipment upgrading.

[0035] The specific steps to obtain the target fish species result through the recognition algorithm are as follows:

[0036] S1: Obtain fish picture data from the intelligent recognition entity and then process it;

[0037] S2: Use a convolutional neural network to automatically extract the features of the image through the convolutional layer and add labels to the image;

[0038] S3: Divide the collected image dataset into a training set, a validation set, and a test set. The training set is used to train the model so that the model learns the relationship between the picture features and the labels. The validation set is used to adjust the parameters of the model during the training process;

[0039] The model algorithm constructed by the present invention divides the input image into multiple grids, and predicts multiple bounding boxes and class probabilities on each grid. The algorithm includes multiple convolutional layers, fully connected layers, and a specific loss function. Each convolutional layer gradually extracts higher-level features and uses these features to predict the position and class of the target;

[0040] S4: Input the preprocessed image data and the corresponding labels into the selected model for training. After the training is completed, integrate the model into the device to identify the photographed fish photos.

[0041] When the intelligent fish phenotype recognition device is used, the chuck is clamped into the cold shoe interface according to the environmental brightness requirement, and the rotation tightening disc drives the first threaded sleeve to cooperate with the first screw rod to descend and cooperate with the chuck to clamp on the support frame. It is convenient to fix the fill light on the support frame to improve the brightness in the low-light environment. When the fixing bolt is screwed into the threaded hole, the connecting frame and the quick-release handle are fixed on both sides of the support frame. After the intelligent recognition entity is placed between the clamping blocks, rotate the bidirectional screw rod, and the threaded cooperation drives the second threaded sleeve and the clamping blocks to move towards each other along the chute and the bidirectional screw rod to clamp the intelligent recognition entity. After the Bluetooth remote control button is Bluetooth-connected to the intelligent recognition entity, press the Bluetooth remote control button to take a photo. Drive the polarizing lens along the limiting groove and insert it into the limiting ring, and rotate the polarizing lens to obtain the maximum polarization effect, eliminate the reflection interference caused by the water surface, and obtain clear and usable high-quality image data. The algorithm automatically performs real-time multi-object detection and displays the calculation results and confidence levels of each target currently.

[0042] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification only illustrates the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A fish phenotype intelligent identification device, characterized in that: The invention comprises a support frame (1), a clamping frame (2), a module expansion mechanism, a clamping mechanism, a polarization filter mechanism and a handheld stabilization mechanism. The support frame (1) is fixedly provided with a clamping frame (2) in the middle of the front side. The module expansion mechanism comprises a cold shoe interface (3), a chuck (4), a first screw (5), a first threaded sleeve (6), a tightening disk (7) and a fill light (8). The support frame (1) is symmetrically provided with a cold shoe interface (3) in the upper and lower parts. A chuck (4) is slidably inserted into the cold shoe interface (3). A first screw (5) is fixedly provided in the middle of the chuck (4). A first threaded sleeve (6) is threadedly fitted on the first screw (5). A tightening disk (7) is fixedly provided outside the first threaded sleeve (6). A fill light (8) is fixedly provided at the upper end of the first screw (5). The clamping mechanism is provided on the clamping frame (2). A polarization filter mechanism is provided in the middle of the other side of the clamping frame (2). Handheld stabilization mechanisms are provided on both sides of the support frame (1).

2. The fish phenotype intelligent identification device according to claim 1, characterized in that: The cold shoe interface (3) is provided with four groups, and the fill light (8) is provided with a built-in battery.

3. The fish phenotype intelligent identification device according to claim 1, characterized in that: The clamping mechanism comprises a slide groove (9), a clamping block (10), a bidirectional screw (11), a second threaded sleeve (12) and an intelligent identification body (13); a slide groove (9) is provided in the middle of the clamping frame (2); a clamping block (10) is symmetrically slidably provided in the slide groove (9) up and down; a bidirectional screw (11) is rotatably provided in the middle of the front side of the clamping block (10); the bidirectional screw (11) is symmetrically threadedly provided with a second threaded sleeve (12) up and down; the second threaded sleeve (12) is matched and fixed to the front side of the clamping block (10); an intelligent identification body (13) is clamped between the rear sides of the clamping block (10); and the intelligent identification body (13) is connected to the handheld stabilization mechanism via Bluetooth.

4. The fish phenotype intelligent identification device according to claim 1, characterized in that: The polarizing filter mechanism comprises a filter holder (14), a limiting ring (15), a limiting groove (16), a mounting ring (17) and a polarizing lens (18); the filter holder (14) is fixedly provided in the middle of the other side of the clamping frame (2); a limiting ring (15) is fixedly provided at the end of the filter holder (14); a limiting groove (16) is provided on the limiting ring (15); a mounting ring (17) is inserted into the limiting groove (16); and a polarizing lens (18) is fixedly provided in the mounting ring (17).

5. The fish phenotype intelligent identification device according to claim 3, characterized in that: The handheld stabilizing mechanism comprises a threaded hole (19), a connecting frame (20), a fixing bolt (21), a quick-release handle (22) and a power supply control mechanism. The threaded holes (19) are symmetrically opened in the middle of both sides of the support frame (1). The connecting frames (20) are symmetrically arranged on both sides of the support frame (1). The fixing bolts (21) are threadedly matched between the connecting frames (20) and the support frame (1). The quick-release handle (22) is fixedly arranged on the outside of the connecting frame (20), and the power supply control mechanism is arranged inside the quick-release handle (22).

6. The fish phenotype intelligent identification device according to claim 5, characterized in that: The power supply control mechanism comprises a Bluetooth remote control button (23), a battery compartment (24) and a battery cover (25); the Bluetooth remote control button (23) is arranged above the quick release handle (22); the battery compartment (24) is arranged at the bottom of the quick release handle (22); and the battery cover (25) is screwed into the bottom of the battery compartment (24).

7. The intelligent fish phenotype identification device according to claim 6, characterized in that: The Bluetooth remote control button (23) is connected to the intelligent recognition body (13) via Bluetooth for remote control photography, and a battery is provided in the battery compartment (24) for power supply.

8. The fish phenotype intelligent identification device according to claim 3, characterized in that: The intelligent recognition body (13) is provided with a photographing module, an algorithm module, an image segmentation module and a recognition module. The photographing module is used to photograph and obtain fish pictures. The algorithm module is used to utilize feature points such as the head, fins, and tail extracted from the fish pictures. The image segmentation module is used to segment the image into networks based on the feature points and predict the boundary box and category of the object in each network unit. The recognition module is used to identify the segmented grid units using a recognition algorithm and provide the type result of the target fish in real time.

9. The fish phenotype intelligent identification device according to claim 1, characterized in that: The specific steps to obtain the target fish species result through the recognition algorithm are: S1: Obtain fish image data from the intelligent recognition subject and then process it; S2: Use convolutional neural networks to automatically extract image features and add labels to images through convolutional layers; S3: Divide the collected image dataset into training set, validation set and test set. The training set is used to train the model so that the model can learn the relationship between image features and labels. The validation set is used to adjust the parameters of the model during the training process. S4: The preprocessed image data and the corresponding labels are input into the selected model for training. After the training is completed, the model is integrated into the software to identify the captured fish photos.

Images