Automatic squid feeding device based on vision intelligence and squid recognition method

Abstract

According to the intelligent vision-based automatic squid feeding device and the squid recognition method, multiple squids can enter a special channel in a scattered and orderly manner, the squids are prevented from being stacked and stopped in the channel, and it is guaranteed that the squids slide into the conveying belt in sequence. Meanwhile, the squid conveying interval is controlled according to the running speed and start-stop time of the conveying belt, squid postures and positions are recognized according to images, and squid data collection is completed in combination with visual intelligence and deep learning. And the squids are accurately grabbed according to the collected data, aerial posture adjustment is completed, squid carcasses are grabbed, and the squids are accurately placed according to the lengths of the squid carcasses. In conclusion, the characteristics of the squids are fully considered, multiple squids can be screened, automatically recognized, fed and accurately placed, the production efficiency is effectively improved, and the labor cost is reduced.

Description

Technical Field

[0001] The present application relates to the technical field of seafood processing, and in particular to an automatic squid feeding device and a squid identification method based on visual intelligence. Background Art

[0002] Squid is a seafood that is loved by consumers. Its meat is delicious and juicy. With the improvement of living quality, consumers have higher and higher requirements for squid. For the sake of appearance and convenience, the squid head needs to be peeled off, ears and internal organs removed, so that the cooked squid is cleaner and easier to eat, which improves the appearance and texture of the dish.

[0003] At present, most squid processing still relies on manual operation, and the proficiency of squid workers varies. Grasping the squid too tightly can easily damage the squid's skin, while grasping it too loosely can cause the squid to slip, requiring multiple grasping, which can easily damage the squid. Frequent manual grasping increases the labor intensity of workers. A small part of squid processing has also achieved automatic operation of the equipment, but these equipment still require manual screening of squids one by one, accurately controlling the position and angle of the squid, and then placing them on the designated loading station for subsequent processing. This loading method has low production efficiency and insufficient process automation, which greatly increases labor costs and work intensity. Summary of the invention

[0004] In order to solve the shortcomings of the prior art, the present invention provides a squid automatic feeding device and a squid identification method based on visual intelligence, which can realize the automatic separation of multiple squids. Based on intelligent visual recognition and deep learning, the squid-specific gripper mechanism of the manipulator can realize precise gripping and adjust the squid posture in the air, thereby achieving precise placement.

[0005] On the one hand, the present invention provides a squid automatic feeding device based on visual intelligence, including a vibrating screen mechanism, a conveying mechanism, a grasping and feeding mechanism, and an image acquisition mechanism.

[0006] The vibrating screen mechanism is arranged at the front end entrance of the conveying mechanism. After the squids are vibrated by the vibrating screen mechanism, they are dispersed and slide into the conveying mechanism. A squid-specific gripper assembly is provided at the top of the grabbing and feeding mechanism. The grabbing and feeding mechanism is arranged on one side of the conveying mechanism, and the squid-specific gripper assembly is suspended above the conveying mechanism. The image acquisition mechanism is arranged on the front side and directly opposite to the grabbing and feeding mechanism.

[0007] Preferably, the vibrating screen mechanism comprises a guide plate and a vibrating disk, and the guide plate is mounted on the vibrating disk to form multiple channels.

[0008] Furthermore, the multi-channel entrances are all in an eight-shaped shape, and behind the eight-shaped shape are parallel channels. There is no straight section between each channel entrance and adjacent entrances, so as to prevent the squid from being stranded in the straight section.

[0009] Preferably, the vibrating screen mechanism further comprises an angle adjustment component and / or a brush component. The angle adjustment component changes the angle with the horizontal plane by moving the telescopic rod up and down, thereby adjusting the sliding speed of the squid and preventing the squid from overlapping again; the brush component is located at the outlet of the vibrating screen mechanism to prevent the squid from stagnating in a certain place.

[0010] Preferably, the conveying mechanism is designed with a brush assembly at the inlet of the conveyor belt to adjust the squid's lying posture and prevent the grasping part from being blocked.

[0011] Furthermore, the rotation direction of the roller brush of the brush assembly is opposite to the movement direction of the conveyor belt. By adjusting the movable plate up and down, the distance between the roller brush and the conveyor belt is made to be the thickness of a single squid carcass, so as to clean the obstructed part of the squid and the overlapping phenomenon of multiple squids.

[0012] Preferably, the conveying mechanism arranges the squids in compartments and controls the squid conveying intervals according to controlling the running speed and start and stop time of the conveyor belt.

[0013] Preferably, the image acquisition mechanism on the front side of the grabbing and feeding mechanism is used to accurately capture the squid image. After identifying whether the squid carcass is defective based on the image, the squid-specific gripper assembly selectively does not grab the defective squid, and the defective squid is sent to centralized processing via a conveyor belt. The image acquisition mechanism directly opposite the grabbing and feeding mechanism is used to reconfirm the squid's aerial posture after grabbing, and complete the subsequent precise adjustment of the squid's posture.

[0014] On the other hand, the present invention also provides a squid identification method using the squid automatic feeding device based on visual intelligence as described above, comprising the following steps:

[0015] Step 1: Read the camera RTSP stream, perform binarization on the squid image, remove clutter, reduce tentacles, extract the squid outline, obtain the minimum circumscribed rectangle of the squid, and calculate the length and angle of the squid.

[0016] Step 2: Read the camera RTSP stream, save the image, create squid fin tags, install and configure YOLOv5 parameters, train the fin recognition model, and put the trained fin recognition model parameters into the weights directory.

[0017] Step 3: Identify the acquired camera RTSP stream, output the box position of the squid fin, and calculate the center point of the fin based on the squid length and angle.

[0018] Further, the step 1 specifically includes:

[0019] Step 1-1: Read the camera RTSP stream;

[0020] Step 1-2: Binarize the squid image and remove the conveyor belt image;

[0021] Step 1-3: Use corrosion and expansion algorithms to remove clutter and reduce tentacles;

[0022] Step 1-4: Extract the squid outline;

[0023] Step 1-5: Obtain the minimum circumscribed matrix of the squid image contour point set;

[0024] Step 1-6: Calculate the length and angle of the circumscribed matrix based on the squid's slender features.

[0025] The beneficial effects of the present invention are:

[0026] The vibrating screen mechanism provided by the present invention can realize the dispersion of multiple squids and orderly enter the dedicated channel. At the same time, a brush assembly is arranged to prevent the squids from overlapping and stagnating in the channel, and ensure that the squids slide into the conveyor belt in sequence; the conveying mechanism controls the squid conveying interval according to the running speed and start and stop time of the conveyor belt, and a brush assembly is arranged at the front end of the conveyor belt to straighten the curled squids that slide off the vibrating screen and prevent the squids from overlapping again; the image acquisition mechanism recognizes the posture and position of the squid according to the image, and completes the squid data acquisition in combination with visual intelligence and deep learning; the grabbing and feeding mechanism accurately grabs the squid according to the collected data, completes the mid-air posture adjustment, grabs the squid carcass, and accurately places the squid carcass according to the length of the squid carcass.

[0027] The squid identification method provided by the present invention can automatically identify squids and obtain squid parameters. The grabbing and feeding mechanism grabs and feeds the squids in combination with the obtained parameters, thus saving manual screening time and improving squid feeding and production efficiency.

[0028] The present invention fully considers the characteristics of squids, realizes the screening of multiple squids, automatic identification and feeding, and accurate positioning of squids, effectively improves production efficiency and reduces labor costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] The implementation steps of the present invention are further described below with reference to the accompanying drawings and specific examples:

[0030] Figure 1 This is a structural diagram of a squid automatic feeding device based on visual intelligence according to an embodiment of the present application;

[0031] Figure 2 This is a structural diagram of a vibrating screen mechanism in a squid automatic feeding device based on visual intelligence according to an embodiment of the present application;

[0032] Figure 3 This is a structural diagram of a conveying mechanism in a squid automatic feeding device based on visual intelligence according to an embodiment of the present application;

[0033] Figure 4This is a structural diagram of an image acquisition mechanism in a squid automatic feeding device based on visual intelligence according to an embodiment of the present application;

[0034] Figure 5 This is a structural diagram of a grabbing and feeding mechanism in a squid automatic feeding device based on visual intelligence according to an embodiment of the present application;

[0035] Figure 6 This is a structural diagram of a squid-specific gripper assembly in a squid automatic feeding device based on visual intelligence according to an embodiment of the present application;

[0036] Figure 7 The squid image obtained in the embodiment of the present application;

[0037] Figure 8 This is a flowchart of step 1 in a squid identification method based on visual intelligence according to an embodiment of the present application;

[0038] Fig. 9 This is a flowchart of steps 2 and 3 in a squid identification method based on visual intelligence in an embodiment of the present application. DETAILED DESCRIPTION

[0039] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0040] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] like Figure 1 As shown, an embodiment of the present application provides a squid automatic feeding device based on visual intelligence, which mainly includes a vibrating screen mechanism 1, an image acquisition mechanism 2, a conveying mechanism 3, and a grabbing and feeding mechanism 4.

[0042] The vibrating screen mechanism 1 is arranged at the front end entrance of the conveying mechanism 3. After the squids are vibrated by the vibrating screen mechanism 1, they are dispersed and slide into the conveying mechanism 3; a squid-specific gripper assembly is provided at the top of the grabbing and feeding mechanism 4, and the grabbing and feeding mechanism 4 is arranged on one side of the conveying mechanism 3, and the squid-specific gripper assembly is suspended above the conveying mechanism 3; the image acquisition mechanism 2 is arranged on the front side and directly opposite to the grabbing and feeding mechanism 4.

[0043] In a preferred embodiment, a set of image acquisition mechanism 2 is arranged on the front side of the grabbing and feeding mechanism, 30 cm away from the conveyor belt brush outlet, to realize three-dimensional intelligent recognition of squids. The squids are clamped by a special squid gripper assembly, and a set of image acquisition mechanism 2 is arranged opposite the robot arm to accurately calibrate the aerial posture, adjust the aerial posture to the required posture, and place the squid carcass precisely at the designated work station according to its size.

[0044] In a preferred embodiment, the vibration screen mechanism 1 includes a vibration screen component 10 , an elastic component 11 , an angle adjustment component 12 , a brush component 13 , a frame 14 and a vibration machine component 15 .

[0045] like Figure 2 As shown, the vibrating screen assembly 10 is composed of a guide plate 101, a vibrating plate 102, a left baffle 103, a right baffle 104, and a rear baffle 105. The guide plate 101 is installed on the vibrating plate 102 to form three squid channels, and roller brushes are arranged above the three channels. The left baffle 103, the right baffle 104, and the rear baffle 105 are all installed on the vibrating plate 102 to prevent the squid from falling off to the surroundings during vibration.

[0046] The elastic component 11 is composed of a rotating shaft 111, a rotating shaft seat 112, a spring seat 113, and a spring 114. The rotating shaft 111 is connected to the left baffle 103 and the right baffle 104 respectively, and a rotating shaft 111 is installed at the front and rear of each baffle, and a rotating shaft seat 112 is installed on each rotating shaft, and the rotating shaft seat 112 rotates around the center of the rotating shaft 111. The spring 114 is provided with spring seats 113 at the top and bottom, and the upper spring seats are connected to the rotating shaft seat 112, the front lower spring seat is connected to the frame 14, and the rear lower spring seat is connected to the telescopic rod 121.

[0047] The angle adjustment assembly 12 is composed of a telescopic rod 121, a set screw 122, a fixed sleeve 123, and an adjusting bolt 124. The telescopic rod 121 is installed in the fixed sleeve 123, and the fixed sleeve 123 is connected to the frame 14. The telescopic rod 121 can be adjusted in and out by turning the adjusting bolt 124. After the position adjustment is completed, the set screw 122 needs to be tightened to fix it.

[0048] The brush assembly 13 is composed of a bearing seat bottom plate 131, a bearing seat 132, an adjusting bolt 133, a roller brush 134, a coupling 135, a motor seat 136, and a servo motor 137. The bearing seat 132 is fixed on the bearing seat bottom plate 131 and the motor seat 136, and is respectively installed at both ends 134 of the roller brush. The servo motor 137 is installed on the motor seat 136, and the coupling 135 is installed between the roller brush 134 and the servo motor 137. The bearing seat bottom plate 131, the motor seat 136 and the frame 14 are connected by the adjusting bolt 133.

[0049] The vibrator assembly 15 is composed of a vibrator mounting plate 151 and a vibrator 152 . The vibrator mounting plate 151 is connected to the left baffle 103 and the right baffle 104 , and the vibrator 152 is fixed on the vibrator mounting plate 151 .

[0050] The squid is poured into the vibration plate 102, and after being vibrated by the vibrator 152, the squid is dispersed and slides into the three channels under the action of the spring 114. The entrances of the three channels are all in an eight-shaped shape, and the eight-shaped shape is followed by parallel channels. The channel width is adjustable within a range of 3 cm to 5 cm. The corresponding width is set according to the width of the squid, and the channel width is consistent with the width of the squid carcass. There is no straight line section between the entrance of each channel and the adjacent entrance to prevent the squid from being shelved in the straight line section.

[0051] The squid's sliding speed can be achieved by adjusting the angle of the vibration plate 102. The best angle between the vibration plate 102 and the horizontal plane is set at 10° to 15°. If the angle is too small, it is not conducive to the squid's sliding. If the angle is too large, the squid is easy to overlap again. The angle of the vibration plate 102 can be adjusted by adjusting the bolt 124. The angle between the vibration plate 102 and the horizontal plane can be changed by moving the telescopic rod 121 up and down. The sliding speed is adjusted to the best angle according to actual needs.

[0052] Controlling the rotation speed of the roller brush 134 and the distance between the roller brush 134 and the vibration plate 102 can effectively ensure that the squids will not overlap or stagnate at a certain place. By adjusting the rotation speed of the roller brush 134, the squids can be controlled to have a certain initial velocity when sliding onto the conveyor belt, ensuring that they slide a certain distance on the conveyor.

[0053] The distance between the roller brush 134 and the vibration plate 102 can be increased or decreased by adjusting the bolt 133, and the rotation speed of the roller brush 134 can be controlled by the servo motor 137. Under the dual effects of the vibrator 152 and the roller brush 134, the squids slide onto the conveyor belt in an orderly manner.

[0054] like Figure 3 As shown, the conveying mechanism 3 is composed of a motor 301, a torque arm 302, a driving roller 303, a bearing seat 304, a conveyor belt 305, a roller brush motor 306, a motor seat 307, a baffle 308, a roller brush 309, a roller brush bearing seat 310, a driven roller 311, a movable plate 312 and a frame 313.

[0055] The bearing seat 304 is connected to the frame 313, the two ends of the active roller 303 are respectively connected to the bearing seat 304, the motor 301 is connected to the active roller 303, one end of the torque arm 302 is connected to the motor 301, and the other end is connected to the frame 313. The motor seat 307 and the movable plate 312 are both connected to the frame 313, and can be adjusted up and down. The roller brush bearing seat 310 is fixed on the motor seat 307 and the movable plate 312 respectively, the roller brush motor 306 is fixed on the motor seat 307, the roller brush motor 306 is connected to the roller brush 309 using a coupling, and baffles 308 are arranged on both sides of the roller brush. One end of the conveyor belt 305 wraps the driven roller 311, and the other end wraps the active roller 303, and the conveyor belt 305 is tensioned by moving the driven roller 311.

[0056] When the squid slides from the vibrating screen mechanism 1 to the conveying mechanism 3, the squid may be in a curled-up state or the tail of the squid may be blocked, which is not conducive to subsequent grasping and intelligent identification. Therefore, a brush assembly is designed at the conveyor belt inlet to effectively adjust the squid's lying posture and prevent the grasping part from being blocked. The direction of rotation of the roller brush 309 is opposite to the direction of movement of the conveyor belt, and the speed is relatively slow. A harder specification nylon wire is selected. By adjusting the moving plate 312 up and down, the distance between the roller brush and the conveyor belt is the thickness of a single squid carcass, which can clean the obstructed part of the squid and the stacking phenomenon of multiple squids. Control the running speed of the conveyor belt 305 (i.e., control the speed of the motor 301) and the start and stop time, and the squid can be arranged in compartments and the squid conveying interval can be controlled. The distance between the roller brush 309 and the conveyor belt 305 is set to be adjustable, and the distance between the roller brush 309 and the conveyor belt 305 can be fine-tuned by adjusting the motor seat 307 and the moving plate 312.

[0057] like Figure 4 As shown, the image acquisition mechanism 2 is composed of a bracket fixing frame 201, a fixing plate 202, a rotating shaft plate 203, a camera 204 and a sliding block 205.

[0058] The support fixing frame 201 is connected to the frame 313, and the fixing plate 202 is installed on the support fixing frame 201. One end of the rotating shaft plate 203 is connected to the fixing plate 202, and the other end is connected to the sliding block 205, and both ends can rotate. The camera 204 is connected to the sliding block 205, and the camera 204 can move forward and backward along the direction of the sliding block 205. By rotating the fixing plate 202, the sliding block 205 and moving the position of the camera 204, the camera 204 is adjusted to a suitable angle, and the squid image can be accurately captured. The image acquisition mechanism 2 is arranged in a left-right symmetrical manner, and the three-dimensional image and position of the squid can be accurately captured in all directions. The length of the squid carcass can be identified according to the image, so that the grasping and feeding mechanism 4 can accurately grasp the squid and accurately place it at the back-end equipment station according to the length of the squid carcass; the camera 204 is arranged directly opposite the grasping and feeding mechanism 4 to confirm the squid's aerial posture again after grasping, so as to facilitate the subsequent precise adjustment of the squid posture. The image acquisition mechanism 2 can identify whether the squid carcass is defective based on the image, and the grabbing and feeding mechanism 4 selectively does not grab it, and sends it to a certain point for centralized processing through the conveying mechanism 3.

[0059] like Figure 5 As shown, the grabbing and feeding mechanism 4 is composed of a mechanical arm 41 and a squid-specific gripper assembly 42.

[0060] The robotic arm 41 can complete the control of any position in the space, and the squid-specific gripper assembly 42 accurately grasps the tip of the squid's tail. According to the intelligent image recognition, the required posture is adjusted in the air and then the carcass is accurately grasped to achieve the precise placement of the squid.

[0061] like Figure 6 As shown, the squid-specific gripper assembly 42 is composed of a fixed rotating plate 421, a carcass gripper 422, a tail tip gripper 423, a push-flattening block 424, a push-flattening cylinder 425, a connecting block 426, a downward pressing cylinder 427 and a servo motor 428.

[0062] The fixed rotating plate 421 is connected to the rotating disk of the mechanical arm 41, and the carcass gripper 422, the tail tip gripper 423 and the downward pressure cylinder 427 are respectively connected to the fixed rotating plate 421. The push-flattening block 424 is fixed to the push-flattening cylinder 425 and is pushed along with the push-flattening cylinder 425. The downward pressure cylinder 427 is fixed to the fixed rotating plate 421, and the push-flattening cylinder 425 and the downward pressure cylinder 427 are fixed by the connecting block 426. The servo motor 428 is installed on the tail tip gripper 423, which will drive the tail tip gripper 423 to rotate.

[0063] After the squid is image-recognized by the image acquisition mechanism on the front side of the mechanical claw, the conveyor belt stops within the grabbing range of the grabbing and feeding mechanism 4, and the grabbing and feeding mechanism 4 starts to grab the squid. The image recognizes the tail tip of the squid and the length of the carcass, and the tail tip gripper 423 starts to grab the tail tip of the squid. The tail tip gripper 423 is in a pointed mouth shape with a width of 2-3mm to prevent the gripper from being too wide and the squid from being easy to droop. The mechanical arm 41 lifts the squid, and the squid carcass will face vertically downward. The camera 204 is arranged opposite the grabbing and feeding mechanism 4 to identify the posture of the squid in the air. The servo motor 428 rotates one circle to identify the position of the squid tail tip. According to the position of the tail tip, the servo motor 428 rotates a corresponding angle so that the back of the squid carcass faces the carcass gripper 422, and the carcass gripper 422 is flat and in an open state, with an opening width greater than 5cm. At this time, the downward pressure cylinder 427 drives the flattening cylinder 425 to move downward, so that the flattening block 424 is slightly lower than the tail tip gripper 423, and the flattening cylinder 425 pushes the flattening block 424 to make the squid carcass lie flat on the flattening block in a horizontal state. In particular, the initial contact surface between the flattening block 424 and the squid is a large arc transition to prevent damage to the squid carcass. The ribs are set on both sides of the flattening block 424 to form a squid pushing channel, which guides the squid when it is pushed to prevent the squid carcass from bending or tilting. The carcass gripper 422 grabs the squid carcass part, and the gripper gap needs to reserve 2-3mm to prevent the gripper from damaging the squid carcass. After the squid is grabbed, the flattening cylinder 425 and the downward pressure cylinder 427 are returned in turn, and finally accurately placed at the loading station according to the length of the squid carcass.

[0064] like Figure 8-9 As shown, the embodiment of the present application also provides a squid identification method based on visual intelligence using the squid automatic feeding device based on visual intelligence as described above, for obtaining squid parameters. Figure 7 As shown, the squid parameters are obtained including the following steps on the conveyor belt 305: Figure 7 The fin 501, body 502 and tentacle 503 of the squid shown are identified, the length L504 of the squid and the angle θ505 with the moving direction of the conveyor belt 305 are calculated, the distance X506 and Y507 from the fin 501 to the edge of the image are calculated, and then the parameters are sent to the grabbing and feeding mechanism 4.

[0065] In an embodiment of the present application, the squid identification method based on visual intelligence specifically comprises the following steps:

[0066] Step 1: Read the camera RTSP stream, perform binarization on the squid image, remove clutter, reduce tentacles, extract the squid outline, obtain the minimum circumscribed rectangle of the squid, and calculate the length and angle of the squid.

[0067] Step 2: Use the VideoCapture function of opencv to read the camera RTSP stream, save the image, create squid fin labels based on LabelImg, install and configure YOLOv5 parameters, train the fin recognition model, and put the trained fin recognition model parameters into the weights directory.

[0068] Step 3: Use the detect function to identify the acquired camera RTSP stream, output the box position 625 of the squid fin, and calculate the fin center points X506 and Y507 based on the squid length L504 and the angle θ505.

[0069] Further, if Figure 8 As shown, the specific process of step 1 is:

[0070] Step 1-1: Use OpenCV's VideoCapture function to read the camera RTSP stream.

[0071] Step 1-2: Use the Threshold function to binarize the squid image and remove the conveyor belt image.

[0072] Step 1-3: Use the erode function to perform an erosion operation to eliminate image noise, and use the dilate function to perform a dilation algorithm to remove clutter and connect adjacent pixels to make the image smoother and reduce tentacles.

[0073] Step 1-4: Extract the squid contours through findContours function.

[0074] Step 1-5: Calculate the minimum enclosing rectangle of the squid image contour point set through the minAreaRect function.

[0075] Step 1-6: Calculate the length L504 and angle θ505 of the circumscribed matrix according to the squid's slender characteristics.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the spirit and scope of the technical solution of the present invention.

Claims

1. A squid automatic feeding device based on visual intelligence, characterized in that: Including vibrating screen mechanism, conveying mechanism, grabbing and feeding mechanism, and image acquisition mechanism; The vibrating screen mechanism is arranged at the front entrance of the conveying mechanism, and the squids are dispersed and slide into the conveying mechanism after being vibrated by the vibrating screen mechanism; A squid-specific gripper assembly is provided at the top of the gripping and feeding mechanism, the gripping and feeding mechanism is arranged on one side of the conveying mechanism, and the squid-specific gripper assembly is suspended above the conveying mechanism; The image acquisition mechanism is arranged in front of and directly opposite to the grabbing and feeding mechanism.

2. The squid automatic feeding device based on visual intelligence according to claim 1 is characterized in that: The vibrating screen mechanism comprises a guide plate and a vibrating disk, wherein the guide plate is mounted on the vibrating disk to form multiple channels.

3. The squid automatic feeding device based on visual intelligence according to claim 2 is characterized in that: The multi-channel entrances are all in an eight-shaped shape, and parallel channels are arranged behind the eight-shaped shape. There is no straight section between each channel entrance and adjacent entrances, so as to prevent the squid from being stranded in the straight section.

4. The squid automatic feeding device based on visual intelligence according to any one of claims 1 to 3, characterized in that: The vibrating screen mechanism further comprises an angle adjustment component and / or a brush component; The angle adjustment component changes the angle with the horizontal plane by moving the telescopic rod up and down, thereby adjusting the sliding speed of the squid and preventing the squid from overlapping again; The brush assembly is located at the outlet of the vibrating screen mechanism and is used to prevent the squids from stagnating in one place.

5. The squid automatic feeding device based on visual intelligence according to claim 1 is characterized in that: The conveying mechanism is designed with a brush assembly at the inlet of the conveyor belt, which is used to adjust the squid's lying posture and prevent the grasping part from being blocked.

6. The squid automatic feeding device based on visual intelligence according to claim 5 is characterized in that: The rotating direction of the roller brush of the brush assembly is opposite to the moving direction of the conveyor belt. The distance between the roller brush and the conveyor belt is made to be the thickness of a single squid carcass by adjusting the moving plate up and down, so as to clean the obstructed part of the squid and the overlapping phenomenon of multiple squids.

7. The squid automatic feeding device based on visual intelligence according to claim 1 is characterized in that: The conveying mechanism arranges the squids in compartments and controls the squid conveying intervals according to the running speed and start and stop time of the conveying belt.

8. The squid automatic feeding device based on visual intelligence according to claim 1 is characterized in that: The image acquisition mechanism on the front side of the grabbing and feeding mechanism is used to accurately capture the squid image. After identifying whether the squid carcass is defective according to the image, the squid-specific gripper assembly selectively does not grab the defective squid, and the defective squid is sent to centralized processing via a conveyor belt; The image acquisition mechanism directly opposite the grasping and feeding mechanism is used to reconfirm the squid's aerial posture after grasping and complete the subsequent precise adjustment of the squid's posture.

9. A squid identification method using the squid automatic feeding device based on visual intelligence according to any one of claims 1 to 8, characterized in that: The following steps are involved: Step 1: Read the camera RTSP stream, perform binarization on the squid image, remove clutter, reduce tentacles, extract the squid outline, obtain the minimum circumscribed rectangle of the squid, and calculate the length and angle of the squid; Step 2: Read the camera RTSP stream, save the image, create squid fin tags, install and configure YOLOv5 parameters, train the fin recognition model, and put the trained fin recognition model parameters into the weights directory; Step 3: Identify the acquired camera RTSP stream, output the box position of the squid fin, and calculate the center point of the fin based on the squid length and angle.

10. The squid identification method based on visual intelligence according to claim 9, characterized in that: The step 1 specifically includes: Step 1-1: Read the camera RTSP stream; Step 1-2: Binarize the squid image and remove the conveyor belt image; Step 1-3: Use corrosion and expansion algorithms to remove clutter and reduce tentacles; Step 1-4: Extract the squid outline; Step 1-5: Obtain the minimum circumscribed matrix of the squid image contour point set; Step 1-6: Calculate the length and angle of the circumscribed matrix based on the squid's slender features.

Images