Self-adaptive optimal angle image acquisition method and system

Abstract

The invention provides an adaptive optimal angle image acquisition method and system, and the method comprises the steps: training a target detection model through a first training set, and training a regression model through a second training set; the position of a packaging block in the unpacking process is detected in real time through the target detection model, and then the deformation length of the packaging block is obtained; judging whether the deformation length of the packing block reaches a preset threshold value, if so, recording the position of the packing block at the moment, outputting the orientation angle of the unpacking block through the regression model according to the picture of the packing block at the moment, converting to obtain the PTZ of the corresponding camera, and controlling the camera to collect the internal image of the packing block; and the camera shoots residual particle impurities on the unpacking table. In the field of unpacking detection of waste steel packing blocks, machine vision and a regression algorithm are used, and two unpacking internal optimum angle photos and one table-board residue soil residue photo are taken in a self-adaptive mode in the unpacking process of each waste steel packing block to serve as the basis and the basis of detection and judgment.

Description

Technical Field

[0001] The present invention belongs to the field of industrial machine vision technology, and more specifically, relates to an adaptive optimal angle image acquisition method for use in the unpacking detection process of scrap steel blocks. The optimal angle is the angle at which the torn surface after unpacking can be fully displayed in the camera image; Background Art

[0002] Scrap recycling and reuse is a crucial component of the metallurgical industry. Unpacking quality inspection directly impacts scrap sorting efficiency and resource utilization. Traditional unpacking inspection techniques rely primarily on manual visual inspection or image capture using fixed industrial cameras, resulting in low detection efficiency and incomplete image information. With the advancement of machine vision technology, the industry has experimented with applying intelligent imaging systems to identify scrap morphology. However, conventional approaches often rely on single-view static imaging, making it difficult to capture the three-dimensional morphological changes of scrap blocks during unpacking.

[0003] Although the existing deep learning-based image recognition algorithm has been applied to the classification of metal materials, it still faces technical bottlenecks such as difficulty in feature extraction and insufficient model generalization ability when dealing with unstructured scrap steel stacking scenarios.

[0004] Patent document CN113810605A discloses a target object processing method and apparatus. The method comprises determining a target identification area of ​​a target object based on a first image captured by a first camera; photographing the target object in the target identification area using a second camera to obtain multiple second images; processing the multiple second images to determine the proportion of objects to be processed within the target object; and determining a target deduction weight for the target object based on the proportion of objects to be processed. The method can capture images of a target object, such as scrap steel, using the first and second cameras during the unloading process to obtain a variety of multi-angle images. By processing these rich multi-angle images, various objects to be processed within the target object, such as various debris, can be detected and identified, ultimately calculating the target deduction weight.

[0005] Patent document CN114189629A discloses an image acquisition method, an image acquisition device, and an intelligent scrap grading system. This solution provides an image acquisition method, device, and intelligent scrap grading system. The image acquisition method includes: after the scrap transport vehicle's docking position meets preset docking specifications, a gun camera determines an effective area, which includes a discharge area and / or a drop area. After discharge begins, when the gun camera determines that the discharge device has entered the gun camera's field of view, an image acquisition device tracks the discharge device. The gun camera determines the discharge position and / or drop position of the discharge device based on the change in the number of points within the effective area, and transmits the discharge position and / or drop position of the discharge device to the image acquisition device. After receiving the discharge position and / or drop position, the image acquisition device takes a photo of the discharge position and / or drop position to obtain an image of the scrap at the discharge position and / or drop position until discharge is complete. A disadvantage of this solution is that it only captures images of bulk material on board a vehicle and is not applicable to packaged blocks. This problem urgently needs to be addressed. Summary of the Invention

[0006] In view of the defects in the prior art, the purpose of the present invention is to provide an adaptive optimal angle image acquisition method and system.

[0007] An adaptive optimal angle image acquisition method provided by the present invention includes:

[0008] Step S1: Collect the operation video of the unpacking machine to obtain the first training set and the second training set;

[0009] Step S2: training a target detection model using the first training set and a regression model using the second training set; the target detection model is capable of detecting the packing block and the unpacking table of the unpacking machine; and the regression model is capable of deriving the orientation angle of the unpacking block;

[0010] Step S3: Detecting the position of the packaged block in the unpacking process in real time using the target detection model, thereby obtaining the deformation length of the packaged block;

[0011] Step S4: Determine whether the deformation length of the packed block reaches a preset threshold. If yes, record the position of the packed block at this time, and output the orientation angle of the unpacked block through the regression model based on the packed block image at this time. Convert the angle into the corresponding PTZ of the camera according to the packed block position and the packed block orientation angle, and control the camera to rotate to the target position to capture the internal image of the unpacked packed block, and execute step S5. If no, execute step S3 again.

[0012] Step S5: Determine whether the packaged block has left the unpacking table. If yes, enable the camera to photograph the granular impurities remaining on the unpacking table, and end; if no, wait for the packaged block to leave, and end.

[0013] Preferably, in step S1, the packaging blocks in the pictures of the second training set are marked with angle values; the packaging blocks and unpacking platforms in the pictures of the first training set are marked with target detection frames; and the packaging blocks are scrap steel blocks generated by the baler.

[0014] Preferably, in step S2, the target detection model is a Yolo series model;

[0015] The unpacking block is the unpacked packing block;

[0016] The number of training rounds of the regression model is greater than or equal to 50 rounds;

[0017] The target detection model is trained for 50 or more rounds.

[0018] Preferably, in step S2, the Yolo series model includes yolov10;

[0019] In step S4, the unpacking block orientation angle, i.e., the angle between the torn surface of the unpacking block and the horizontal line, i.e., θ, is converted into the angle that the camera needs to rotate, i.e., the optimal angle, and then the camera is rotated to the target position. The mathematical expression of the closest angle is:

[0020] y=θ-90

[0021] Where y represents the optimal angle.

[0022] Preferably, in step S5, the position of the packaging block is detected by the target detection model, and it is determined whether the position of the packaging block intersects with the position of the unpacking table. If the result is yes, the camera is made to photograph the granular impurities remaining on the unpacking table, and the process ends; if the result is no, the process waits for the packaging block to leave, and the process ends.

[0023] In step S5, the camera is a box-type camera;

[0024] The height of the gun is greater than or equal to the height of the table top of the unpacking table, which is 5 meters.

[0025] An adaptive optimal angle image acquisition system provided by the present invention includes:

[0026] Module M1: Collect the operation video of the unpacking machine to obtain the first training set and the second training set;

[0027] Module M2: trains a target detection model using the first training set and a regression model using the second training set; the target detection model is capable of detecting the packing block and the unpacking platform of the unpacking machine; and the regression model is capable of deriving the orientation angle of the unpacking block;

[0028] Module M3: Detecting the position of the packaged block in the unpacking process in real time through the target detection model, and then obtaining the deformation length of the packaged block;

[0029] Module M4: Determine whether the deformation length of the packed block reaches a preset threshold. If yes, record the position of the packed block at this time, and output the orientation angle of the unpacked block through the regression model based on the packed block image at this time. According to the packed block position and the packing block orientation angle, it is converted into the corresponding PTZ of the camera, and the camera is controlled to rotate to the target position to collect the internal image of the unpacked packed block, triggering the operation of module M5; if no, re-trigger module M3;

[0030] Module M5: Determine whether the packing block has left the unpacking table. If yes, the camera is used to photograph the particle impurities remaining on the unpacking table, and the process ends. If no, the process waits for the packing block to leave, and the process ends.

[0031] Preferably, in the module M1, the packaging blocks in the pictures of the second training set are provided with angle value annotations; the packaging blocks and the unpacking table in the pictures of the first training set are provided with target detection frame annotations; and the packaging blocks are scrap steel blocks generated by the baler.

[0032] Preferably, in the module M2, the target detection model is a Yolo series model;

[0033] The unpacking block is the unpacked packing block;

[0034] The number of training rounds of the regression model is greater than or equal to 50 rounds;

[0035] The target detection model is trained for 50 or more rounds.

[0036] Preferably, in the module M2, the Yolo series model includes yolov10;

[0037] In the module M4, the unpacking block orientation angle, i.e., the angle between the tearing surface of the unpacking block and the horizontal line, i.e., θ, is converted into the angle that the camera needs to rotate, i.e., the optimal angle, and then the camera is rotated to the target position. The mathematical expression of the closest angle is:

[0038] y=θ-90

[0039] Where y represents the optimal angle.

[0040] Preferably, in the module M5, the position of the packaging block is detected by the target detection model, and it is determined whether the position of the packaging block intersects with the position of the unpacking table. If the result is yes, the camera is made to photograph the granular impurities remaining on the unpacking table, and the process ends; if the result is no, the process waits for the packaging block to leave, and the process ends.

[0041] In the module M5, the camera is a gun camera;

[0042] The height of the gun is greater than or equal to the height of the table top of the unpacking table, which is 5 meters.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] 1. In the field of scrap steel block unpacking detection, the present invention uses machine vision, regression algorithms and a unique camera setup scheme to adaptively capture two photos of the unpacking interior at the best angle and one photo of the residual soil on the table as the basis and basis for detection during the unpacking process of each scrap steel block.

[0045] 2. The present invention adopts machine vision, regression algorithm and specific camera arrangement scheme to adaptively provide two photos of the best angle of the unpacking interior and one photo of the residual soil on the table for each scrap steel package, that is, the unpacking process of the package. This solves the problem of being unable to effectively inspect the package due to the invisible interior of the package and the inconsistent degree of unpacking of the package.

[0046] 3. The output of the present invention is the angle between the tearing surface and the horizontal line, and the angle is converted into the angle that the ball camera needs to rotate. The camera rotates to the corresponding position by the specified angle and can adaptively take two photos of the best angle inside the unpacking. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Other features, objects and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0048] Figure 1 This is a top view of the location overview of the unpacking machine and nearby collection equipment provided by the present invention; wherein 1 represents camera No. 1; 2 represents camera No. 2; 3 represents camera No. 3;

[0049] Figure 2 This is a schematic diagram of the single unpacking camera acquisition process provided by the present invention. DETAILED DESCRIPTION

[0050] The present invention will be described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present invention, but are not intended to limit the present invention in any form. It should be noted that, for those skilled in the art, several changes and improvements can be made without departing from the scope of the present invention. These all fall within the scope of protection of the present invention.

[0051] According to the present invention, an adaptive optimal angle image acquisition system is provided, and its hardware includes: a package unpacking machine and a plurality of cameras; the plurality of cameras are circumferentially arranged around the package unpacking machine;

[0052] Specifically, the number of the plurality of cameras is three;

[0053] The camera is an industrial camera; specifically, the camera is an industrial network camera.

[0054] The unpacking machine comprises an unpacking platform, a plurality of unpacking machine grippers and electrical components, wherein the unpacking platform is a part of the unpacking machine.

[0055] Specifically, a bag unpacking machine gripper is provided on each side of the unpacking table.

[0056] Specifically, the plurality of cameras include: camera 1, camera 2 and camera 3;

[0057] In the embodiment provided by the present invention, the camera No. 1 and the camera No. 2 are both ball cameras; the camera No. 3 is a box camera; the camera No. 1 is the first camera; the camera No. 2 is the second camera; and the camera No. 3 is the third camera.

[0058] The ball camera is kept at the same height as the unpacking table through the vertical pole; the gun camera is 5 meters higher than the unpacking table through the vertical pole;

[0059] The camera No. 1, camera No. 2 and camera No. 3 are arranged at intervals in the circumferential position of the unpacking machine. Specifically, the camera No. 1 and camera No. 2 are arranged on both sides of camera No. 3.

[0060] According to the present invention, an adaptive optimal angle image acquisition method is provided, and its software system process includes:

[0061] Step S1: Collect unpacking operation video to obtain the first training set and the second training set;

[0062] Step S2: training a target detection model using the first training set and a regression model using the second training set; the target detection model is capable of detecting the packing block and the unpacking station; and the regression model is capable of deriving the orientation angle of the unpacking block;

[0063] Step S3: The object detection model is used to detect the position of the baling block in real time during the unpacking process, and then calculate and obtain the deformation length of the baling block; the baling block is the result of the baling machine. The baling machine compresses the scrap steel into a baling block.

[0064] Step S4: Determine whether the deformation length of the packaged block reaches a preset threshold. If yes, record the position of the packaged block at this time, and output the orientation angle of the unpacked block through the regression model based on the packaged block image at this time. Convert the angle into the corresponding camera PTZ according to the packaged block position and the packaged block orientation angle, and control the No. 1 and No. 2 cameras in front of the unpacking machine to rotate to the optimal position and take pictures of the interior of the unpacked packaged block, and execute step S5; if no, execute step S3 again;

[0065] Specifically, camera rotation is achieved through the SDK provided by the manufacturer.

[0066] Specifically, after a packaged block is disassembled, it will be divided into two pieces. Due to the uncertain length of the displacement, the imaging effect of a fixed camera position is poor, for example, the torn surface cannot be captured. However, a regression model can be trained to determine the tilt angle of each small piece after disassembly, and then the optimal angle can be calculated using the normal vector. The camera only needs to be controlled to the calculated optimal angle.

[0067] Specifically, the camera rotation parameters are PTZ, meaning the position of a camera lens is uniquely determined by the three parameters P, T, and Z. In other words, the present invention uses the tilt angle and normal vector to control the camera to a calculated optimal angle. This optimal angle is the angle facing the tearing surface of the unpacking block.

[0068] In other words, the regression model can determine the orientation angle of the unpacked block, that is, the angle between the torn surface of the unpacked block and the horizontal line, or θ. This angle is the tilt angle determined by the data used during model training. Therefore, this angle can be converted into the angle at which the camera needs to rotate, that is, the optimal angle.

[0069] The camera is rotated to a corresponding position through an optimal angle.

[0070] The mathematical expression of the closest angle is:

[0071] y=θ-90

[0072] Where y represents the optimal angle.

[0073] Among them, the PTZ of the corresponding camera is converted according to the package block position and package block orientation angle.

[0074] Step S5: Determine whether the packaged block has left the unpacking table. If yes, camera 3 takes a picture of the particle impurities remaining on the unpacking table. If no, wait for the packaged block to leave.

[0075] Specifically, in step S1, the packing blocks in the pictures of the second training set are annotated with angle values; the packing blocks and unpacking stations in the pictures of the first training set are annotated with target detection frames;

[0076] Specifically, in step S2, the target detection model is a Yolo series model, such as yolov10;

[0077] The regression model is a model with a regression head that can realize fitting numerical values;

[0078] In step S2, the model is trained; the number of training rounds is greater than or equal to 50 rounds;

[0079] Specifically, in step S5, the determination of whether the packaged block has left the unpacking station is implemented as follows: the position of the packaged block is detected through the target detection model. The position of the packaged block is fixed, and it can be determined whether it has left by simply determining that there is no intersection between the detected position of the packaged block and the position of the unpacking station.

[0080] Step 1: Collect training set photos of the packing blocks and unpacking stations. A camera near the unpacking machine station records the daily unpacking operations on site. Sample photos of the unpacking process are extracted from the video and annotated to obtain the training set.

[0081] Step 2: Collect training set photos of the packing blocks and unpacking stations. A camera near the unpacking machine station records the daily unpacking operations on site. Frames are extracted from the video to obtain fully unpacked sample photos, which are then annotated to obtain the training set.

[0082] Step 3: Train the object detection model on the training set obtained in step 1. Obtain a model that can detect packing blocks and unpacking stations.

[0083] Step 4: Train the regression model on the training set obtained in step 2. Obtain a model that can calculate the orientation angle of the unpacked block by looking at the image after the unpacked block is unpacked.

[0084] Step 5: During the on-site unpacking process, the model obtained in step 2 is used to detect the position of the packaged block, and the deformation length of the packaged block is calculated in real time.

[0085] Step 6: When the deformation length of the packaged block reaches the preset threshold, the position of the packaged block at this time is recorded, and based on the packaged block image at this time, the regression model in step 4 is used to output the packaged block orientation angle. According to the packaged block position and the packaged block orientation angle, it is converted into the corresponding camera PTZ, and the No. 1 and No. 2 cameras in front of the unpacking machine are controlled to rotate to the optimal position and take pictures of the inside of the unpacked packaged block.

[0086] Step 7: After the camera in front of the unpacking machine takes a picture, it continues to determine in real time whether the packaged block has left the unpacking table.

[0087] Step 8: After the packaged block leaves the unpacking table, control the camera No. 3 which is tilted upwards to take pictures of the granular impurities remaining on the unpacking table.

[0088] The step 1 comprises:

[0089] Step 1.1: Take the unpacking operation video at different time periods.

[0090] Step 1.2: Filter the video obtained in step 1.1 by extracting frames to obtain usable unpacking process images.

[0091] Step 1.3: Annotate the target detection frames for the packing blocks and unpacking stations in the image obtained in step 1.2.

[0092] The step 2 includes:

[0093] Step 2.1: Take the unpacking operation video at different time periods.

[0094] Step 2.2: Frame extraction and screening are performed on the video obtained in step 2.1 to obtain a usable complete unpacking end image.

[0095] Step 2.3: Annotate the angle values ​​of the packed blocks in the image obtained in step 2.2.

[0096] The step 3 comprises the following steps:

[0097] Step 3.1: Perform data augmentation, resolution unification, and histogram equalization on the training set in step 1.

[0098] Step 3.2: Use the dataset processed in step 3.1 to train a target detection model to obtain a model.

[0099] The step 4 comprises the following steps:

[0100] Step 4.1: Perform data augmentation, resolution unification, and histogram equalization on the training set in step 2.

[0101] Step 4.2: Perform regression model training on the data set processed in step 4.1 to obtain a model.

[0102] The step 5 comprises the following steps:

[0103] Step 5.1: Establish the scale from pixels to actual distances in advance by measuring and calibrating.

[0104] Step 5.2: Calculate the deformation length of the packed block in real time using the scale in step 5.1.

[0105] The present invention also provides an adaptive optimal angle image acquisition system, which can be implemented by executing the process steps of the adaptive optimal angle image acquisition method. That is, those skilled in the art can understand the adaptive optimal angle image acquisition method as a preferred implementation of the adaptive optimal angle image acquisition system.

[0106] An adaptive optimal angle image acquisition system provided by the present invention includes:

[0107] Module M1: Collect the operation video of the unpacking machine to obtain the first training set and the second training set;

[0108] Module M2: trains a target detection model using the first training set and a regression model using the second training set; the target detection model is capable of detecting the packing block and the unpacking platform of the unpacking machine; and the regression model is capable of deriving the orientation angle of the unpacking block;

[0109] Module M3: Detecting the position of the packaged block in the unpacking process in real time through the target detection model, and then obtaining the deformation length of the packaged block;

[0110] Module M4: Determine whether the deformation length of the packed block reaches a preset threshold. If yes, record the position of the packed block at this time, and output the orientation angle of the unpacked block through the regression model based on the packed block image at this time. According to the packed block position and the packing block orientation angle, it is converted into the corresponding PTZ of the camera, and the camera is controlled to rotate to the target position to collect the internal image of the unpacked packed block, triggering the operation of module M5; if no, re-trigger module M3;

[0111] Module M5: Determine whether the packing block has left the unpacking table. If yes, the camera is used to photograph the particle impurities remaining on the unpacking table, and the process ends. If no, the process waits for the packing block to leave, and the process ends.

[0112] In the description of this application, it should be understood that the terms "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on this application.

[0113] The above describes specific embodiments of the present invention. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art may make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. The embodiments of this application and the features in the embodiments may be combined with each other in any manner unless there is a conflict.

Claims

1. An adaptive optimal angle image acquisition method, characterized in that: include: Step S1: Collect the operation video of the unpacking machine to obtain the first training set and the second training set; Step S2: training the target detection model using the first training set, and training the regression model using the second training set; The target detection model can detect the packing block and the unpacking platform of the unpacking machine; the regression model can derive the orientation angle of the unpacking block; Step S3: Detecting the position of the packaged block in the unpacking process in real time using the target detection model, thereby obtaining the deformation length of the packaged block; Step S4: Determine whether the deformation length of the packed block reaches a preset threshold. If yes, record the position of the packed block at this time, and output the orientation angle of the unpacked block through the regression model based on the packed block image at this time. Convert the angle into the corresponding PTZ of the camera according to the packed block position and the packed block orientation angle, and control the camera to rotate to the target position to capture the internal image of the unpacked packed block, and execute step S5. If no, execute step S3 again. Step S5: Determine whether the packaged block has left the unpacking table. If yes, enable the camera to photograph the granular impurities remaining on the unpacking table, and end; if no, wait for the packaged block to leave, and end.

2. The adaptive optimal angle image acquisition method according to claim 1, characterized in that: In step S1, the packaging blocks in the pictures of the second training set are annotated with angle values; the packaging blocks and the unpacking platform in the pictures of the first training set are annotated with target detection frames; and the packaging blocks are scrap steel blocks generated by the baler.

3. The adaptive optimal angle image acquisition method according to claim 1, characterized in that: In step S2, the target detection model is a Yolo series model; The unpacking block is the unpacked packing block; The number of training rounds of the regression model is greater than or equal to 50 rounds; The target detection model is trained for 50 or more rounds.

4. The adaptive optimal angle image acquisition method according to claim 3, characterized in that: In the step S2, the Yolo series model includes yolov10; In step S4, the unpacking block orientation angle, i.e., the angle between the torn surface of the unpacking block and the horizontal line, i.e., θ, is converted into the angle that the camera needs to rotate, i.e., the optimal angle, and then the camera is rotated to the target position. The mathematical expression of the closest angle is: y=θ-90 Where y represents the optimal angle.

5. The adaptive optimal angle image acquisition method according to claim 1, characterized in that: In step S5, the target detection model is used to detect the position of the packaging block, and it is determined whether the position of the packaging block intersects with the position of the unpacking table. If yes, the camera is used to photograph the granular impurities remaining on the unpacking table, and the process ends; if no, the process waits for the packaging block to leave, and the process ends. In step S5, the camera is a box-type camera; The height of the gun is greater than or equal to the height of the table top of the unpacking table, which is 5 meters.

6. An adaptive optimal angle image acquisition system, characterized in that: include: Module M1: Collect the operation video of the unpacking machine to obtain the first training set and the second training set; Module M2: training the target detection model using the first training set and training the regression model using the second training set; The target detection model can detect the packing block and the unpacking platform of the unpacking machine; the regression model can derive the orientation angle of the unpacking block; Module M3: Detecting the position of the packaged block in the unpacking process in real time through the target detection model, and then obtaining the deformation length of the packaged block; Module M4: Determine whether the deformation length of the packed block reaches a preset threshold. If yes, record the position of the packed block at this time, and output the orientation angle of the unpacked block through the regression model based on the packed block image at this time. According to the packed block position and the packing block orientation angle, it is converted into the corresponding PTZ of the camera, and the camera is controlled to rotate to the target position to collect the internal image of the unpacked packed block, triggering the operation of module M5; if no, re-trigger module M3; Module M5: Determine whether the packing block has left the unpacking table. If yes, the camera is used to photograph the particle impurities remaining on the unpacking table, and the process ends. If no, the process waits for the packing block to leave, and the process ends.

7. The adaptive optimal angle image acquisition system according to claim 6, characterized in that: In the module M1, the packaging blocks in the pictures of the second training set are labeled with angle values; the packaging blocks and unpacking tables in the pictures of the first training set are labeled with target detection frames; and the packaging blocks are scrap steel blocks generated by the baler.

8. The adaptive optimal angle image acquisition system according to claim 6, characterized in that: In the module M2, the target detection model is a Yolo series model; The unpacking block is the unpacked packing block; The number of training rounds of the regression model is greater than or equal to 50 rounds; The target detection model is trained for 50 or more rounds.

9. The adaptive optimal angle image acquisition system according to claim 8, characterized in that: In the module M2, the Yolo series models include yolov10; In the module M4, the unpacking block orientation angle, i.e., the angle between the tearing surface of the unpacking block and the horizontal line, i.e., θ, is converted into the angle that the camera needs to rotate, i.e., the optimal angle, and then the camera is rotated to the target position. The mathematical expression of the closest angle is: y=θ-90 Where y represents the optimal angle.

10. The adaptive optimal angle image acquisition system according to claim 6, characterized in that: In the module M5, the target detection model is used to detect the position of the packaging block, and it is determined whether the position of the packaging block intersects with the position of the unpacking table. If so, the camera is used to photograph the granular impurities remaining on the unpacking table, and the process ends; if not, the process waits for the packaging block to leave, and the process ends. In the module M5, the camera is a gun camera; The height of the gun is greater than or equal to the height of the table top of the unpacking table, which is 5 meters.

Images