An automatic insertion method of a robotic arm based on an optical fiber connector end face detector

Abstract

The present invention proposes a robotic arm automatic insertion method based on an optical fiber connector end face detector. The automatic insertion system is arranged above the optical fiber connector end face detector, and the automatic insertion system includes: an image acquisition device, an identification and positioning system, a robotic arm and a mechanical gripper; the image acquisition device is used to obtain an image of the optical fiber connector; the identification and positioning system uses a target object recognition neural network based on the YOLOv5 algorithm to identify and position the optical fiber connector in the image; the robotic arm is used to drive the mechanical gripper to move to the coordinate position to pick up the optical fiber connector and move to the feeding area of the detector to insert the optical fiber connector. Through the technical solution of the present invention, it realizes the automatic insertion operation of the detector by using the target object recognition neural network to guide the robotic arm to pick up the optical fiber connector, and improves the versatility and efficiency of the optical fiber connector end face detector.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic feeding, and more specifically, to a method for automatic insertion of a robotic arm on an optical fiber connector end face detector. Background Art

[0002] An optical fiber connector is a device for detachably connecting optical fibers. It precisely docks the two end faces of the optical fibers so that the optical energy output from the transmitting optical fiber can be maximally coupled into the receiving optical fiber, and the impact on the system caused by its insertion into the optical link is minimized. This is the basic requirement of an optical fiber connector. To a certain extent, the optical fiber connector affects the reliability and various performances of the optical transmission system.

[0003] The optical fiber connector end face detector involved in the present invention is mainly based on the principle of white light interference, integrating multiple technical modules such as temperature control, light control, pixel discrimination, noise detection, filtering, and qualification detection. It can calculate and output 4 non-reference evaluation indicators of the MPO optical fiber connector and evaluate whether it is qualified according to the IEC standard, thereby increasing the reliability and correctness of the detection results. This optical fiber connector end face detector is divided into two parts: a multi-parameter MPO connector detection hardware device and an MPO connector detection software platform.

[0004] The hardware device uses white light interference technology. By integrating accessories such as a piezoelectric ceramic control platform and a CCD camera, and carrying a self-developed single-photon light intensity detection circuit and a high-precision temperature-sensitive control circuit, it can capture more accurate interference pictures and achieve the application standard of rapid detection.

[0005] The software platform is based on the interference image. It uses neural networks to learn the characteristics of environmental noise in the image and applies ensemble learning to optimize the discrimination algorithm to identify noise pixel points. For the distinguished noise positions, various filtering algorithms are applied for noise reduction processing, and finally the restored result is displayed on the visualization software platform for parameter measurement and display.

[0006] With the increasing development of science and technology, the requirements for the functions and working efficiency of products in society are getting higher and higher. However, the plugging and unplugging efficiency of the manual plugging and unplugging method of optical fiber connectors is not high, and the inspectors are prone to fatigue, resulting in unstable plugging and unplugging, which may lead to misjudgment of the quality of the connectors during detection, affecting the detection accuracy. In addition, when manually inserting the optical fiber connector, it is easy to tilt, causing damage to the connector and the optical fiber slot. Moreover, when the inspector is inconvenient to use the detector, the automatic insertion method of the robotic arm based on the optical fiber connector end face detector designed by the present invention can also solve this problem. Therefore, it is imperative to make an automated device for automatic insertion and judgment. Summary of the Invention

[0007] The object of the present invention is to solve at least one technical problem existing in the process of using machine vision recognition to guide a robotic arm to grasp an optical fiber connector for feeding a detector, namely, the accurate recognition of the target object (optical fiber connector), the analysis of its position information, and the design of a detachable robotic arm and an image acquisition device to improve the versatility and efficiency of the automatic insertion system of the detector.

[0008] The technical solution of the present invention is: a robotic arm automatic insertion method based on an optical fiber connector end face detector is proposed. The automatic insertion system is arranged above the optical fiber connector end face detector and is used for feeding the detector. The automatic insertion system includes: an image acquisition device, a recognition and positioning system, a robotic arm and a mechanical gripper; the fixed end of the image acquisition device is arranged on the upper left side of the optical fiber connector end face detector through a detachable base, and the image acquisition device is used for acquiring an image of the optical fiber connector beside the optical fiber connector end face detector; the recognition and positioning system is configured to recognize and position the optical fiber connector in the image by using a target object recognition neural network based on the YOLOv5 algorithm; the fixed end of the robotic arm is arranged on the upper right side of the optical fiber connector end face detector through a detachable base, a mechanical gripper is installed at the free end of the robotic arm, the robotic arm is used for driving the mechanical gripper to move to a coordinate position to grasp the optical fiber connector at the coordinate position, and the robotic arm is also used for moving to the feeding area of the detector and inserting the optical fiber connector.

[0009] In the above technical solution, the image acquisition device further includes a camera, a bracket and a detachable base. The camera is connected and fixed to the detachable base through the bracket, and the camera captures a 3-channel RGB color image with a length and width of 640 at a fixed position.

[0010] In the above technical solution, the recognition and positioning system is further configured to: recognize and position the optical fiber connector in the image by using a target object recognition neural network based on the YOLOv5 algorithm. First, the RGB color image containing the optical fiber connector is input into the YOLOv5 neural network, and the YOLOv5 neural network processes the image to recognize the optical fiber connector and two corner points on one side of the optical fiber connector in the image. The optical fiber connector and the two corner points on one side of the optical fiber connector in the image belong to two objects respectively. According to the target object, the bounding box of the optical fiber connector, the bounding boxes of the two corner points on one side of the optical fiber connector and the relevant parameters x, y, w, h of the bounding boxes, the bounding box confidence to and the confidence tci of the i-th category, and the class score Pr(Class1) corresponding to the target grid are obtained, and thus the target position information of the target object (i.e., the optical fiber connector) is analyzed.

[0011] In the above technical solution, before the target object recognition neural network based on the YOLOv5 algorithm performs recognition and positioning, the target object recognition neural network needs to be trained. The datasets composed of multiple data including fiber optic connector category objects and the datasets composed of multiple data including two corner point category objects on one side of the fiber optic connector are classified respectively. That is, the datasets are divided into a training set, a validation set, and a test set according to the ratio of 7:2:1, and the parameter model is saved after 800 iterations of training to obtain the trained target object recognition neural network.

[0012] In the above technical solution, the robotic arm is further configured to: determine whether there is an intersection between the bounding box of the second fiber optic connector to be clamped and the bounding box of the first fiber optic connector to be clamped. If there is no intersection between the bounding box of the second fiber optic connector to be clamped and the bounding box of the first fiber optic connector to be clamped, grab the first fiber optic connector to be clamped and the second fiber optic connector to be clamped in sequence. If there is an intersection between the bounding box of the second fiber optic connector to be clamped and the bounding box of the first fiber optic connector to be clamped, determine the next fiber optic connector as the second fiber optic connector to be clamped, and re-determine whether there is an intersection in the bounding box until the first fiber optic connector to be clamped and the second fiber optic connector to be clamped are determined.

[0013] In the above technical solution, the target position information of the target object further includes the fiber optic connector coordinates and the fiber optic connector angle. Among them, the fiber optic connector coordinates are the center point coordinates of the fiber optic connector bounding box, and the fiber optic connector angle is determined by trigonometric functions through the center point coordinates of the bounding boxes of two corner points on one side of the fiber optic connector.

[0014]

[0015] In the above technical solution, the detachable bases of the image acquisition device and the robotic arm are further configured to be fixed to the fiber optic connector end face detector through the screws on the detachable base, achieving the purpose of convenient transportation and carrying and firm installation.

[0016] Compared with the general fiber optic connector end face detector in the technical solution of the present invention, considering that the plugging and unplugging efficiency of the manual plugging and unplugging method of the fiber optic connector is not high, and it is easy to tilt when manually inserting the fiber optic connector, which may damage the connector and the fiber optic slot, thus affecting the detection accuracy. Furthermore, an automatic insertion method of the robotic arm based on the fiber optic connector end face detector is realized. Compared with the prior art, the automatic insertion system in the present invention omits the manual plugging and unplugging process of the fiber optic, improves the efficiency, and eliminates the situation that the fiber is prone to tilt and damage the connector and the fiber optic slot during the plugging and unplugging process, with higher stability and better versatility, and thus also improves the detection accuracy. Description of the Drawings

[0017] Figure 1Schematic flowchart of each step included in the robotic arm automatic insertion method based on an optical fiber connector end face detector provided by an embodiment of the present invention.

[0018] Figure 2 Schematic diagram of the construction process of the first model M1_YOLO and the second model M2_YOLO provided by an embodiment of the present invention.

[0019] Figure 3 Schematic diagram of the structure of the robotic arm based on an optical fiber connector end face detector provided by an embodiment of the present invention.

[0020] Figure 4 Schematic diagram of the structure of the detachable base of the image acquisition device provided by an embodiment of the present invention.

[0021] Figure 5 Schematic diagram of the structure of the detachable base of the robotic arm provided by an embodiment of the present invention.

[0022] Explanation of reference numerals: 1 optical fiber connector end face detector, 2 robotic arm, 3 mechanical clamp, 4 bracket, 5 camera, 6 detachable base, 7 screw. Detailed implementation manners

[0023] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Generally, the components of the embodiments of the present invention described and illustrated in the drawings here can be arranged and designed in various different configurations.

[0024] It should be clear that all other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts belong to the scope of protection of the present invention.

[0025] Refer to Figure 3For the content shown, an embodiment of the present invention provides a robotic arm automatic insertion method based on an optical fiber connector end face detector. The automatic insertion system is arranged above the optical fiber connector end face detector 1, and the automatic insertion system is used for feeding the detector. The automatic insertion system includes: an image acquisition device, an identification and positioning system, a robotic arm 2, and a mechanical gripper 3; the fixed end of the image acquisition device is arranged on the upper left side of the optical fiber connector end face detector through a detachable base 6, and the image acquisition device is used to acquire images of the optical fiber connector beside the optical fiber connector end face detector 1; the identification and positioning system is configured to identify and position the optical fiber connector in the image by using a target object recognition neural network based on the YOLOv5 algorithm; the fixed end of the robotic arm 2 is arranged on the upper right side of the optical fiber connector end face detector 1 through a detachable base 6, and a mechanical gripper 3 is installed at the free end of the robotic arm 2. The robotic arm 2 is used to drive the mechanical gripper 3 to move to the coordinate position to clamp the optical fiber connector at the coordinate position, and the robotic arm 2 is also used to move to the feeding area of the detector and insert the optical fiber connector.

[0026] Specifically, the automatic insertion system mainly includes an image acquisition device, an identification and positioning system, a robotic arm 2, and a mechanical gripper 3. The image acquisition device mainly consists of a camera 5, a bracket 4, and a detachable base 6, and the camera is connected and fixed to the detachable base through the bracket.

[0027] After the image acquisition device obtains a 3-channel RGB color image with a length and width of 640, in order to be able to identify a complete optical fiber connector, an identification and positioning system is introduced. The identification and positioning system uses a target object recognition neural network based on the YOLOv5 algorithm to identify and position the optical fiber connector in the image.

[0028] Specifically, before the identification and positioning system identifies, first, according to the shooting area of the camera 5, the robotic arm 2 is adjusted to delimit the clamping area as the target object recognition area, and the optical fiber connectors are arranged in sequence and spaced apart in the target object recognition area.

[0029] Specifically, before applying the identification and positioning system, the target object recognition neural network needs to be trained first so that it can accurately identify and position the optical fiber connector. The specific steps are as follows:

[0030] Step 1: Obtain a large number of images containing optical fiber connectors through the image acquisition device, and use the target detection annotation tool LabelImg to annotate the optical fiber connectors in each image one by one, thereby obtaining a data set D1, which is saved in txt format. The label data is stored in the txt, and each line of numbers represents: target category, x, y, w, h.

[0031] Step 2: Divide the dataset D1 with fiber optic connector annotations obtained in Step 1 into a training dataset D1_train, a validation dataset D1_val, and a test dataset D1_test. Input the training dataset D1_train into the YOLOv5 algorithm to obtain an object detection model M1_train. Further input the validation dataset D1_val into this object detection model for validation to obtain the first model M1_YOLO that meets the validation accuracy requirements. Then input the test dataset D1_test into the first model M1_YOLO that meets the validation accuracy requirements for testing, and finally confirm that the model M1_YOLO is the first model that meets the validation accuracy requirements.

[0032] Specifically, the confirmation model for the fiber optic connector angle in the target position information of the target object will also be trained using the above method. The specific steps are as follows:

[0033] Step 1: Use an image acquisition device to obtain a large number of images containing fiber optic connectors, and use the object detection annotation tool LabelImg to label the two corner points on one side of the fiber optic connector in each image one by one. Thus, obtain datasets D2 and D3 (D2 and D3 represent the datasets of the two corner points respectively), and save them in txt format. The label data is stored in the txt, and each line of numbers represents: target category, x, y, w, h.

[0034] Step 2: Divide the dataset D2 with the annotations of the two corner points on one side of the fiber optic connector obtained in Step 1 into a training dataset D2_train, a validation dataset D2_val, and a test dataset D2_test, and divide D3 into a training dataset D3_train, a validation dataset D3_val, and a test dataset D3_test. Input the training datasets D2_train and D3_train into the YOLOv5 algorithm to obtain an object detection model M2_train. Further input the validation datasets D2_val and D3_val into this object detection model for validation to obtain the second model M2_YOLO that meets the validation accuracy requirements. Then input the test datasets D2_test and D3_test into the second model M2_YOLO that meets the validation accuracy requirements for testing, and finally confirm that the model M2_YOLO is the second model that meets the validation accuracy requirements.

[0035] And, in combination with Figure 1 and Figure 2 , and provide the following specific embodiment in detail according to the actual model construction and the order of image detection and recognition:

[0036] As Figure 1 shown in the process schematic diagram,

[0037] S1. AsFigure 4 and Figure 5 As shown, the image acquisition device and the mechanical arm 2 are installed and fixed above the optical fiber connector end face detector 1 through a detachable base 6.

[0038] S2. According to the shooting area of the camera 5, the robot arm 2 is mobilized to mark the clamping area as the target object recognition area, and the optical fiber connectors are arranged in sequence and at intervals in the target object recognition area.

[0039] S3. Use the camera 5 carried by the image acquisition device to obtain a large number of 3-channel RGB color images with a length and width of 640, and select the images containing the optical fiber connector through preliminary screening.

[0040] S4. Perform manual frame selection and labeling processing on the image. The inspector uses the target detection and labeling tool LabelImg to select the rectangular frame of the optical fiber connector and the two corner points on one side of the optical fiber connector.

[0041] S5. Perform enhancement operations on the image in step 4, including rotation, scaling, blurring, etc., to enhance the sample quality and obtain sample data sets D1, D2, and D3 (D1 represents the data set of the optical fiber connector, and D2 and D3 represent the data sets of two corner points, respectively). The main purpose of the enhancement operation is to enhance the recognition of the image, especially to improve the recognition of the rectangular frame. Of course, the obtained image can also be enhanced and then the annotation process in step 4 can be performed to obtain more and clearer sample images.

[0042] S6. Randomly divide the sample data set D1 generated in step 5 into a training data set D1_train, a validation data set D1_val and a test data set D1_test in a ratio of 7:2:1; randomly divide D2 into a training data set D2_train, a validation data set D2_val and a test data set D2_test; and randomly divide D3 into a training data set D3_train, a validation data set D3_val and a test data set D3_test.

[0043] S7. Input the training data sets D1_train, D2_train and D3_train data (D1_train belongs to the fiber optic connector category object, D2_train and D3_train belong to the two corner point category objects on one side of the fiber optic connector) into the YOLOv5 algorithm for model training, adjust the algorithm parameters, and update the model parameters through multiple rounds of iterations. Each iteration will calculate the loss value until the loss value reaches an acceptable value, and obtain the target detection model of the fiber optic connector and the two corner points on one side of the fiber optic connector.

[0044] S8. Use the optimal object detection model generated in step 7 to predict and verify the model accuracy for the validation datasets D1_val, D2_val, and D3_val. Adjust the parameters and repeat step 7 until the model meets the accuracy requirements, such as the mean average precision mAP > 95%, to obtain the first model M1_YOLO and the second model M2_YOLO.

[0045] S9. Input the test dataset D1_test into the first model M1_YOLO that meets the validation accuracy requirements for testing, and input the test datasets D2_test and D3_test into the second model M2_YOLO that meets the validation accuracy requirements for testing. Finally, confirm that the model M1_YOLO is the first model that meets the validation accuracy requirements, and the model M2_YOLO is the second model that meets the validation accuracy requirements.

[0046] S10. For the image to be recognized and located, first use the first model M1_YOLO to identify the target object of the fiber optic connector in the image. Based on the target object, obtain the bounding box of the fiber optic connector and the related parameters x1, y1, w1, h1 of the bounding box, the bounding box confidence to1, the confidence tci1 of the i-th category, and the class score Pr(Class1) corresponding to the target grid. And predict the offsets (tx1, ty1, tw1, th1) through the sigmoid function, where (tw1, th1) needs to be processed by sigmoid twice.

[0047]

[0048]

[0049]

[0050] S11. Decode the offsets (tx1, ty1, tw1, th1) of the above fiber optic connector bounding box to obtain the final predicted bounding box of the fiber optic connector and the related parameters bx1, by1, bw1, bh1 of the predicted bounding box, which are the center point coordinates and width and height dimensions of the fiber optic connector predicted bounding box respectively. score1 is the confidence score, and cx1, cy1 are the distances between the grid occupied by the center of the fiber optic connector bounding box and the top-left grid. Thus, obtain the coordinates of the fiber optic connector in the target position information.

[0051] bx1 = 2σ(tx1) - 0.5 + cx1

[0052] by1 = 2σ(ty1) - 0.5 + cy1

[0053] bw1 = Pw1(2σ(tw1)) 2

[0054] bh1 = Ph1(2σ(th1)) 2

[0055] score1 = confidence × Pr(Class1) - σ(to1) × σ(tci1)

[0056] Among them, confidence has two cases, namely IOU and 0. When there is indeed an object in the grid, then Pr(Object) is equal to 1, so the predicted confidence is directly equal to IOU. If there is no object in the grid, then Pr(Object) is equal to 0, so confidence is directly equal to 0. Here, IOU is the intersection over union of the predicted bounding box and the true bounding box.

[0057] coafidence = Pr(Object) × IOU

[0058] S12. The image to be recognized and located is processed by the second model M2_YOLO again to identify the target objects at two corner points on one side of the fiber optic connector in the image. Based on the target objects, the bounding boxes of the two corner points on one side of the fiber optic connector and the related parameters x2, y2, and x3, y3 of the bounding boxes are obtained, and the offsets (tx2, ty2) and (tx3, ty3) are predicted through the sigmoid function.

[0059]

[0060]

[0061] S13. Decode the offsets (tx2, ty2) and (tx3, ty3) of the two corner point bounding boxes on one side of the above fiber optic connector to obtain the final predicted boxes of the two corner points on one side of the fiber optic connector and the related parameters bx2, by2, and bx3, by3 of the predicted boxes, which are the center point coordinates of the two corner point predicted boxes on one side of the fiber optic connector respectively. cx2, cy2, and cx3, cy3 are the distances from the center of the two corner point bounding boxes on one side of the fiber optic connector to the grid in the upper left corner of the grid (grid) respectively.

[0062] bx2 = 2σ(tx2) - 0.5 + cx2

[0063] by2 = 2σ(ty2) - 0.5 + cy2

[0064] bx3 = 2σ(tx3) - 0.5 + cy3

[0065] by3 = 2σ(ty3) - 0.5 + cy3

[0066] S14. The center point coordinates of the two corner prediction boxes on one side of the fiber optic connector obtained in step 13.

[0067]

[0068] S15. The recognition and positioning system transmits the target position information including the coordinates and angle of the fiber optic connector to the robotic arm 2. The robotic arm 2 drives the mechanical gripper 3 to move to the coordinate position, pick up the fiber optic connector at the coordinate position, and move to the loading area of the fiber optic connector end face detector 1 to insert the fiber optic connector.

[0069] The technical solution of the present invention has been described in detail above with reference to the accompanying drawings. The present invention provides a robotic arm automatic insertion method based on a fiber optic connector end face detector. The automatic insertion system is arranged above the fiber optic connector end face detector and is used for loading the detector. The automatic insertion system includes: an image acquisition device, a recognition and positioning system, a robotic arm, and a mechanical gripper. The fixed end of the image acquisition device is arranged on the upper left side of the fiber optic connector end face detector through a detachable base. The image acquisition device is used to acquire the image of the fiber optic connector beside the fiber optic connector end face detector. The recognition and positioning system is configured to recognize and locate the fiber optic connector in the image by using a target object recognition neural network based on the YOLOv5 algorithm. The fixed end of the robotic arm is arranged on the upper right side of the fiber optic connector end face detector through a detachable base. A mechanical gripper is installed at the free end of the robotic arm. The robotic arm is used to drive the mechanical gripper to move to the coordinate position to pick up the fiber optic connector at the coordinate position. The robotic arm is also used to move to the loading area of the detector and insert the fiber optic connector. Through the technical solution of the present invention, the operation of guiding the robotic arm to pick up the fiber optic connector for automatic insertion into the detector by using the target object recognition neural network is realized, and the versatility and efficiency of the fiber optic connector end face detector are improved.

Claims

1. An automatic insertion method of a robotic arm on an optical fiber connector end face detector, characterized in that, An automatic insertion system is applied. The automatic insertion system is arranged above the optical fiber connector end face detector, and the automatic insertion system is used for feeding the optical fiber connector end face detector. The automatic insertion system includes: an image acquisition device, a recognition and positioning system, a robotic arm, and a mechanical gripper; The fixed end of the image acquisition device is arranged on the upper left side of the optical fiber connector end face detector through a detachable base, and is used for acquiring images of the optical fiber connector beside the optical fiber connector end face detector; The recognition and positioning system is configured to recognize and position the optical fiber connector in the image by using a target object recognition neural network based on the YOLOv5 algorithm; The fixed end of the robotic arm is arranged on the upper right side of the optical fiber connector end face detector through a detachable base. A mechanical gripper is installed at the free end of the robotic arm. The robotic arm is used to drive the mechanical gripper to move to the coordinate position to pick up the optical fiber connector at the coordinate position. The robotic arm is also used to move to the feeding area of the detector and insert the optical fiber connector; The recognition and positioning system is configured to recognize and position the optical fiber connector in the image by using a target object recognition neural network based on the YOLOv5 algorithm. Specifically: The target object recognition neural network is trained to accurately recognize and position the optical fiber connector and two corner points on one side of the optical fiber connector. The specific steps are as follows: Step 1: A large number of images containing optical fiber connectors are acquired through the image acquisition device, and the optical fiber connectors in the images are labeled one by one by using the target detection annotation tool LabelImg, thereby obtaining data sets D1, D2, and D3; Step 2: Divide the dataset D1 obtained in Step 1 into a training dataset D1_train, a validation dataset D1_val, and a test dataset D1_test, divide D2 into a training dataset D2_train, a validation dataset D2_val, and a test dataset D2_test, and divide D3 into a training dataset D3_train, a validation dataset D3_val, and a test dataset D3_test; input the training dataset D1_train into the YOLOv5 algorithm to obtain an object detection model M1_train, and input the training datasets D2_train and D3_train into the YOLOv5 algorithm to obtain an object detection model M2_train; further input the validation dataset D1_val into M1_train for validation to obtain a first model M1_YOLO that meets the validation accuracy requirements, input the validation datasets D2_val and D3_val into M2_train for validation to obtain a second model M2_YOLO that meets the validation accuracy requirements; then input the test dataset D1_test into the first model M1_YOLO that meets the validation accuracy requirements for testing, and finally confirm that the model M1_YOLO is the first model that meets the validation accuracy requirements. Input the test datasets D2_test and D3_test into the second model M2_YOLO that meets the validation accuracy requirements for testing, and finally confirm that the model M2_YOLO is the second model that meets the validation accuracy requirements. Input the RGB color image containing the fiber optic connector into the YOLOv5 neural network. The YOLOv5 neural network processes the image to identify the fiber optic connector and two corner points on one side of the fiber optic connector in the image. The fiber optic connector and the two corner points on one side of the fiber optic connector in the image belong to two objects respectively. Based on the target objects, obtain the bounding box of the fiber optic connector, the bounding boxes of the two corner points on one side of the fiber optic connector, and the relevant parameters of the bounding boxes, and thus analyze the target position information of the fiber optic connector. The robotic arm is further configured to: determine whether there is an intersection between the bounding box of the second fiber optic connector to be picked and the bounding box of the first fiber optic connector to be picked. If there is no intersection between the bounding box of the second fiber optic connector to be picked and the bounding box of the first fiber optic connector to be picked, pick the first fiber optic connector to be picked and the second fiber optic connector to be picked in sequence. If there is an intersection between the bounding box of the second fiber optic connector to be picked and the bounding box of the first fiber optic connector to be picked, determine the next fiber optic connector as the second fiber optic connector to be picked, and re-determine whether there is an intersection between the bounding boxes until the first fiber optic connector to be picked and the second fiber optic connector to be picked are determined.

2. The automatic insertion method of the robotic arm on the fiber optic connector end face detector according to claim 1, wherein The image acquisition device is further configured to: include a camera, a lens, a bracket, and a detachable base. The camera is fixedly connected to the detachable base through the bracket, and the camera captures a 3-channel RGB color image with a length and width of 640 at a fixed position.

3. The automatic insertion method of the robotic arm on the fiber optic connector end face detector according to claim 1, characterized in that, The target position information includes the coordinates of the fiber optic connector and the angle of the fiber optic connector. Among them, the coordinates of the fiber optic connector are the coordinates of the center point of the bounding box of the fiber optic connector, and the angle of the fiber optic connector is determined by trigonometric functions through the coordinates of the center points of the bounding boxes of two corner points on one side of the fiber optic connector.

4. A robotic arm automatic insertion method on an optical fiber connector end face detector according to claim 1, characterized in that, The detachable base of the image acquisition device and the robotic arm is further configured to be fixed to the end face detector of the fiber optic connector through the screws at the four corners of the detachable base, achieving the purpose of convenient transportation and installation.

Images