Machine vision-based strand unmanned following traction intelligent identification system and method

Abstract

The application discloses a kind of based on machine vision's strand unmanned following traction intelligent identification system, comprising: data acquisition module, the front view picture and rear view picture of the pulling device and the position information of pulling device are collected;Data wireless transmission module is used for the transmission of the data of data acquisition module;Edge computing terminal is used to receive the front view picture and rear view picture of the pulling device and the position information of pulling device, the edge computing terminal also includes strand posture anomaly identification module and abnormal state early warning module, the strand posture anomaly identification module is used to identify strand abnormal state and alarm through abnormal state early warning module;Receiving end receives the abnormal alarm information sent by abnormal state early warning module.The application also discloses a kind of based on machine vision's strand unmanned following traction intelligent identification method.The application improves the intelligent level of on-site strand erection, saves artificial, improves traction efficiency, operability is strong, applicable to engineering site.

Description

Technical Field

[0001] This invention relates to the field of suspension bridge cable strand erection construction. More specifically, this invention relates to a machine vision-based intelligent recognition system and method for unmanned cable strand following and traction. Background Technology

[0002] In existing suspension bridge cable-stayed bridge construction, the common method is to manually monitor the cable-stayed erection status and identify any abnormalities by following the traction device. This involves manually observing and judging whether the traction device is functioning properly, whether the erected cable strands are precisely positioned within the support rollers, and whether any strands are loose. If a problem is detected, an alarm is triggered via walkie-talkie, the traction device is stopped, and the fault is rectified. This method requires a dedicated person to monitor the cable-stayed bridge from the catwalk, which is time-consuming, labor-intensive, and involves personnel working at height, posing safety hazards.

[0003] Emerging machine vision-based image recognition technology has many advantages such as non-contact, long distance, high precision, time and labor saving, and real-time monitoring, and has been widely used in the field of bridge construction. However, there are no construction cases to date that have used this technology for cable strand traction status recognition. Summary of the Invention

[0004] One objective of this invention is to provide a machine vision-based intelligent identification system and method for unmanned cable traction, in order to solve the aforementioned problem of identifying the cable traction erection status.

[0005] To achieve these objectives and other advantages according to the present invention, a machine vision-based unmanned following traction intelligent recognition system is provided, comprising:

[0006] The data acquisition module includes a front-view high-definition camera, a rear-view high-definition camera, and a Beidou positioning terminal. The front-view high-definition camera and the rear-view high-definition camera respectively acquire the front-view and rear-view images of the puller. The Beidou positioning terminal is connected to the rear-view high-definition camera and is used to acquire the position information of the puller.

[0007] A wireless data transmission module, used for data transmission from the data acquisition module;

[0008] An edge computing terminal is used to receive the front view and rear view of the puller and the position information of the puller transmitted by the data wireless transmission module. The edge computing terminal also includes a cable posture abnormality identification module and an abnormal state early warning module. The cable posture abnormality identification module is used to identify the abnormal state of the cable through the front view and rear view of the puller and to issue an alarm through the abnormal state early warning module.

[0009] The receiving end receives abnormal alarm information sent by the abnormal state early warning module.

[0010] Preferably, the data wireless transmission module adopts a combination of wireless AP and wireless bridge.

[0011] Preferably, the abnormal alarm information includes the type of abnormal cable posture, a screenshot of the abnormal posture, and the current position information of the puller.

[0012] Preferably, the edge computing terminal further includes a winch control module, which is used to control the start and stop of the winch. After the abnormal state of the cable posture is identified, the cable posture abnormality identification module also controls the winch to stop through the winch control module.

[0013] This invention also provides a machine vision-based intelligent recognition method for unmanned traction of stranded cables, comprising the following steps:

[0014] Step 1: Install the intelligent recognition system on the puller and establish a communication connection with the backend receiver;

[0015] Step 2: Activate the intelligent identification system and begin cable strand traction construction;

[0016] Step 3: The data acquisition module acquires the front and rear view images of the puller through the front-view high-definition camera and the rear-view high-definition camera respectively, and acquires the position information of the puller through the Beidou positioning terminal. The data acquired by the data acquisition module is transmitted to the edge computing terminal through the data wireless transmission module. After the edge computing terminal obtains the front and rear view image data of the puller, it identifies the abnormal state of the cable through the cable posture abnormality identification module. When the edge computing terminal identifies the abnormality, it sends the abnormality alarm information to the receiving end of the background through the abnormal state early warning module, and at the same time controls the winch to stop through the winch control module.

[0017] Step 4: After receiving the abnormal alarm information, the operator at the receiving end checks the screenshot of the abnormal posture in the background to determine if there is an abnormality. If there is an abnormality, the operator quickly goes to the position of the puller according to the current position information of the puller to handle it. After handling, the operator controls the winch to start working through the winch control module. If the operator confirms that the abnormal alarm information is incorrect, the operator controls the winch to continue working through the winch control module.

[0018] Step 5: Repeat steps 3 and 4 above until the entire traction of the cable strand is completed.

[0019] Preferably, the specific method for the abnormal state of the strand posture identification module to identify the abnormal state of the strand is as follows:

[0020] 3.1 Perform semantic segmentation on the acquired strand posture image to obtain the semantic segmentation regions of strand pixels and rolling pixels in the image;

[0021] 3.2 Using the semantic segmentation region of the scrolling pixels obtained in step 3.1, extract the bounding rectangle of the region;

[0022] 3.3 Using the semantic segmentation region of the indexed pixels obtained in step 3.1, calculate the intersection region of this region and the bounding rectangle obtained in step 3.2, and calculate the geometric parameters of each region as follows:

[0023] (X w1 ,Y w1 — Coordinates of the leftmost point of the circumscribed rectangle of the support roll;

[0024] (X w2 ,Y w1 — Coordinates of the rightmost point of the circumscribed rectangle of the support roller;

[0025] (X rmin ,Y r1 — Coordinates of the leftmost point in the tracing area;

[0026] (X rmax ,Y r2 — Coordinates of the rightmost point in the stock-collecting area;

[0027] w rmax —Maximum width of the thread-locking area;

[0028] 3.4. Based on the geometric parameters of the circumscribed rectangle of the support roll and the cable strand region, anomaly detection is performed, specifically including:

[0029] ① Normal: If X w1 <X rmin And X rmax <X w2 And w rmax / (X w2 -X w1 If the threshold is less than 1, the system operates normally and the winch does not stop; where threshold1 is the threshold value.

[0030] ② Loose threads: If X w1 <X rmin And X rmax <X w2 And w rmax / (X w2 -X w1 If the threshold is greater than or equal to 1, the system operates normally and the winch does not stop. The system sends the abnormal alarm information to the receiving end in the background through the abnormal state early warning module, and the relevant operators decide on the next action.

[0031] ③Rolling off:

[0032] Condition 1: If X rmin <Xw1 And X w1 <X rmax At this point, the cable detaches from the roller on the left side, but not completely. At this time, the abnormal status warning module sends the abnormal alarm information to the receiving end in the background, and the relevant operators decide on the next action.

[0033] Condition 2: If X rmax <X w1 At this point, the cable strand is completely detached from the support roller on the left side. At this time, the abnormal state early warning module sends the abnormal alarm information to the receiving end of the background, and at the same time, the winch control module controls the winch to stop.

[0034] Condition 3: If X w2 <X rmax And X rmin <X w2 At this point, the cable detaches from the support roller on the right side, but not completely. At this time, the abnormal status warning module sends the abnormal alarm information to the receiving end in the background, and the relevant operators decide on the next action.

[0035] Condition 4: If X w2 <X rmin At this point, the cable strand is completely detached from the support roller on the right side. At this time, the abnormal status early warning module sends the abnormal alarm information to the receiving end in the background, and at the same time, the winch control module controls the winch to stop.

[0036] Preferably, the specific steps for semantic segmentation of the strand posture image in step 3.1 are as follows:

[0037] 3.1.1 Data annotation: Collect a large number of cable posture images from the data acquisition module, including images of normal and abnormal cable postures, and annotate them using annotation tools;

[0038] 3.1.2 Model Training: A deep learning-based semantic segmentation model is used to train the model on the data;

[0039] 3.1.3 Model Inference: Using the model trained in 3.1.2, semantic segmentation is performed on the new input image to obtain the semantic segmentation image of the tether pose. The semantic segmentation regions of the rolling pixels and the tether pixels in the image can be obtained.

[0040] Preferably, in step 3.4, the threshold value is 0.45.

[0041] Preferably, step 3.1.2 uses the DeeplabV3+ model to train the data to obtain a semantic segmentation model for the strand posture image.

[0042] The present invention has at least the following beneficial effects:

[0043] The intelligent identification system and method of this invention can monitor abnormal conditions of the cable strand status and the position information of the puller during traction construction in real time, and provide proactive real-time warnings, eliminating the need for workers to accompany the cable strand throughout the entire process. This method improves the level of intelligence in on-site cable strand erection, saves labor, increases traction efficiency, is highly operable, and is suitable for engineering sites.

[0044] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0045] Figure 1 This is a schematic diagram of the intelligent recognition system of the present invention;

[0046] Figure 2 This is a schematic diagram of the cable strand traction structure of the present invention;

[0047] Figure 3 This is a schematic diagram of semantic segmentation according to the present invention;

[0048] Figure 4 This is a schematic diagram of the geometric parameters of each region of the present invention;

[0049] Figure 5 This is a schematic diagram of the anomaly detection structure of the present invention.

[0050] Explanation of reference numerals in the attached diagram: 1. Forward-looking high-definition camera; 2. Rear-looking high-definition camera; 3. Beidou positioning terminal; 4. Edge computing terminal; 5. Puller; 6. Traction cable strand. Detailed Implementation

[0051] The present invention will now be described in further detail with reference to the accompanying drawings, so that those skilled in the art can implement it based on the description.

[0052] It should be noted that, unless otherwise specified, the experimental methods described in the following embodiments are all conventional methods, and the reagents and materials described are all commercially available unless otherwise specified. In the description of this invention, the terms "lateral", "longitudinal", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", and "outer" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0053] like Figure 1 and Figure 2As shown, this invention provides a machine vision-based intelligent recognition system for unmanned cable-stayed traction, comprising:

[0054] The data acquisition module includes a front-view high-definition camera 1, a rear-view high-definition camera 2, and a Beidou positioning terminal 3. The front-view high-definition camera 1 and the rear-view high-definition camera 2 respectively acquire the front-view and rear-view images of the puller 5. The Beidou positioning terminal 3 is connected to the rear-view high-definition camera 2 as a whole and is used to acquire the position information of the puller 5. The puller 5 is connected to the traction cable strand 6.

[0055] A wireless data transmission module, used for data transmission from the data acquisition module;

[0056] Edge computing terminal 4 is used to receive the front view and rear view of the puller 5 and the position information of the puller 5 transmitted by the data wireless transmission module. The edge computing terminal 4 also includes a cable posture abnormality identification module and an abnormal state warning module. The cable posture abnormality identification module is used to identify the abnormal state of the cable through the front view and rear view of the puller 5 and to issue an alarm through the abnormal state warning module.

[0057] The receiving end receives abnormal alarm information sent by the abnormal state early warning module.

[0058] In the above technical solution, the data acquisition module includes two high-definition cameras and one Beidou positioning terminal 3. The two high-definition cameras respectively acquire the front-view and rear-view images (the key focus area) of the puller 5, providing high-definition images of the puller 5 from both sides. The Beidou positioning terminal 3 is bound to the rear-view high-definition camera 2 to provide the location information of the puller 5. The edge computing terminal 4 receives the high-definition image data of the puller 5 from both sides and the positioning data of the puller 5 transmitted through the data wireless transmission module. The edge computing terminal 4 also includes a cable posture anomaly recognition module to identify abnormal cable states. When the edge computing terminal 4 identifies an anomaly, it promptly sends an anomaly alarm message to the on-site operators at the back-end receiving end through the anomaly state early warning module.

[0059] In another technical solution, the data wireless transmission module adopts a fusion of wireless AP and wireless bridge. To achieve normal transmission of high-definition images on site, a fusion of wireless AP and wireless bridge is used. Compared with a simple wireless AP, this method provides a more stable network and enables stable high-definition data transmission for the puller 5 in motion mode and under signal interference conditions, while also offering high cost-effectiveness.

[0060] In another technical solution, the abnormal alarm information includes the type of abnormal cable posture, a screenshot of the abnormal posture, and the current position information of the puller 5. On-site personnel confirm the cable status and take further action. For example, if the system detects "rolling off," on-site personnel can confirm the "rolling off" by viewing the screenshot of the abnormal posture on their mobile phone (i.e., the receiving end backend), and quickly reach the position of the puller 5 based on its current location information to handle the situation. Alternatively, since the algorithm cannot be 100% accurate, when the system detects an anomaly and on-site personnel confirm it is an error, they will instruct the winch to continue operating.

[0061] In another technical solution, the edge computing terminal 4 also includes a winch control module, which controls the start and stop of the winch. After the strand attitude anomaly identification module identifies an abnormal strand state, it also controls the winch to stop via the winch control module. When the edge computing terminal 4 identifies an anomaly, it will also promptly control the winch via the winch control module. For example, if the system detects that the strand has detached, it will immediately control the winch to stop. The anomaly warning module and the winch control module work simultaneously and are controlled by the edge computing terminal 4.

[0062] This invention also provides a machine vision-based intelligent recognition method for unmanned traction of stranded cables, comprising the following steps:

[0063] Step 1: Install the intelligent recognition system on the puller 5 and establish a communication connection with the backend receiver;

[0064] Step 2: Activate the intelligent identification system and begin cable strand traction construction;

[0065] Step 3: The data acquisition module acquires the front view and rear view images of the puller 5 through the front view high-definition camera 1 and the rear view high-definition camera 2 respectively, and acquires the position information of the puller 5 through the Beidou positioning terminal 3. The data acquired by the data acquisition module is transmitted to the edge computing terminal 4 through the data wireless transmission module. After the edge computing terminal 4 obtains the front view and rear view image data of the puller 5, it identifies the abnormal state of the cable through the cable posture abnormality identification module. When the edge computing terminal 4 identifies the abnormality, it sends the abnormality alarm information to the receiving end of the background through the abnormal state early warning module, and at the same time controls the winch to stop through the winch control module.

[0066] Step 4: After receiving the abnormal alarm information, the operator at the receiving end checks the screenshot of the abnormal posture in the background to determine whether there is an abnormality. If there is an abnormality, the operator quickly goes to the position of the puller 5 according to the current position information of the puller 5 to handle it. After the handling is completed, the operator controls the winch to start working through the winch control module. If the operator confirms that the abnormal alarm information is incorrect, the operator controls the winch to continue working through the winch control module.

[0067] Step 5: Repeat steps 3 and 4 above until the entire traction of the cable strand is completed.

[0068] In another technical solution, the abnormal posture recognition of the strand is the core of the edge computing terminal 4. The specific method for the strand posture abnormality recognition module to identify the abnormal state of the strand is as follows:

[0069] 3.1. Perform semantic segmentation on the acquired strand posture image to obtain the semantic segmentation regions of strand pixels and rolling pixels in the image, such as... Figure 3 As shown, the background color represents the background, the rectangular frame divides the area to represent the support roller, and the trapezoidal frame divides the area inside to represent the cable strand;

[0070] 3.2 Using the semantic segmentation region of the scrolling pixels (the outer rectangular region) obtained in step 3.1, extract the bounding rectangle of this region;

[0071] 3.3 Using the semantic segmentation region (the inner trapezoidal bounding box region) obtained in step 3.1, calculate the intersection region of this region image and the bounding rectangle obtained in step 3.2, and calculate the geometric parameters of each region, such as... Figure 4 As shown, the parameters are as follows:

[0072] (X w1 ,Y w1 — Coordinates of the leftmost point of the circumscribed rectangle of the support roll;

[0073] (X w2 ,Y w1 — Coordinates of the rightmost point of the circumscribed rectangle of the support roller;

[0074] (X rmin ,Y r1 — Coordinates of the leftmost point in the tracing area;

[0075] (X rmax ,Y r2 — Coordinates of the rightmost point in the stock-collecting area;

[0076] w rmax —Maximum width of the thread-locking area;

[0077] 3.4, such as Figure 5 As shown, anomaly detection is performed based on the geometric parameters of the roll circumscribed rectangle and the cable strand region, specifically including:

[0078] ① Normal: If X w1 <X rmin And X rmax <X w2 And w rmax / (X w2 -X w1If the threshold is less than 1, the system operates normally and the winch does not stop; where threshold1 is the threshold value; it can be adjusted adaptively according to different projects, and optionally, threshold1 can be set to 0.45.

[0079] ② Loose threads (abnormality): If X w1 <X rmin And X rmax <X w2 And w rmax / (X w2 -X w1 If the threshold is greater than or equal to 1, the system operates normally and the winch does not stop. The system sends the abnormal alarm information to the receiving end in the background through the abnormal state early warning module, and the relevant operators decide on the next action.

[0080] ③Rolling off (abnormal):

[0081] Condition 1: If X rmin <X w1 And X w1 <X rmax At this point, the cable detaches from the roller on the left side, but not completely. At this time, the abnormal status warning module sends the abnormal alarm information to the receiving end in the background, and the relevant operators decide on the next action.

[0082] Condition 2: If X rmax <X w1 At this point, the cable strand is completely detached from the support roller on the left side. At this time, the abnormal state early warning module sends the abnormal alarm information to the receiving end of the background, and at the same time, the winch control module controls the winch to stop.

[0083] Condition 3: If X w2 <X rmax And X rmin <X w2 At this point, the cable detaches from the support roller on the right side, but not completely. At this time, the abnormal status warning module sends the abnormal alarm information to the receiving end in the background, and the relevant operators decide on the next action.

[0084] Condition 4: If X w2 <X rmin At this point, the cable strand is completely detached from the support roller on the right side. At this time, the abnormal status early warning module sends the abnormal alarm information to the receiving end in the background, and at the same time, the winch control module controls the winch to stop.

[0085] In another technical solution, the specific steps for semantic segmentation of the strand posture image in step 3.1 are as follows:

[0086] 3.1.1 Data Labeling: Collect a large number of cable posture images from the data acquisition module, including images of normal and abnormal cable postures, and label them using the Labelme annotation tool; among them, the background label is 0, the rolling label is 1, and the cable label is 2.

[0087] 3.1.2 Model Training: A deep learning-based semantic segmentation model is used to train the data; optionally, the DeeplabV3+ model is used to obtain a semantic segmentation model for the strand posture image.

[0088] 3.1.3 Model Inference: Using the model trained in 3.1.2, semantic segmentation is performed on the new input image to obtain the semantic segmentation image of the tether pose. The semantic segmentation regions of the rolling pixels and the tether pixels in the image can be obtained.

[0089] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A machine vision-based unmanned following traction intelligent recognition system for cable strands, characterized in that, include: The data acquisition module includes a front-view high-definition camera, a rear-view high-definition camera, and a Beidou positioning terminal. The front-view high-definition camera and the rear-view high-definition camera respectively acquire the front-view and rear-view images of the puller. The Beidou positioning terminal is connected to the rear-view high-definition camera and is used to acquire the position information of the puller. A wireless data transmission module, used for data transmission from the data acquisition module; An edge computing terminal is used to receive the front view and rear view of the puller and the position information of the puller transmitted by the data wireless transmission module. The edge computing terminal also includes a cable posture abnormality identification module and an abnormal state early warning module. The cable posture abnormality identification module is used to identify the abnormal state of the cable through the front view and rear view of the puller and to issue an alarm through the abnormal state early warning module. The receiving end receives abnormal alarm information sent by the abnormal state early warning module; The abnormal alarm information includes the type of abnormal cable posture, a screenshot of the abnormal posture, and the current position information of the puller. The edge computing terminal also includes a winch control module, which is used to control the start and stop of the winch. After the abnormal state of the cable strand is identified, the cable strand posture abnormality identification module also controls the winch to stop through the winch control module. The identification method using a machine vision-based unmanned traction intelligent recognition system includes the following steps: Step 1: Install the intelligent recognition system on the puller and establish a communication connection with the backend receiver; Step 2: Activate the intelligent identification system and begin cable strand traction construction; Step 3: The data acquisition module acquires the front and rear view images of the puller through the front-view high-definition camera and the rear-view high-definition camera respectively, and acquires the position information of the puller through the Beidou positioning terminal. The data acquired by the data acquisition module is transmitted to the edge computing terminal through the data wireless transmission module. After the edge computing terminal obtains the front and rear view image data of the puller, it identifies the abnormal state of the cable through the cable posture abnormality identification module. When the edge computing terminal identifies the abnormality, it sends the abnormality alarm information to the receiving end of the background through the abnormal state early warning module, and at the same time controls the winch to stop through the winch control module. Step 4: After receiving the abnormal alarm information, the operator at the receiving end checks the screenshot of the abnormal posture in the background to determine if there is an abnormality. If there is an abnormality, the operator quickly goes to the position of the puller according to the current position information of the puller to handle it. After handling, the operator controls the winch to start working through the winch control module. If the operator confirms that the abnormal alarm information is incorrect, the operator controls the winch to continue working through the winch control module. Step 5: Repeat steps 3 and 4 above until the entire traction strand of the cable to be pulled is completed; The specific method for the abnormal stock-picking posture identification module to identify abnormal stock-picking states is as follows: 3.1 Perform semantic segmentation on the acquired strand posture image to obtain the semantic segmentation regions of strand pixels and rolling pixels in the image; 3.2 Using the semantic segmentation region of the scrolling pixels obtained in step 3.1, extract the bounding rectangle of the region; 3.3 Using the semantic segmentation region of the indexed pixels obtained in step 3.1, calculate the intersection region of this region and the bounding rectangle obtained in step 3.2, and calculate the geometric parameters of each region as follows: (X w1 ,Y w1 ) —Coordinates of the leftmost point of the circumscribed rectangle of the roll; (X w2 ,Y w1 ) —Coordinates of the rightmost point of the circumscribed rectangle of the support; (X rmin ,Y r1 ) —Coordinates of the leftmost point in the tracing area; (X rmax ,Y r2 ) —Coordinates of the rightmost point in the stock-collecting area; w rmax —— Maximum width of the strand area; 3.

4. Based on the geometric parameters of the circumscribed rectangle of the support roll and the cable strand region, anomaly detection is performed, specifically including: ① Normal: If X w1 < X rmin and X rmax < X w2 and w rmax / ( X w2 - X w1 If the threshold is less than 1, the system operates normally and the winch does not stop; where threshold1 is the threshold value. ② Loose silk: If X w1 < X rmin and X rmax < X w2 and w rmax / ( X w2 - X w1 If the threshold is greater than or equal to 1, the system operates normally and the winch does not stop. The system sends the abnormal alarm information to the receiving end in the background through the abnormal state early warning module, and the relevant operators decide on the next action. ③Rolling off: Operating Condition 1: If X rmin < X w1 and X w1 < X rmax At this point, the cable detaches from the roller on the left side, but not completely. At this time, the abnormal status warning module sends the abnormal alarm information to the receiving end in the background, and the relevant operators decide on the next action. Operating Condition 2: If X rmax < X w1 At this point, the cable strand is completely detached from the support roller on the left side. At this time, the abnormal state early warning module sends the abnormal alarm information to the receiving end of the background, and at the same time, the winch control module controls the winch to stop. Operating Condition 3: If X w2 < X rmax and X rmin < X w2 At this point, the cable detaches from the support roller on the right side, but not completely. At this time, the abnormal status warning module sends the abnormal alarm information to the receiving end in the background, and the relevant operators decide on the next action. Operating Condition 4: If X w2 < X rmin At this point, the cable strand is completely detached from the support roller on the right side. At this time, the abnormal status early warning module sends the abnormal alarm information to the receiving end in the background, and at the same time, the winch control module controls the winch to stop.

2. The machine vision-based unmanned traction intelligent recognition system for cable strands as described in claim 1, characterized in that, The specific steps for semantic segmentation of the stock posture image in step 3.1 are as follows: 3.1.1 Data annotation: Collect a large number of cable posture images from the data acquisition module, including images of normal and abnormal cable postures, and annotate them using annotation tools; 3.1.2 Model Training: A deep learning-based semantic segmentation model is used to train the model on the data; 3.1.3 Model Inference: Using the model trained in 3.1.2, semantic segmentation is performed on the new input image to obtain the semantic segmentation image of the tether pose. The semantic segmentation regions of the rolling pixels and the tether pixels in the image can be obtained.

3. The machine vision-based unmanned traction intelligent recognition system for cable strands as described in claim 1, characterized in that, In step 3.4, the threshold value is 0.

45.

4. The machine vision-based unmanned traction intelligent recognition system for cable strands as described in claim 2, characterized in that, Step 3.1.2 uses the DeeplabV3+ model to train the data and obtains a semantic segmentation model for the strand posture image.

5. The machine vision-based unmanned traction intelligent recognition system for cable strands as described in claim 1, characterized in that, The data wireless transmission module adopts a combination of wireless AP and wireless bridge.

Images