Low-computing-power-demand double twister target detection method, system, medium and computer

Abstract

The application relates to a low-computing-power-demand double twister target detection method, a system, a medium and a computer, and the technical scheme comprises the following steps: a camera with a fixed height is arranged to shoot the double twister, and the photo is first cut, so that the size of the photo can be effectively reduced, and the time for identifying and processing the photo is reduced; according to the corresponding relationship among the positions of a wire passing wheel, a three-element and a raw material pipe on the photo, only the wire passing wheel is identified, and the position information of the three-element and the raw material pipe is calculated according to the position information of the wire passing wheel; and according to the above-mentioned scheme, the position information of the three-element and the raw material pipe can be calculated with extremely low computing power, the corresponding region is cut and then identified, the demand for computing resources can be reduced without reducing the definition of the camera photo, and through the cutting of the image, the proportion of the target region in the photo can be improved, and the identification accuracy can be effectively improved.

Description

Technical Field

[0001] This invention relates to the field of target detection technology, and more specifically, to a target detection method, system, medium, and computer for a low-computing-power-requirement twisting machine. Background Technology

[0002] Image acquisition is a crucial application in visual inspection, finding wide applicability across various fields, including industry, healthcare, agriculture, security monitoring, and autonomous driving. The purpose of image acquisition is to capture images of target objects or scenes and use these images for subsequent analysis, recognition, detection, and decision-making processes.

[0003] In the textile industry, a doubling machine is a twisting device, essentially a yarn-binding device, that can twist and bond two or more single yarns together to form a ply, thereby enhancing the performance of the original yarn. Modern doubling machines typically consist of a power unit, a doubling unit, and a transmission system. To ensure the normal operation of the doubling machine, it is necessary to monitor the three components of the doubling machine, rollers, guide rollers, and the remaining material in the raw material tube.

[0004] In the current production process, the monitoring of the doubling machine is usually based on manual monitoring, which requires workers to patrol constantly, check and judge the status of the doubling machine, and avoid the production of defective products due to problems with the spindles on a certain doubling machine.

[0005] A single existing doubling machine can twist 256 strands or even more at the same time. Each strand requires monitoring of the ternary winding, rollers, spindles, and raw material tubes. Often, a factory has dozens or more doubling machines, resulting in a large number of areas that need to be monitored. Relying solely on manual inspections would lead to an excessive workload for the inspectors, and the large number of doubling machines could easily result in missed inspections.

[0006] Therefore, some manufacturers have proposed a vision-based detection-based spindle monitoring solution for doubling twisters. This involves taking photos of the doubling twister and then using a pre-trained deep learning model to locate the target to be monitored in the photos and determine if the target is abnormal. If so, an alarm signal is issued to prompt relevant personnel to check immediately. However, this method still has the following problems: As the demand for image clarity continues to increase, the number of camera pixels continues to increase, the size of each frame image increases, and the number of markers increases, leading to a continuous increase in memory demand and pressure. This further leads to a continuous increase in image processing time. With the increase in the number of markers and image quality, the detection time of subsequent algorithms increases sharply, causing system performance bottlenecks. The increased processing time and memory pressure lead to a backlog of image processing tasks, eventually causing memory exhaustion and system failure. Therefore, how to reduce the computational demand on the system while ensuring the accuracy of target detection and maintaining photo clarity is a problem that the applicant urgently needs to solve. Summary of the Invention

[0007] To address the shortcomings of existing technologies, the present invention aims to provide a target detection method, system, medium, and computer with low computing power requirements for twisting machines, so as to reduce the computing power requirements of the target detection system without reducing the clarity of photographs or maintaining the accuracy of target detection.

[0008] The above-mentioned technical objective of the present invention is achieved through the following technical solution: a target detection method for a low-computing-power-requirement twisting machine, comprising:

[0009] S1. Use a pre-set inspection robot to film the working status of the twisting machine to obtain a working video; extract all frame images from the working video and record them as the first frame image;

[0010] S2. Based on the position of the yarn guide roller of the doubling machine in each first frame image, each first frame image is cut to obtain the corresponding second frame image; wherein, the width of the second frame image is the same as the width of the corresponding first frame image; the height of the second frame image is the height of the corresponding yarn guide roller in the corresponding first frame image;

[0011] S3. According to the preset size transformation rules, adjust the size of each second frame image to obtain the corresponding third frame image; perform target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel;

[0012] S4. Based on each first position information, determine whether the thread guide wheel on the corresponding third frame image is located inside the preset area. If yes, record the third frame image as the fourth frame image and execute step S5; otherwise, discard the third frame image.

[0013] S5. Based on the preset size transformation rules and the first position information corresponding to each fourth frame image, determine the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image; based on each second position information, calculate the third position information of the three-element double twister on the corresponding first frame image and the fourth position information of the raw material tube.

[0014] S6. Based on each third position information, crop the corresponding first frame image to obtain the corresponding fifth frame image; based on each fourth position information, crop the corresponding first frame image to obtain the corresponding sixth frame image.

[0015] S7. Use the pre-trained target detection neural network model to identify each fifth frame image and determine whether there is an abnormality in the corresponding twisting machine ternary. If so, send an alarm signal. Use the pre-trained target detection neural network model to identify each sixth frame image and determine whether there is an abnormality in the corresponding raw material pipe. If so, send an alarm signal.

[0016] Optionally, step S1 further includes: extracting all frame images from the working video, and extracting one frame image from the video at predetermined intervals, all of which are recorded as the first frame image.

[0017] Optionally, the step of adjusting the size of each second frame image according to a preset size transformation rule to obtain the corresponding third frame image includes: using OpenCV's resize algorithm to adjust the size of the second frame image, reducing the width and height of each second frame image to half of the second frame image to obtain the corresponding third frame image.

[0018] Optionally, the step of cropping each first frame image according to the position of the yarn guide roller of the twisting machine in each first frame image to obtain the corresponding second frame image includes: determining the first pixel coordinates corresponding to the highest point of the yarn guide roller and the second pixel coordinates corresponding to the lowest point of the yarn guide roller in the first frame image; cropping the first frame image with the horizontal line where the first pixel coordinates are located as the first cutting line and the horizontal line where the second pixel coordinates are located as the second cutting line to obtain the corresponding second frame image.

[0019] Optionally, the step of performing target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel includes: inputting each third frame image into a pre-trained target detection neural network model for detection, thereby obtaining the first bounding box that defines the wire guide wheel and the first position information corresponding to the bounding box.

[0020] Optionally, the step of determining the second position information of the thread guide wheel on the first frame image corresponding to the fourth frame image according to the preset size transformation rules and the first position information; and calculating the third position information of the twisting machine ternary and the fourth position information of the raw material tube on the first frame image according to the second position information of the thread guide wheel, includes: calculating the inverse affine transformation matrix according to the preset size transformation rules; calculating the second position information of the thread guide wheel on the corresponding first frame image according to the inverse affine transformation matrix and the first position information; obtaining the first position correspondence between the twisting machine ternary and the thread guide wheel, and the second position correspondence between the raw material tube and the thread guide wheel; determining the third position information of the twisting machine ternary on the first frame image according to the second position information and the first position correspondence; and determining the fourth position information of the raw material tube on the first frame image according to the second position information and the second position correspondence.

[0021] Optionally, determining whether the thread guide roller on the corresponding third frame image is located inside the preset area based on each first position information includes: setting a preset area with a width of 120 pixels at the center position of each third frame image; when the center line of the thread guide roller is located inside the preset area, the thread guide roller is considered to be located inside the preset area.

[0022] A target detection system for low-computing-power-requirement twisting machines includes:

[0023] Frame image extraction module: used to use a pre-set inspection robot to capture the working state of the twisting machine and obtain a working video; extract all frame images from the working video and record them as the first frame image;

[0024] Frame image cropping module: This module crops each first frame image according to the position of the yarn guide roller in each first frame image to obtain a corresponding second frame image. The width of the second frame image is the same as the width of the corresponding first frame image; the height of the second frame image is the height of the corresponding yarn guide roller in the corresponding first frame image; based on each third position information, it crops the corresponding first frame image to obtain a corresponding fifth frame image; based on each fourth position information, it crops the corresponding first frame image to obtain a corresponding sixth frame image.

[0025] Position information recognition module: used to adjust the size of each second frame image according to the preset size transformation rules to obtain the corresponding third frame image; and to perform target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel;

[0026] Frame image filtering module: It is used to determine whether the threading wheel on the corresponding third frame image is located inside the preset area according to each first position information. If it is, the third frame image is recorded as the fourth frame image and step S5 is executed; otherwise, the third frame image is discarded.

[0027] Position information calculation module: It is used to determine the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image according to the preset size transformation rules and the first position information corresponding to each fourth frame image; and to calculate the third position information of the twisting machine ternary and the fourth position information of the raw material tube on the corresponding first frame image according to each second position information.

[0028] Equipment status recognition module: Used to identify each fifth frame image using a pre-trained target detection neural network model to determine whether there is an abnormality in the corresponding twisting machine ternary element. If so, an alarm signal is sent. Used to identify each sixth frame image using a pre-trained target detection neural network model to determine whether there is an abnormality in the corresponding raw material pipe. If so, an alarm signal is sent.

[0029] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0030] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the method described above.

[0031] In summary, the present invention has the following beneficial effects: The technical solution of this application includes: taking pictures of the twisting machine by setting a camera at a fixed height, and first cropping the picture, which can effectively reduce the size of the picture, thereby reducing the time for picture recognition and processing. Based on the correspondence between the positions of the yarn guide roller, the ternary electrode, and the raw material tube in the picture, only the yarn guide roller is recognized, and the position information of the ternary electrode and the raw material tube is calculated according to the position information of the yarn guide roller. In summary, by adopting the solution of this application, the position information of the ternary electrode and the raw material tube can be calculated with extremely low computing power. After cropping the corresponding area, recognition can be performed, which can reduce the demand for computing resources without reducing the clarity of the camera picture. Moreover, by cropping the image, the proportion of the target area in the picture can be increased, which can effectively improve the accuracy of recognition. Attached Figure Description

[0032] Figure 1 This is a flowchart of the target detection method for a low-computing-power-requirement twisting machine according to the present invention;

[0033] Figure 2 This is a structural diagram of the low-computing-power-requirement twisting machine target detection system of the present invention;

[0034] Figure 3 This is an internal structural diagram of the computer device in an embodiment of the present invention.

[0035] Figure 4 This is a schematic diagram of the inspection robot in an embodiment of the present invention;

[0036] Figure 5 This is a schematic diagram illustrating the video recording operation of the inspection robot in an embodiment of the present invention.

[0037] In the diagram: 1. Frame image extraction module; 2. Frame image cropping module; 3. Location information recognition module; 4. Frame image filtering module; 5. Location information calculation module; 6. Device status recognition module. Detailed Implementation

[0038] To make the objectives, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Several embodiments of the present invention are shown in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein.

[0039] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances. The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0040] In this invention, unless otherwise expressly specified and limited, "above" or "below" a second feature can include direct contact between the first and second features, or contact between the first and second features through another feature between them. Furthermore, "above," "over," and "on top" of a second feature includes the first feature being directly above or diagonally above the second feature, or simply indicates that the first feature is at a higher horizontal level than the second feature. "Below," "below," and "under" of a second feature includes the first feature being directly below or diagonally below the second feature, or simply indicates that the first feature is at a lower horizontal level than the second feature. The terms "vertical," "horizontal," "left," "right," "above," "below," and similar expressions are for illustrative purposes only and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed or operated in a specific orientation, and therefore should not be construed as limiting the invention.

[0041] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0042] This invention provides a target detection method for a low-computing-power-requirement twisting machine, such as... Figure 1 As shown, it includes:

[0043] S1. Use a pre-set inspection robot to film the working status of the twisting machine to obtain a working video; extract all frame images from the working video and record them as the first frame image;

[0044] S2. Based on the position of the yarn guide roller of the doubling machine in each first frame image, each first frame image is cropped to obtain the corresponding second frame image; wherein, the width of the second frame image is the width of the corresponding first frame image; and the height of the second frame image is the height of the corresponding yarn guide roller in the corresponding first frame image.

[0045] S3. According to the preset size transformation rules, adjust the size of each second frame image to obtain the corresponding third frame image; perform target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel;

[0046] S4. Based on each first position information, determine whether the thread guide wheel on the corresponding third frame image is located inside the preset area. If yes, record the third frame image as the fourth frame image and execute step S5; otherwise, discard the third frame image.

[0047] S5. Based on the preset size transformation rules and the first position information corresponding to each fourth frame image, determine the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image; based on each second position information, calculate the third position information of the three-element double twister on the corresponding first frame image and the fourth position information of the raw material tube.

[0048] S6. Based on each third position information, crop the corresponding first frame image to obtain the corresponding fifth frame image; based on each fourth position information, crop the corresponding first frame image to obtain the corresponding sixth frame image.

[0049] S7. Use the pre-trained target detection neural network model to identify each fifth frame image and determine whether there is an abnormality in the corresponding twisting machine ternary. If so, send an alarm signal. Use the pre-trained target detection neural network model to identify each sixth frame image and determine whether there is an abnormality in the corresponding raw material pipe. If so, send an alarm signal.

[0050] In practical applications, due to the needs of image recognition, camera pixels are increasing, image sizes are growing, and the amount of information in images is gradually increasing. This leads to longer image processing times, slower image recognition speeds, and higher demands on device computing power. As the number of images to be processed increases, they accumulate in memory, causing memory exhaustion and device crashes. To address the trade-off between the computational pressure on image processing and image clarity, the applicant proposes a low-computing-power image recognition method for the textile industry, which can balance the requirements of image clarity and low computing power.

[0051] Specifically, this application is applied to the inspection of doubling machines in the textile industry. In actual production, to ensure the normal operation of the doubling machine, it is necessary to monitor the three components of the doubling machine, rollers, yarn guide rollers, and raw material tube. To ensure that all components are monitored, it is necessary to take as many pictures as possible. To improve the image clarity, it is necessary to increase the camera's pixel count. To ensure that each component can be accurately identified and judged, it is necessary to increase the camera's speed limit accordingly, thereby ensuring that the position of each component has sufficient impact information.

[0052] For commonly used doubling machines on the market, the doubling spindles installed on each specific type of doubling machine are usually located at the same height, and the arrangement of the doubling spindles follows a certain pattern; that is, the spindles on the doubling machine are arranged repeatedly. The inventors discovered that after fixing the camera height, in the photos or videos taken, the doubling spindles are always located at a certain predetermined height. For example, in photos or videos taken with a high-definition camera, the pixel dimensions are 2880*1960. Then, the area where the doubling machine spindle is located only appears in a certain horizontal area. For example, the area where the doubling machine spindle is located usually only appears in the middle area, and the dimensions of this area are 2420*1960. Furthermore, since the settings of the ternary element, the thread guide roller, and the raw material tube of each doubling spindle are identical, the areas where each ternary element, each thread guide roller, and each raw material tube appears are independent of each other, yet aligned with each other. For example, in a photograph, if a thread guide roller appears in the area belonging to the thread guide roller, then in the area of ​​the raw material tube, there will definitely be a raw material tube corresponding to the thread guide roller, and in the area of ​​the ternary element, there will definitely be a ternary element. Moreover, the lateral positions of the ternary element, the raw material tube, and the thread guide roller will also remain consistent. For instance, if the position of the thread guide roller shifts 10 pixels to the right, the ternary element and the raw material tube will also shift 10 pixels to the right accordingly.

[0053] Therefore, based on the above approach, in the actual shooting and recognition process, it is first necessary to set up a mobile device that can fix the camera at a specific height. For example... Figure 4The image shows an AGV mobile robot with a column on it. A camera is fixedly mounted on the column, and the camera's height needs to be adjusted according to the actual height of the spindle so that the camera can center the spindle in the frame when the AGV robot moves laterally. During video recording, as the AGV robot moves, the spindle slides laterally from left to right in the frame. That is, the area where the spindle appears is a long strip, the width of which is the same as the actual width of the photograph, and the height of the photograph is the same as the height of the spindle in the photograph. Similarly, within this spindle area, the width of the areas containing the thread guide roller, the raw material tube, and the ternary electrode is also the same as the actual width of the photograph, and their heights are the same as the pixel heights of the thread guide roller, the raw material tube, and the ternary electrode in the photograph, respectively. Therefore, in the actual inspection process, to identify the working condition, it is necessary to extract frame images from the video (i.e., the first frame image in this application). This only requires cropping the frame image into several regions according to the location of the parts, corresponding to smaller frame images (i.e., the second frame image in this application). By cropping the frame image using the corresponding positions of the components and then recognizing the cropped second frame image, the size of the frame image can be effectively reduced, thereby reducing the computing power required for image processing and recognition. In step S3, the size of the second frame image is adjusted, which can further reduce the computing power requirements for computer target detection.

[0054] Specifically, in actual shooting, the clarity and integrity of the target can only be guaranteed when the target is within the predetermined area. Due to lens distortion, targets located at the edge of the photo will be deformed due to fiber refraction, affecting the recognition effect. Therefore, it is necessary to identify the position of the wire guide wheel in the second frame image to determine whether the wire guide wheel is in the preset center area. If the wire guide wheel is not completely in the center area, the photo needs to be discarded. This improves the accuracy of the result and saves subsequent recognition steps, reducing the burden on the equipment. In step S5, the position of the wire guide wheel is first determined, that is, the position information of the wire guide wheel is obtained using the target detection network model. The actual position information usually includes: using a bounding box to frame the wire guide wheel and obtaining the position information of the bounding box as the first position information. Since the photo has been resized in the above steps, it is necessary to reverse the photo restoration rules according to the size transformation rules. Generally speaking, this is calculated using a transformation matrix, that is, calculating the inverse matrix of the transformation matrix and multiplying it by the position information of the bounding box to obtain the position information of the bounding box on the original photo.

[0055] After obtaining the positional information of the original image, the positions of the ternary electrode and the raw material tube can be directly inferred from the original image based on the positional relationship between the wire guide roller, the ternary electrode, and the raw material tube. For example, if the ternary electrode is located 800 pixels above the wire guide roller in the original image, and its area size is 300*400 pixels, its position can be quickly located based on this information. Similarly, the raw material tube is located below the wire guide roller, and its position can be quickly located using the same approach. This method eliminates the need for neural network models for detection, allowing the determination of the raw material tube and ternary electrode positions based on the position of the wire guide roller. Its detection and positioning effect is sufficiently accurate and does not consume excessive computing power. After determining the positions of the ternary electrode and the raw material tube, the area is cropped to isolate the ternary electrode and the raw material tube for target detection to determine whether they are functioning properly. The detected images are small in size and the targets are clearly defined, effectively reducing the computational power required. This improves the speed of image detection while still meeting the requirement for image clarity.

[0056] In summary, the technical solution of this application includes: taking pictures of the twisting machine by setting a camera at a fixed height, and first cropping the picture to effectively reduce the size of the picture, thereby reducing the time for picture recognition and processing; based on the positional correspondence between the yarn guide roller, the ternary winding, and the raw material tube in the picture, only the yarn guide roller is recognized, and the positional information of the ternary winding and the raw material tube is calculated based on the positional information of the yarn guide roller. In summary, by adopting the solution of this application, the positional information of the ternary winding and the raw material tube can be calculated with extremely low computing power. After cropping the corresponding area, recognition can be performed, which can effectively improve the accuracy of recognition and reduce the size of the picture to be processed, thus effectively reducing the computing power requirement.

[0057] Furthermore, step S1 also includes: extracting all frame images from the working video, and extracting one frame image from the video at predetermined intervals, all of which are recorded as the first frame image.

[0058] In practical applications, in order to reduce the overall computational load, the frame images can also be downsampled. That is, every two or three frames, the frame images in the video are extracted. Since the changes between adjacent frame images are not significant, interval sampling can also meet the corresponding requirements.

[0059] Furthermore, the step of adjusting the size of each second frame image according to the preset size transformation rules to obtain the corresponding third frame image includes: using OpenCV's resize algorithm to adjust the size of the second frame image, reducing the width and height of each second frame image to half of the second frame image to obtain the corresponding third frame image.

[0060] OpenCV (Open Source Computer Vision Library) is an open-source software library for computer vision and machine learning. It consists of a series of C functions and a small number of C++ classes, while also providing interfaces for languages ​​such as Python, Java, and MATLAB. It implements many common algorithms in image processing and computer vision. The resize algorithm can enlarge or shrink the original image to a predetermined size as needed. Furthermore, based on the transformation matrix in the resize function, the inverse matrix of the transformation matrix can be obtained. Therefore, when calculating the second position information, the bounding box of the first position information is multiplied by the inverse matrix to obtain the second position information.

[0061] Furthermore, the step of cropping each first frame image according to the position of the yarn guide roller of the twisting machine in each first frame image to obtain the corresponding second frame image includes: determining the first pixel coordinates corresponding to the highest point of the yarn guide roller and the second pixel coordinates corresponding to the lowest point of the yarn guide roller in the first frame image; cropping the first frame image with the horizontal line where the first pixel coordinates are located as the first cutting line and the horizontal line where the second pixel coordinates are located as the second cutting line to obtain the corresponding second frame image.

[0062] In practical applications, since the AGV robot moves horizontally, the pixel height of the wire guide wheel in the photo is the same as the height of the second frame image. Alternatively, the height of the second frame image can be set slightly larger than the height of the wire guide wheel to ensure that the wire guide wheel can be fully captured. This slightly larger pixel height setting can be inferred from the method of this application, and therefore should fall within the scope of protection of this application.

[0063] Furthermore, the step of performing target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel includes: inputting each third frame image into a pre-trained target detection neural network model for detection, thereby obtaining the first bounding box that defines the wire guide wheel and the first position information corresponding to the bounding box.

[0064] In practical applications, the object detection neural network uses the YOLO V5 neural network to detect objects and obtain bounding boxes of the selected objects. By decoding the bounding boxes, we can obtain the (x, y, w, h) coordinates of the bounding box, which are the horizontal and vertical coordinates of the top-left corner of the bounding box, as well as the width and height of the bounding box.

[0065] Furthermore, the step of determining the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image based on the preset size transformation rules and the first position information corresponding to each fourth frame image; and calculating the third position information of the twisting machine ternary and the fourth position information of the raw material tube on the corresponding first frame image based on the second position information, includes: calculating the inverse affine transformation matrix based on the preset size transformation rules; calculating the second position information of the thread guide wheel on the corresponding first frame image based on the inverse affine transformation matrix and the first position information; obtaining the first position correspondence between the twisting machine ternary and the thread guide wheel, and the second position correspondence between the raw material tube and the thread guide wheel; determining the third position information of the twisting machine ternary on the first frame image based on the second position information and the first position correspondence; and determining the fourth position information of the raw material tube on the first frame image based on the second position information and the second position correspondence.

[0066] In practical applications, since the resize function is used, the inverse matrix of the affine transformation can be calculated accordingly. Using the bounding box (x,y,w,h) and the inverse matrix of the affine transformation, the position and size of the rectangle in the first frame image can be deduced in reverse, thereby determining the position and size of the wire wheel in the first frame image.

[0067] Furthermore, the step of determining whether the thread guide roller on the corresponding third frame image is located inside the preset area based on each first position information includes: setting a preset area with a width of 120 pixels at the center position of each third frame image; when the center line of the thread guide roller is located inside the preset area, it is determined that the thread guide roller is located inside the preset area.

[0068] In practical applications, to avoid distortion of the target due to lens edge distortion or partial loss of the target due to its location in the boundary area, it is necessary to determine whether the feed roller is located in the center area. If the feed roller is not located in the center area, the image can be discarded, and subsequent calculation and processing of the image can be stopped to avoid wasting computer computing resources.

[0069] like Figure 2 As shown, the present invention also provides a target detection system for a low-computing-power-requirement twisting machine, comprising:

[0070] Frame image extraction module: used to use a pre-set inspection robot to capture the working state of the twisting machine and obtain a working video; extract all frame images from the working video and record them as the first frame image;

[0071] Frame image cropping module: This module crops each first frame image according to the position of the yarn guide roller in each first frame image to obtain a corresponding second frame image. The width of the second frame image is the same as the width of the corresponding first frame image; the height of the second frame image is the height of the corresponding yarn guide roller in the corresponding first frame image; based on each third position information, it crops the corresponding first frame image to obtain a corresponding fifth frame image; based on each fourth position information, it crops the corresponding first frame image to obtain a corresponding sixth frame image.

[0072] Position information recognition module: used to adjust the size of each second frame image according to the preset size transformation rules to obtain the corresponding third frame image; and to perform target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel;

[0073] Frame image filtering module: It is used to determine whether the threading wheel on the corresponding third frame image is located inside the preset area according to each first position information. If it is, the third frame image is recorded as the fourth frame image and step S5 is executed; otherwise, the third frame image is discarded.

[0074] Position information calculation module: It is used to determine the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image according to the preset size transformation rules and the first position information corresponding to each fourth frame image; and to calculate the third position information of the twisting machine ternary and the fourth position information of the raw material tube on the corresponding first frame image according to each second position information.

[0075] Equipment status recognition module: Used to identify each fifth frame image using a pre-trained target detection neural network model to determine whether there is an abnormality in the corresponding twisting machine ternary element. If so, an alarm signal is sent. Used to identify each sixth frame image using a pre-trained target detection neural network model to determine whether there is an abnormality in the corresponding raw material pipe. If so, an alarm signal is sent.

[0076] Specific limitations regarding the target detection system for low-computing-power-requirement twisting machines can be found in the limitations of the target detection method for low-computing-power-requirement twisting machines described above, and will not be repeated here. Each module in the aforementioned target detection system for low-computing-power-requirement twisting machines can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0077] In one embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. When the computer program is executed by the processor, it implements a target detection method for a low-computing-power-requirement twisting machine.

[0078] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0079] In one embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps: including:

[0080] include:

[0081] S1. Use a pre-set inspection robot to film the working status of the twisting machine to obtain a working video; extract all frame images from the working video and record them as the first frame image;

[0082] S2. Based on the position of the yarn guide roller of the doubling machine in each first frame image, each first frame image is cropped to obtain the corresponding second frame image; wherein, the width of the second frame image is the width of the corresponding first frame image; and the height of the second frame image is the height of the corresponding yarn guide roller in the corresponding first frame image.

[0083] S3. According to the preset size transformation rules, adjust the size of each second frame image to obtain the corresponding third frame image; perform target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel;

[0084] S4. Based on each first position information, determine whether the thread guide wheel on the corresponding third frame image is located inside the preset area. If yes, record the third frame image as the fourth frame image and execute step S5; otherwise, discard the third frame image.

[0085] S5. Based on the preset size transformation rules and the first position information corresponding to each fourth frame image, determine the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image; based on each second position information, calculate the third position information of the three-element double twister on the corresponding first frame image and the fourth position information of the raw material tube.

[0086] S6. Based on each third position information, crop the corresponding first frame image to obtain the corresponding fifth frame image; based on each fourth position information, crop the corresponding first frame image to obtain the corresponding sixth frame image.

[0087] S7. Use the pre-trained target detection neural network model to identify each fifth frame image and determine whether there is an abnormality in the corresponding twisting machine ternary. If so, send an alarm signal. Use the pre-trained target detection neural network model to identify each sixth frame image and determine whether there is an abnormality in the corresponding raw material pipe. If so, send an alarm signal.

[0088] In one embodiment, step S1 further includes: extracting all frame images from the working video, and extracting one frame image from the video at predetermined intervals, all of which are recorded as the first frame image.

[0089] In one embodiment, the step of adjusting the size of each second frame image according to a preset size transformation rule to obtain the corresponding third frame image includes: using OpenCV's resize algorithm to adjust the size of the second frame images, reducing the width and height of each second frame image to half of the second frame image, to obtain the corresponding third frame image.

[0090] In one embodiment, the step of cropping each first frame image according to the position of the yarn guide roller of the twisting machine in each first frame image to obtain the corresponding second frame image includes: determining the first pixel coordinates corresponding to the highest point of the yarn guide roller and the second pixel coordinates corresponding to the lowest point of the yarn guide roller in the first frame image; cropping the first frame image using the horizontal line where the first pixel coordinates are located as the first cutting line and the horizontal line where the second pixel coordinates are located as the second cutting line to obtain the corresponding second frame image.

[0091] In one embodiment, the step of performing target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel includes: inputting each third frame image into a pre-trained target detection neural network model for detection, thereby obtaining the first bounding box that defines the wire guide wheel and the first position information corresponding to the bounding box.

[0092] In one embodiment, the step of determining the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image according to a preset size transformation rule and the first position information corresponding to each fourth frame image; and calculating the third position information of the twisting machine ternary and the fourth position information of the raw material tube on the corresponding first frame image according to the second position information, includes: calculating the inverse affine transformation matrix according to the preset size transformation rule; calculating the second position information of the thread guide wheel on the corresponding first frame image according to the inverse affine transformation matrix and the first position information; obtaining the first position correspondence between the twisting machine ternary and the thread guide wheel, and the second position correspondence between the raw material tube and the thread guide wheel; determining the third position information of the twisting machine ternary on the first frame image according to the second position information and the first position correspondence; and determining the fourth position information of the raw material tube on the first frame image according to the second position information and the second position correspondence.

[0093] In one embodiment, determining whether the thread guide roller on the corresponding third frame image is located inside the preset area based on each first position information includes: setting a preset area with a width of 120 pixels at the center position of each third frame image; when the center line of the thread guide roller is located inside the preset area, the thread guide roller is considered to be located inside the preset area.

[0094] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0095] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0096] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principle of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A target detection method for a low-computing-power-requirement twisting machine, characterized in that, include: S1. Use a pre-set inspection robot to film the working status of the twisting machine to obtain a working video; extract all frame images from the working video and record them as the first frame image; S2. Based on the position of the yarn guide roller of the doubling machine in each first frame image, each first frame image is cropped to obtain the corresponding second frame image; wherein, the width of the second frame image is the width of the corresponding first frame image; and the height of the second frame image is the height of the corresponding yarn guide roller in the corresponding first frame image. S3. According to the preset size transformation rules, adjust the size of each second frame image to obtain the corresponding third frame image; perform target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel; S4. Based on each first position information, determine whether the threading wheel on the corresponding third frame image is located inside the preset area. If yes, record the third frame image as the fourth frame image and execute step S5; otherwise, discard the third frame image. S5. Based on the preset size transformation rules and the first position information corresponding to each fourth frame image, determine the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image; based on each second position information, calculate the third position information of the three-element double twister on the corresponding first frame image and the fourth position information of the raw material tube. S6. Based on each third position information, crop the corresponding first frame image to obtain the corresponding fifth frame image; based on each fourth position information, crop the corresponding first frame image to obtain the corresponding sixth frame image. S7. Use the pre-trained target detection neural network model to identify each fifth frame image and determine whether there is an abnormality in the corresponding twisting machine ternary. If so, send an alarm signal. Use the pre-trained target detection neural network model to identify each sixth frame image and determine whether there is an abnormality in the corresponding raw material pipe. If so, send an alarm signal.

2. The target detection method for a low-computing-power-requirement twisting machine according to claim 1, characterized in that, Step S1 further includes: Extract all frame images from the working video, and extract one frame image from the video at predetermined time intervals, which is recorded as the first frame image.

3. The target detection method for a low-computing-power-requirement twisting machine according to claim 1, characterized in that, The step of adjusting the size of each second frame image according to a preset size transformation rule to obtain the corresponding third frame image includes: The second frame image is resized using OpenCV's resize algorithm. The width and height of each second frame image are reduced to half of the second frame image to obtain the corresponding third frame image.

4. The target detection method for a low-computing-power-requirement twisting machine according to claim 1, characterized in that, The step of cropping each first frame image according to the position of the yarn guide roller of the twisting machine in each first frame image to obtain the corresponding second frame image includes: In the first frame image, determine the first pixel coordinates corresponding to the highest point of the thread wheel and the second pixel coordinates corresponding to the lowest point of the thread wheel; Using the horizontal line where the first pixel coordinates are located as the first cropping line and the horizontal line where the second pixel coordinates are located as the second cropping line, the first frame image is cropped to obtain the corresponding second frame image.

5. The target detection method for a low-computing-power-requirement twisting machine according to claim 1, characterized in that, The step of performing target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel includes: Each third frame image is input into a pre-trained target detection neural network model for detection, thereby obtaining the first bounding box that defines the wire wheel and the first position information corresponding to the bounding box.

6. The target detection method for a low-computing-power-requirement twisting machine according to claim 1, characterized in that, The second position information of the thread guide wheel is determined on the first frame image corresponding to each fourth frame image according to the preset size transformation rules and the first position information corresponding to each fourth frame image; Based on the second position information, the third position information of the three elements of the twisting machine and the fourth position information of the raw material tube on the corresponding first frame image are calculated, including: The inverse affine transformation matrix is ​​calculated according to the preset size transformation rules; the second position information of the through-wheel on the corresponding first frame image is calculated according to the inverse affine transformation matrix and the first position information; Obtain the first positional correspondence between the three elements of the doubling twister and the thread guide roller, and the second positional correspondence between the raw material tube and the thread guide roller; Based on the second position information and the correspondence between the first and second positions, the third position information of the three elements of the twisting machine on the first frame image is determined; Based on the second position information and the second position correspondence, the fourth position information of the raw material tube on the first frame image is determined.

7. The target detection method for a low-computing-power-requirement twisting machine according to claim 1, characterized in that, The step of determining whether the thread guide roller on the corresponding third frame image is located inside the preset area based on each first position information includes: setting a preset area with a width of 120 pixels at the center position of each third frame image; when the center line of the thread guide roller is located inside the preset area, it is determined that the thread guide roller is located inside the preset area.

8. A target detection system for a low-computing-power-requirement twisting machine, characterized in that, include: Frame image extraction module: used to capture the working status of the twisting machine using a pre-set inspection robot to obtain working video; All frame images are extracted from the working video and recorded as the first frame image; Frame image cropping module: This module crops each first frame image according to the position of the yarn guide roller in each first frame image to obtain a corresponding second frame image. The width of the second frame image is the same as the width of the corresponding first frame image; the height of the second frame image is the height of the corresponding yarn guide roller in the corresponding first frame image; based on each third position information, it crops the corresponding first frame image to obtain a corresponding fifth frame image; based on each fourth position information, it crops the corresponding first frame image to obtain a corresponding sixth frame image. Position information recognition module: used to adjust the size of each second frame image according to the preset size transformation rules to obtain the corresponding third frame image; and to perform target recognition on each third frame image to obtain the first position information of the corresponding wire guide wheel; Frame image filtering module: It is used to determine whether the threading wheel on the corresponding third frame image is located inside the preset area according to each first position information. If it is, the third frame image is recorded as the fourth frame image and step S5 is executed; otherwise, the third frame image is discarded. Position information calculation module: It is used to determine the second position information of the corresponding thread guide wheel on the first frame image corresponding to each fourth frame image according to the preset size transformation rules and the first position information corresponding to each fourth frame image; and to calculate the third position information of the twisting machine ternary and the fourth position information of the raw material tube on the corresponding first frame image according to each second position information. Equipment status recognition module: Used to identify each fifth frame image using a pre-trained target detection neural network model to determine whether there is an abnormality in the corresponding twisting machine ternary element. If so, an alarm signal is sent. Used to identify each sixth frame image using a pre-trained target detection neural network model to determine whether there is an abnormality in the corresponding raw material pipe. If so, an alarm signal is sent.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 7.

Images