Image recognition-based power tool state automatic detection method and system

Abstract

The application discloses an image recognition-based power tool state automatic detection method and system, relates to the technical field of power tool detection, and comprises the following steps: collecting video data of a power tool in an operation process; then, the surface of the power tool is regionally divided to obtain a plurality of target subareas; then, periodic frame sampling is performed on the video data, and image registration is performed on the sampled frames, so that the positions of the target subareas correspond in different frames; then, differential processing is performed on the registered current frame image and a non-contact reference image to obtain the contact states of the target subareas; the application realizes the transformation from "passive periodical detection" to "active risk early warning", intelligently identifies the contact modes of different regions by analyzing the actual operation video of the tool, and evaluates the wear risk grade according to the contact modes. The system can automatically perform more frequent detection on high-risk regions and prolong the detection period of low-risk regions.

Description

Technical Field

[0001] This invention belongs to the field of power tool detection technology, specifically an automatic detection method and system for the status of power tools based on image recognition. Background Technology

[0002] Electrical tools, especially insulating operating rods, voltage detectors, grounding wires, and other insulating equipment, are critical devices for ensuring the safety of workers during power system operation and maintenance. The integrity of their insulation performance directly affects the safety of operators and the stable operation of power grid equipment. Currently, the industry primarily relies on two methods for condition monitoring of these tools, especially for detecting defects such as insulation wear and aging: One method is periodic manual general inspection, which involves conducting a uniform visual inspection, dimensional measurement, or pressure test on all tools in inventory at fixed time intervals (such as every six months or a year). This method is labor-intensive, inefficient, and highly dependent on the experience and sense of responsibility of the inspectors, carrying the subjective risk of inspection and misjudgment. Furthermore, the "periodic general inspection" model lacks specificity and cannot reflect the rapid, localized wear and tear caused by differences in operating habits and usage scenarios during actual use. Problems are often only discovered at the end of the cycle, resulting in insufficient preventative measures.

[0003] Secondly, there is the simple recording based on video surveillance. Some sites have cameras installed to record the operation process, but the video content is only used as evidence for retrospective investigation. It lacks automatic and intelligent analysis methods, cannot identify the contact status of tools in real time from the operation behavior, and cannot predict potential wear and tear risks.

[0004] Therefore, existing technologies lack a method to intelligently link tool usage with condition monitoring. There is an urgent need for a technology that can automatically analyze operational behavior, identify vulnerable parts of the tool, and based on this, shift from "uniform periodic inspection" to "risk-based precision inspection," thereby significantly improving inspection efficiency and enabling preventative maintenance while ensuring safety. Summary of the Invention

[0005] This invention aims to address the shortcomings of existing technologies by proposing a method that can intelligently associate the tool usage process with state detection. Therefore, this invention proposes an automatic detection method for the status of power tools based on image recognition, including: S1: Collect video data of electrical tools during operation; S2: Divide the surface of the electrical equipment into regions to obtain several target zones; S3: Periodically sample the video data frame by frame, and perform image registration on the sampled frames to ensure that the positions of each target partition correspond in different frames; S4: Perform differential processing on the registered current frame image and the non-contact reference image to obtain the contact state of each target partition; S5: Identify the contact mode of each target partition based on the contact state sequence in all sampling frames; the contact mode includes continuous coverage contact, intermittent dynamic contact, and no / very little contact; S6: Assess the behavioral wear risk level of each target zone based on its contact mode; S7: Perform differentiated thickness detection tasks based on the risk level of each target zone.

[0006] Furthermore, the definition and identification of the contact pattern includes: If the target partition remains in contact for a duration exceeding a preset time threshold during a single continuous contact event, it is identified as continuous coverage contact; If the contact state of the target partition changes frequently and the total contact frequency exceeds a preset frequency threshold, it is identified as intermittent dynamic contact; If the target partition does not meet the above two conditions, it is identified as having no / very little contact.

[0007] Furthermore, the behavioral wear and tear risk levels include: high-risk zones, medium-risk zones, and safe zones; Among them: target zones identified as having intermittent dynamic contact are marked as high-risk zones; Target zones identified as having persistent coverage contact are marked as medium-risk zones; target zones identified as having no / very little contact are marked as safe zones.

[0008] Furthermore, the differentiated thickness detection task includes: setting different detection cycles for high-risk zones, medium-risk zones, and safe zones, with the detection cycles being extended sequentially; when the detection cycle of the target zone expires, performing thickness detection on the zone, and determining whether to issue a warning based on the amount of wear.

[0009] Furthermore, the image registration employs a feature-based registration method, extracting and matching inherent SIFT or ORB feature points on the target object, or performing registration based on fixed reference points in the shooting environment.

[0010] Furthermore, the determination of the contact state includes: Calculate the proportion of foreground pixels to the total number of pixels in the differential image within the target partition, and use it as the contact coverage rate R. If R is greater than the high contact threshold, it is determined to be in a contact state; if R is less than the low contact value, it is determined to be in a non-contact state.

[0011] Furthermore, the high contact threshold is 0.7 and the low contact threshold is 0.1.

[0012] Furthermore, it also includes: obtaining multiple detection results of the same target object under different operating scenarios; if the same target partition is classified into different risk levels in different detections, then the highest risk level is used for final labeling.

[0013] Furthermore, it also includes counting the number of times each target partition was classified into each risk level in multiple tests, and then redefining the final risk level of each target partition according to preset rules based on the number of counts.

[0014] The image recognition-based automatic detection system for the condition of electrical tools uses the aforementioned method to detect the condition of electrical tools.

[0015] Compared with the prior art, the beneficial effects of the present invention are: This application realizes a shift from "passive periodic inspection" to "proactive risk warning." By analyzing actual operation videos using analysis tools, it intelligently identifies contact patterns in different areas and assesses the wear risk level accordingly. The system can automatically perform more frequent inspections on high-risk areas and extend the inspection cycle for low-risk areas. This differentiated inspection strategy based on actual usage behavior can detect potential hazards earlier and more accurately, transforming passive waiting for maintenance into proactive prevention and early warning, greatly improving the level of safety management.

[0016] Meanwhile, this application improves the automation, intelligence, and objectivity of status detection; the entire process, from video analysis, contact judgment, pattern recognition to risk classification and detection triggering, is completed automatically by the system, minimizing errors caused by human intervention and subjective judgment. It solves the problems of low efficiency and poor consistency in traditional manual inspection, making the detection results more objective and reliable. Attached Figure Description

[0017] Figure 1 This is a flowchart of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Please see Figure 1 This application provides an automatic detection method and system for the status of power tools based on image recognition; As an embodiment of this application, it specifically includes: S1: Collect video data of electrical tools during operation; S2: Divide the surface of the electrical equipment into regions to obtain several target zones; S3: Periodically sample the video data frame by frame, and perform image registration on the sampled frames to ensure that the positions of each target partition correspond in different frames; S4: Perform differential processing on the registered current frame image and the non-contact reference image to obtain the contact state of each target partition; S5: Identify the contact mode of each target partition based on the contact state sequence in all sampling frames; the contact mode includes continuous coverage contact, intermittent dynamic contact, and no / very little contact; S6: Assess the behavioral wear risk level of each target zone based on its contact mode; S7: Perform differentiated thickness detection tasks based on the risk level of each target zone.

[0020] As a second embodiment of this application, it specifically includes: Step 1: Data collection; A fixed video surveillance acquisition station is set up, which should cover the target object to be monitored. The target object here refers to the electrical tools involved in this application, especially those tools that require insulation; this is mainly achieved using the following equipment: High-definition industrial camera: resolution no less than 1920x1080, frame rate no less than 30fps, fixedly installed in a position that can completely capture the operation process of the tool.

[0021] Uniform lighting source: A ring-shaped LED shadowless lamp is used to ensure uniform illumination on the tool surface and reduce the interference of shadows and reflections on image recognition.

[0022] Data processing unit: an industrial computer or server equipped with image processing algorithms.

[0023] Step 2: When the operator uses the insulated operating rod to perform relevant simulated operations (such as voltage testing and grounding), the system automatically starts recording the video stream of the entire operation process; Step 3: Divide the target object into regions; The surface of the target object is divided into regions. The surface of the tool is unfolded and then divided into several regions of equal area. The area of ​​each region is preset by the administrator. The resulting regions are marked as target partitions. Step 4: Perform periodic frame sampling and alignment of the video. The recorded operation video is periodically sampled. Frame-by-frame analysis is not performed here because the computational load is too large. Therefore, a fixed time interval Δt is selected to extract image frames from the video stream; Δt can generally be 0.5 seconds here. Since the target object will move and rotate during operation, image registration needs to be performed for each sampling frame. A feature-based registration method is adopted. For example, SIFT or ORB feature points that are inherent to the target object and not easily occluded are extracted and matched. Alternatively, fixed reference points preset in the shooting environment can be used to register the images of the target object in each frame so that each region can correspond one by one in different images. Here, for the convenience of analysis, different target partitions are marked as Mi, where i = 1,..., n, indicating that there are a total of n target partitions. Step Five: Analysis and Classification of Contact Areas Based on Difference Images Perform a difference operation on the aligned current frame image and the reference image to obtain a difference image. Here, the reference image refers to the image of the corresponding target object when there is no occlusion, that is, when it is not in contact with the background. Through thresholding, the regions with significant gray-scale changes in the difference image are binarized into the foreground. A significant difference means that the gray-scale difference exceeds a preset value, thereby obtaining a suspected contact area.

[0024] Determination of Region Contact Status: Traverse each target partition Mi. Statistically calculate the ratio of the foreground pixels in any target partition to the total pixels in the region, denoted as the contact coverage rate R, where 0 ≤ R ≤ 1. Set two thresholds: a high contact threshold Th and a low contact threshold Tl. Generally, Th can be set to 0.7 and Tl can be set to 0.1. If R > Th, it is determined that the target partition is in a contact state in the current frame. If R < Tl, it is determined that the target partition is in a non-contact state in the current frame. If Tl ≤ R ≤ Th, it may be a shadow or a slight contact, and this frame is not counted as an effective contact for the time.

[0025] Definition and Statistics of Three Types of Contact Modes: After analyzing a complete operation process video (such as an operation lasting T = 5 minutes), for each target partition, statistically calculate its contact status in all sampling frames to form a time-series contact status sequence. Based on this sequence, define and identify three types of contact modes: The First Type: Continuous Covering Contact; It means that in a single continuous contact event, the duration for which the target partition remains in a contact state (R > Th) exceeds the time threshold Ts, such as 10 seconds. This type of contact is generally used to represent the behavior of holding something tightly with the hand for a long time. Statistically calculate the number of occurrences C1 of such events in the entire video. The Second Type: Intermittent Dynamic Contact; It means that the contact state switches frequently, the duration of a single contact is shorter than Ts, but the total number of contacts is frequent. Statistically calculate the total number of sampling frames Fc that meet the "contact state", and calculate the contact frequency Fr = Fc / total number of sampling frames. If Fr exceeds the frequency threshold Ft (e.g., 0.3), the area is considered to be an intermittent high-frequency contact area; such contact generally indicates that the corresponding part of the tool repeatedly bumps and rubs against other equipment. Category 3: No / very little contact; target areas that do not meet the above two conditions, that is, there is neither long-term continuous contact nor high-frequency intermittent contact. Step Six: Wear Risk Level Assessment and Dynamic Detection Strategy Generation; Generate a behavioral wear risk level for each region: The target zones with obvious second-type contact characteristics, namely the target zones with intermittent dynamic contact, are marked as high-risk zones; because intermittent dynamic contact is often accompanied by friction and impact, which has the greatest impact on material wear; their risk level is marked as Hi.

[0026] The target zone with obvious first-class contact characteristics, that is, the target zone with continuous coverage contact, is marked as medium-risk zone; continuous static pressure may cause material fatigue and slow deformation, and the risk is secondary; its risk level is marked as Med.

[0027] The third type of contact area, which is the remaining area, is marked as a safe area; it is almost unaffected by usage behavior and has the lowest risk; its risk level is marked as Low. Step 7: Risk-based differentiated thickness detection execution. The system automatically adjusts and executes targeted thickness detection tasks based on the risk level. Detection task triggering: The system maintains a dynamic detection plan table; sets the shortest detection cycle for high-risk partitions, for example, triggering once after being used X1 times; Set a medium detection cycle for medium-risk zones, for example, trigger once after being used 3 times X1. Set the longest detection cycle for the security partition, for example, trigger it once after it has been used 10 times (X1). When the detection cycle for any region expires, the system guides or automatically starts the high-precision thickness detection module. For example, it uses laser triangulation or the aforementioned low-angle optical sectioning method to scan the specific region; it then compares the current average thickness of the region with the initial thickness record to calculate the wear amount.

[0028] If the measured wear in a high-risk area exceeds the safety threshold, the system will immediately issue an early warning, prompting the tool to be taken out of service in advance or to undergo a re-inspection, thereby realizing the transformation from "regular general inspection" to "risk-based precise early warning".

[0029] As a third embodiment of this application, this embodiment is based on the first embodiment. The difference is that in this embodiment, it is necessary to perform several repeated simulations for the designated personnel in different scenarios, that is, different maintenance and repair scenarios. When any target partition is divided into areas with different risk levels at the same time, it is marked according to the higher risk level. For example, if any area is marked as a high-risk partition and a medium-risk partition at the same time, it is marked as a high-risk partition. As a fourth embodiment of this application, this embodiment is based on the first embodiment. The difference is that in this embodiment, it is necessary to obtain the number of times Gi, Zi, and Ai of each target partition Mi being divided into high-risk partitions when different personnel of the corresponding company or unit are simulating work with the help of relevant target objects. Here, Mi is in a one-to-one correspondence with Gi, Zi, and Ai, and i=1, ..., n. The number of times each target partition is delineated is calculated according to the formula, and the specific delineation method is as follows: When Gi≥(Zi+Ai) / 2, the corresponding target partition is marked as a high-risk partition; When Zi≥(Gi+Ai) / 2, the corresponding target partition is marked as a medium-risk partition; When Ai≥(Gi+Zi) / 2, the corresponding target partition is marked as a safe partition; Obtain all high-risk zones, medium-risk zones, and safe zones; The subsequent processing shall be carried out in accordance with the method of Example 1.

[0030] Of course, based on the above four embodiments, this application also provides a system for detecting the status of electrical tools using the method of any one of the aforementioned three embodiments.

[0031] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An automatic detection method for the status of electrical tools based on image recognition, characterized in that, include: S1. Collect video data of electrical tools during operation; S2. Divide the surface of the electrical equipment into regions to obtain several target zones; S3. Periodically sample the video data frame by frame, and perform image registration on the sampled frames to ensure that the positions of each target partition correspond in different frames; S4. Perform differential processing on the registered current frame image and the non-contact reference image to obtain the contact state of each target partition; S5. Identify the contact mode of each target partition based on the contact state sequence in all sampling frames; the contact mode includes continuous coverage contact, intermittent dynamic contact, and no / very little contact; S6. Assess the behavioral wear risk level of each target zone based on its contact mode. S7. Perform differentiated thickness detection tasks based on the risk level of each target zone.

2. The automatic detection method for the status of electrical tools based on image recognition according to claim 1, characterized in that, The definition and identification of the contact pattern include: If the target partition remains in contact for a duration exceeding a preset time threshold during a single continuous contact event, it is identified as continuous coverage contact; If the contact state of the target partition changes frequently and the total contact frequency exceeds a preset frequency threshold, it is identified as intermittent dynamic contact; If the target partition does not meet the above two conditions, it is identified as having no / very little contact.

3. The automatic detection method for the status of electrical tools based on image recognition according to claim 1, characterized in that, The behavioral wear and tear risk levels are categorized into: high-risk zone, medium-risk zone, and safe zone. Among them: target zones identified as having intermittent dynamic contact are marked as high-risk zones; Target zones identified as having persistent coverage contact are marked as medium-risk zones; target zones identified as having no / very little contact are marked as safe zones.

4. The automatic detection method for the status of electrical tools based on image recognition according to claim 1, characterized in that, The differentiated thickness detection task includes: setting different detection cycles for high-risk zones, medium-risk zones, and safe zones, with the detection cycles being extended sequentially; when the detection cycle of the target zone expires, performing thickness detection on that zone, and determining whether to issue a warning based on the amount of wear.

5. The automatic detection method for the status of electrical tools based on image recognition according to claim 1, characterized in that, The image registration adopts a feature-based registration method, which extracts and matches the inherent SIFT or ORB feature points on the target object, or performs registration based on fixed reference points in the shooting environment.

6. The automatic detection method for the status of electrical tools based on image recognition according to claim 1, characterized in that, The determination of the contact state includes: Calculate the proportion of foreground pixels to the total number of pixels in the differential image within the target partition, and use it as the contact coverage rate R. If R is greater than the high contact threshold, it is determined to be in a contact state; if R is less than the low contact value, it is determined to be in a non-contact state.

7. The automatic detection method for the status of electrical tools based on image recognition according to claim 6, characterized in that, The high contact threshold is 0.7, and the low contact threshold is 0.

1.

8. The automatic detection method for the status of electrical tools based on image recognition according to claim 1, characterized in that, It also includes: obtaining multiple detection results of the same target object under different operating scenarios; if the same target partition is classified into different risk levels in different detections, the highest risk level will be used for final labeling.

9. The automatic detection method for the status of electrical tools based on image recognition according to claim 1, characterized in that, It also includes counting the number of times each target partition was classified into each risk level in multiple tests, and then redefining the final risk level of each target partition according to preset rules based on the number of counts.

10. An automatic status detection system for electrical tools based on image recognition, characterized in that, The method described in any one of claims 1-9 is used for the condition detection of electrical tools.

Images