Steel belt elevator safety detection method based on computer vision

Abstract

The invention discloses a steel belt elevator safety detection method based on computer vision. The method comprises the steps of 1, collecting video data of a steel belt surface in a steel belt operation area; 2, performing frame extraction processing on the collected video data to obtain images one by one, and performing denoising processing, image enhancement and image alignment on the images; step 3, a YOLOv8 algorithm is utilized to train a model which can learn various abnormal conditions appearing on the surface of the steel strip through the marked steel strip image data set in a supervised learning mode, and the abnormal conditions comprise cracks, abrasion, looseness and fractures; step 4, aiming at each detection result of the steel strip, using a post-processing algorithm to analyze the identified abnormal area, and judging the severity and types of the abnormity, the types including slight cracks, deep cracks and steel strip looseness; if the detected abnormity exceeds a set threshold value, an alarm system is triggered. The elevator safety detection efficiency and precision can be effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of elevators and relates to a safety detection method for steel belt elevators. Background Art

[0002] With the advancement of urbanization, elevators, as an indispensable vertical transportation means in modern buildings, have been widely used in residential buildings, commercial buildings and public facilities. Steel belt elevators, as a special type of elevator system, are widely used in buildings that require high load and high-frequency operation. Steel belt elevators mainly rely on steel belts as driving elements. Compared with traditional wire rope elevators, they have higher stability, tensile strength and longer service life. However, during long-term use, the steel belt may be affected by factors such as wear, corrosion, and fracture, which may lead to elevator safety hazards.

[0003] Currently, the safety detection methods for steel belt elevators mainly rely on manual inspections and some simple mechanical sensors. Although manual inspections can detect some obvious defects through visual inspections or mechanical operations, limited by the professional level, inspection frequency and working environment of the inspection personnel, it is difficult for manual inspections to achieve all-weather and high-precision safety monitoring. At the same time, traditional mechanical sensors can usually only monitor some physical indicators of the elevator (such as the tension and displacement of the steel belt), and it is difficult to comprehensively detect various safety hazards of the steel belt, such as tiny cracks, corrosion marks, fractures, etc. on the surface of the steel belt.

[0004] With the continuous development of computer vision technology and deep learning algorithms, elevator monitoring systems based on image recognition have gradually become a feasible alternative. Compared with traditional methods, the monitoring system based on computer vision can collect image information of the elevator steel belt in real time through high-definition camera devices, and combine image processing and deep learning technologies to automatically analyze and identify various abnormal states of the steel belt, so as to achieve efficient, accurate and all-weather safety detection. However, most current computer vision-based monitoring methods still have limitations, such as insufficient image recognition accuracy, slow detection speed, poor environmental adaptability, etc. There is an urgent need for a more accurate, stable and adaptable safety detection method for steel belt elevators. Summary of the Invention

[0005] In order to overcome the deficiencies of the existing technology, the present invention provides a safety detection method for steel belt elevators based on computer vision, which combines means such as image processing and deep learning to monitor the operating state and safety condition of the steel belt elevator in real time, and can effectively improve the efficiency and accuracy of elevator safety detection.

[0006] The technical solution adopted by the present invention to solve its technical problems is:

[0007] A safety detection method for steel belt elevators based on computer vision, comprising the following steps;

[0008] Step 1, Data Acquisition: Collect video data of the surface of the steel strip in the operating area of the steel strip.

[0009] Step 2, Image Preprocessing: Perform frame extraction on the collected video data to obtain individual images, and then perform denoising, image enhancement, and image alignment on the images.

[0010] Step 3, Target Detection and Analysis: Use the YOLOv8 algorithm to train a model that can learn various abnormal conditions on the surface of the steel strip through supervised learning using a labeled steel strip image dataset. The abnormal conditions include cracks, wear, looseness, and fractures.

[0011] Step 4, Abnormality Detection and Alarm: For each detection result of the steel strip, use a post-processing algorithm to analyze the identified abnormal area, judge the severity and type of the abnormality. The types include minor cracks, deep cracks, and steel strip looseness. If the detected abnormality exceeds the set threshold, the alarm system is triggered.

[0012] Further, in Step 1, install vision sensors in the operating area of the steel strip. The vision sensors include high-definition cameras, infrared cameras, laser scanners, etc. to collect video data of the surface of the steel strip.

[0013] Preferably, the vision sensors are arranged in a multi-angle shooting manner to ensure comprehensive collection of all parts of the steel strip. All parts of the steel strip include: the side, the surface, and the contact surface, avoiding blind spots. In a darker or low-light environment, infrared or night vision technology can be combined to ensure accurate collection of image data under low light conditions.

[0014] More preferably, install vision sensors in the elevator shaft and the car to monitor the operating state of the steel strip and the surrounding environment (such as vibration, foreign object interference, etc.), and support a backend analysis and alarm system to achieve full-range and high-precision safety detection.

[0015] Still further, in Step 2, the denoising process includes using methods such as median filtering and Gaussian filtering to remove noise points in the collected images and maintain image quality. The image enhancement includes contrast adjustment, brightness correction, etc. to improve the detail clarity of the images for subsequent processing. The image alignment includes using image registration technology for image alignment to ensure consistency between images collected at different angles and different times.

[0016] Even further, in Step 4, the alarm system notifies the maintenance personnel for processing via text message, APP, or email.

[0017] The beneficial effects of the present invention are mainly manifested in:

[0018] 1. Through deep learning models and advanced image processing techniques, it is possible to accurately identify the tiny defects on the steel strip, ensuring high-precision and high-reliability detection.

[0019] 2. Through the automated image acquisition and analysis process, the system can achieve all-weather and real-time safety monitoring of the steel strip, significantly improving the detection efficiency and avoiding the omissions and delays of manual inspections.

[0020] 3. By adopting various image acquisition techniques (such as high-definition cameras and infrared imaging), it can adapt to the monitoring requirements under different environmental conditions, ensuring high-precision detection under the influence of various factors such as light and vibration.

[0021] 4. By installing vision sensors at multiple positions, it is possible to achieve all-round and multi-angle monitoring of the steel strip, not limited to the local inspection of a certain position, thereby reducing blind spots and ensuring a comprehensive assessment of the steel strip status.

[0022] 5. The automated detection and alarm system replaces the traditional manual inspection method, which not only reduces labor costs but also reduces the safety risks caused by human errors or omissions. Through real-time alarms, the system can also notify maintenance personnel in a timely manner when serious problems occur with the steel strip, avoiding accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flowchart of a method for detecting the safety of a steel strip elevator based on computer vision. DETAILED DESCRIPTION OF THE INVENTION

[0024] The present invention will be further described below with reference to the accompanying drawings.

[0025] Refer to Figure 1 , a method for detecting the safety of a steel strip elevator based on computer vision, comprising the following steps;

[0026] Step 1, data acquisition: Collect video data of the surface of the steel strip in the operating area of the steel strip; install vision sensors in the operating area of the steel strip, and the vision sensors include high-definition cameras, infrared cameras, laser scanners, etc., to collect image or video data of the surface of the steel strip.

[0027] Preferably, the vision sensors are arranged in a multi-angle shooting manner to ensure comprehensive collection of all parts of the steel strip. All parts of the steel strip include: sides, surfaces, and contact surfaces, avoiding blind spots; in a darker or low-light environment, infrared or night vision technology can be combined to ensure accurate collection of image data under low-light conditions.

[0028] In this embodiment, visual sensors (such as high-definition cameras, infrared cameras, etc.) need to be installed in the running area of the steel belt, covering the surface, side and contact surface of the steel belt to avoid detection blind spots. Visual sensors also need to be deployed in the elevator shaft and car to comprehensively monitor the running state of the steel belt and the surrounding environment (such as vibration, foreign object interference, etc.). Only by supporting with a backend analysis and alarm system can all-round and high-precision safety detection be achieved.

[0029] Step 2: Image preprocessing: The collected video data is frame-extracted into individual images, and then the images are denoised, enhanced, and aligned.

[0030] The denoising process includes using methods such as median filtering and Gaussian filtering to remove noise points in the acquired images and maintain image quality. The image enhancement includes contrast adjustment, brightness correction, etc., to improve the detail clarity of the images for subsequent processing. The image alignment includes using image registration technology to align the images to ensure the consistency between images acquired at different angles and different times.

[0031] Step 3: Object detection and analysis: Using the YOLOv8 algorithm, a model is trained through supervised learning with a labeled steel belt image dataset to learn various abnormal conditions that appear on the surface of the steel belt, including cracks, wear, looseness, and fractures.

[0032] In this embodiment, the model in Step 3 is optimized and configured as follows: In terms of model structure design, based on the lightweight YOLOv8 network architecture, the feature extraction ability for tiny defects on the steel belt surface is enhanced by introducing the attention mechanism SE module, while reducing the number of model parameters to improve real-time performance. The construction of the dataset covers a variety of actual scenarios, including different lighting conditions (such as strong light, weak light, infrared imaging), steel belt running states (stationary, high-speed running), and abnormal types (cracks, wear, looseness, fractures) to ensure the comprehensiveness and reliability of model training. In the data augmentation strategy, methods such as random rotation, brightness jitter, Gaussian noise injection, and affine transformation are used to improve the model's robustness to image distortion and noise; in view of the reflective characteristics of the steel belt surface, mirror reflection simulation enhancement is added to effectively avoid false detection caused by reflective interference. During the training process, the pre-trained COCO dataset model is used as the base through transfer learning technology, and some layer parameters are frozen and the classification head is fine-tuned to accelerate model convergence; the loss function uses the improved Focal Loss to balance the positive and negative sample ratios and solve the problem of the small proportion of abnormal areas on the steel belt.

[0033] Step 4: Abnormal detection and alarm: For each test result of the steel strip, a post-processing algorithm is used to analyze the identified abnormal area to determine the severity and type of the abnormality, which includes slight cracks, deep cracks and loose steel strips. If the detected abnormality exceeds the set threshold, the alarm system is triggered. The alarm system notifies maintenance personnel via SMS, APP or email for processing.

[0034] The post-processing algorithm in step 4 is specifically implemented as follows: First, a morphological closing operation is performed on the abnormal area detected by YOLOv8 to eliminate isolated noise points, and quantitative analysis is performed based on the area, aspect ratio, contour complexity and other features of the abnormal area to distinguish interference items such as cracks and stains. Secondly, based on the principle of temporal consistency, a logical AND operation is performed on the detection results of 5 consecutive frames, and only the abnormal area that persists is retained, thereby reducing the false alarm rate caused by instantaneous noise. In the classification of abnormal severity, the system sets a three-level alarm mechanism: if a slight crack is detected (length <2mm and does not penetrate the surface of the steel strip), a low priority alarm is triggered (notified by APP); if it is a deep crack (length ≥2mm or penetrates the surface), a medium priority alarm is triggered (SMS notification); if the steel strip is detected to be loose or broken (such as a sudden change in local tension or structural separation), a high priority alarm is directly triggered (SMS and buzzer alarm linkage). In addition, the algorithm supports adaptive threshold adjustment function, which dynamically optimizes the abnormal judgment threshold according to the operating environment of the steel strip (such as temperature, humidity) and historical detection data to avoid over-inspection or missed detection caused by fixed thresholds. Finally, the alarm system automatically generates a maintenance work order based on the abnormality type, location and screenshot, pushes it to the maintenance personnel terminal in real time, and stores the complete detection data in the database for subsequent trend analysis and fault tracing.

[0035] In this embodiment, the steel belt elevator safety detection system based on computer vision includes steel belt elevator equipment, visual sensors, video storage devices, video analysis systems and alarm systems. Among them, the visual sensors are deployed in the elevator shaft, car and steel belt running area to capture abnormal conditions of the steel belt; the video storage device is used to store the images captured by the camera; the video analysis system is connected to the video storage device, receives the images stored in the video storage device, and performs real-time analysis of the images through a pre-set intelligent analysis algorithm; when the alarm system receives an abnormal signal from the video analysis system, it immediately triggers the alarm system.

[0036] The steel belt elevator safety detection method of this embodiment applies computer vision technology to the steel belt elevator safety detection, which can not only detect the running status and abnormal conditions of the steel belt in real time, but also initiate emergency measures in the first time, greatly improving the safety of the elevator.

[0037] The content described in the embodiments of this specification is only an enumeration of the implementation forms of the inventive concept and is for illustrative purposes only. The protection scope of the present invention should not be regarded as limited to the specific forms stated in this embodiment, and the protection scope of the present invention also extends to equivalent technical means that can be conceived by those of ordinary skill in the art based on the inventive concept of the present invention.

Claims

1. A steel belt elevator safety detection method based on computer vision, characterized in that: The method The steps include: Step 1, data collection: collect video data of the steel belt surface in the steel belt running area; Step 2: Image preprocessing: extract the collected video data into frames one by one, and then perform denoising, image enhancement and image alignment on the images; Step 3, target detection and analysis: using the YOLOv8 algorithm, a model is trained to learn various abnormal conditions on the surface of the steel strip through the labeled steel strip image dataset in a supervised learning manner, including cracks, wear, looseness and fracture; Step 4: Abnormal detection and alarm: For each detection result of the steel strip, a post-processing algorithm is used to analyze the identified abnormal area to determine the severity and type of the abnormality, which includes slight cracks, deep cracks and loose steel strip; if the detected abnormality exceeds the set threshold, the alarm system is triggered.

2. A steel belt elevator safety detection method based on computer vision as claimed in claim 1, characterized in that: In the step 1, a visual sensor is installed in the running area of ​​the steel strip. The visual sensor includes a high-definition camera, an infrared camera, a laser scanner, etc. to collect images or video data of the surface of the steel strip.

3. A steel belt elevator safety detection method based on computer vision as claimed in claim 2, characterized in that: The visual sensors are arranged in a multi-angle shooting mode to ensure comprehensive collection of various parts of the steel strip, including: the side, the surface and the contact surface; in darker or low-light environments, combined with infrared or night vision technology, it is ensured that image data can still be accurately collected under low-light conditions.

4. A steel belt elevator safety detection method based on computer vision as claimed in any one of claims 1 to 3, characterized in that: Visual sensors are arranged in the elevator shaft and car to monitor the running status of the steel belt and the surrounding environment.

5. A steel belt elevator safety detection method based on computer vision as claimed in any one of claims 1 to 3, characterized in that: In step 2, the denoising process includes using median filtering and Gaussian filtering methods to remove noise in the captured image to maintain image quality. The image enhancement includes contrast adjustment and brightness correction to improve the detail clarity of the image for subsequent processing. The image alignment includes using image registration technology to align the image to ensure consistency between images captured at different angles and at different times.

6. A steel belt elevator safety detection method based on computer vision as claimed in any one of claims 1 to 3, characterized in that: In step 4, the alarm system notifies maintenance personnel via SMS, APP or email for processing.

Images