Research On Image Abnormality Detection Algorithm For Operating Environment Monitoring Of Power Equipment

The monitoring of the operating environment of electrical equipment is of prime importance to their normal operation. Along with the development of electrical systems, the monitoring of the operating environment of electrical equipment has been shifted from manual work to monitoring equipment. The monitoring equipment can obtain the state of the operating environment through processing and the analysis of the environment information that is collected; the monitoring equipment will send alarm signals if any danger occurs in the operating environment, enabling the staff to take timely measures and ensure the safety of the operating environment of the electrical equipment. This paper offers an investigation of how to detect abnormal conditions in the monitoring images of the operating environment of electrical equipment, identify the staff in monitoring images, detect the breakage of cables and fire in monitoring images with achievements made in the following fields:(1) Identification of the Staff in Monitoring Images of the Operating Environment of Electrical Equipment Based on SVMBecause the staff and their complicated actions in the monitoring images of the operating environment of electrical equipment might influence the normal operation of electrical equipment, this paper first extracts the staff area with a mixture of a Gaussian background model and the histogram characteristics of the edge direction of the staff area, and it establishes the SVM classifier. The SVM classifier is trained by edge direction histogram characteristics in order to identify the staff in monitoring images of the operating environment of electrical equipment. The experiment shows that, compared to the personnel identification model that is based on histogram of gradient features, the personnel identification model in the monitoring image of the operating environment of electrical equipment has enhanced the accuracy rate and recall rate by 8.65% and 22.6%, respectively.(2) Cable Breakage Detection Based on Improved Hough TransformIn order to address the cable breakage caused by aging and strong wind, this paper first enhances and denoises the cable monitoring image and extracts the edge direction histogram characteristics of the cables in the monitoring images through improved Canny operators. These edge images are subjected to erosion and dilation in order to obtain more accurate cable edge images. The straight lines in the cable edge images are found through improved Hough transformation and mapped to the rectangular coordinate system. The cable breakage is detected by calculating the coordinate of line intersections and the angles between lines. The experimental results reveal that, different from the cable breakage detection model based on straight line extraction, the cable breakage detection model based on improved Hough transformation has increased the accuracy rate and recall rate by 33.5% and 31.15%, respectively.(3) Fire Detection in the Monitoring Images of Operating Environment of Electrical Equipment Based on Color Coherence Vector and Wavelet Energy FeatureIn light of the failure of traditional sensors for detecting the fire far from the electrical equipment, this paper extracts the flame area by calculating the color distribution distance in the monitoring images of the operating environment of electrical equipment. After the flame area is enhanced and denoised, the color coherence vector and wavelet energy feature are extracted in order to establish the SVM classifier. These features are used to train the SVM classifier and establish a fire detection model. The experimental results show that the fire detection model based on color coherence vector and wavelet energy feature can increase the accuracy rate and recall rate by 97.655 and 87.5%, respectively.(4) Design and Realization of the Anomaly Detection Prototype System of the Monitoring Image of Operating Environment of Electrical EquipmentOn the basis of the aforesaid research achievements, the anomaly detection prototype system is developed in order to identify staff in the monitoring images and detect cable breakage and fire.