Research On Early Warning System For Object Close To High-voltage Power Line Based On Computer Vision

Computer vision is the science and technology using computers to simulate biological explicit or macro visual function, it is an important branch of computer science and artificial intelligence research field, in recent years, since the computer vision technology is widely applied in parts identification and count positioning, intelligent visual surveillance, moving body visual navigation, medical image analysis, virtual reality, remote education, remote sensing image analysis, television film production and other areas, computer vision technology has received considerable attention of academic and project developers in related field.The application of computer vision technology has great significance in the field of early warning, but early warning system in the field of high-voltage power line is very rare. Previous studies on early warning systems are basically detection alarm after the disaster occurs and did not play a role in early warning preventive measures, if there is a system which can intelligent recognize and proactive warn operators and pedestrians who enter the non-secure operating distance, alert operators and pedestrians away from the non-secure area, it is an important means to reduce such accidents. The system is designed on the basis of the accident occurrence scene, for the purpose of avoiding accident.While designing a power early warning system, a key step is spatial positioning.Currently, most spatial orientation techniques based on binocular vision are spatial positioning technology based on binocular parallax. Binocular disparity is mainly based on the principle of triangulation to obtain three-dimensional geometric information. Measurement area is a triangle formed by two camera and the targets in front of the camera. into the difference in the triangulation principle is a key step, but the difference is obtained through a series of algorithms,which is very complex, for that we propose a simple camera binocular positioning technology, the main steps are as follows:firstly, create a perspective projection imaging model similar to the human visual system,determine two space linear through two camera and the target object based on the location of the target in the calibration plate, since the object is got by finding the median point,the two space straight lines basically do not intersect, so it is have to find the public vertical line of two space straight line, the midpoint of the public vertical line as a target point.Finally, obtain the distance between target point and high-voltage power lines.The main work is done as follows:Firstly, give a brief introduction to research status and significance of the power early warning systems, and conduct a summary of some work and on the research status of the power early warning systems done by predecessors including infrared-based electrical equipment fault detection,the position monitoring of the switch disconnectors, transformers fault monitoring and using computer vision technology to complete the transformer fault monitoring. Secondly, for the target detected object obtained from the video images, the article briefly describes several basic methods to detect prospects, as well as background modeling method. And summarized and compared the current most widely used single Gaussian background modeling, mixture Gaussian background modeling method, laid the foundation for the following algorithm theory.Then, we give a brief introduction to the basic knowledge of camera imaging principle,related content of distortion correction and several camera calibration methods, and propose a simple binocular vision of space-based positioning technology based on these concepts and methods, the technology utilized the method for solving linear public space perpendicular surface the midpoint to calculate from the actual distance between target point and high-voltage power lines.Finally,we use the former basic knowledge and theory to guide the study of the early warning system for the high-voltage power lines to prevent the foreign object close. The system consists of input modules, the distortion correction module, the target foreground detection module, the target position-location module and alarm module, and the paper describes the function of each module. In Chapter Five, the paper analyzes the experiment systematically, the experiment show that in certain scenarios the method can accurately and effectively detect the target, and real-time alarm reminder when the target gets into the alert zone...