Research On Image Recognition Technology In Monitoring Distribution Network Equipment Condition

There are a lot of distribution network equipment,and widely distributed,for a long time have not been the attention of the relevant departments,leading to the aging distribution equipment,and the technology is relatively backward.As the demand for electricity increases,the disadvantages in the distribution network gradually emerge.With the continuous development of technology,on-line monitoring of power equipment has been widely used in the power systems such as substations and power plants,which has played an important role.The traditional power patrol not only consume a lot of manpower and resources,but also because of the subjective factors of the inspectors lead to misjudgment of the state of the equipment,and because of the harsh environment where many power distribution equipment,resulting in daily inspection is not easy,and has security risk.Therefore,this paper proposes an on-line monitoring system of distribution network equipment based on image recognition technology,to save manpower and resources while greatly improving the inspection efficiency.This paper first introduces the state monitoring system of the distribution network equipment.Secondly,the paper describes the traditional image processing techniques,including image preprocessing,image feature extraction,SIFT-based image matching and state recognition.Then this paper mainly introduces image processing technology in special environment,including image de-fogging and low-illumination image enhancement.According to the limitations of traditional algorithms,the improved algorithms about haze removal algorithm using dark channel prior and image enhancement algorithm using Retinex.In image matching,aiming at the high complexity of traditional SIFT algorithm,this paper proposes an improved image matching algorithm.Finally,this article selected the drop-down switch,vacuum circuit breakers and air switch,these three kinds of distribution network equipment for image recognition processing,to achieve the status of three kinds of equipment to determine.Experimental results show that the three improved algorithms proposed in this paper can improve the processing efficiency while ensuring the processing efficiency,can be applied to the condition monitoring system of distribution network equipment.At the same time,this paper proposes three different identification methods for the different distribution network equipment characteristics,and obtains good judgment results in the experiment,which provides a train of thought for the subsequent condition monitoring of the power network equipment.