Research On Infrared Ima Ge Segmentation In Zero Value Insulator Detection

The insulator has the function of mechanically fastened and electrical insulation, and is widely used in transmission and distribution networks. Porcelain insulator has a large share in current market, and some of them will become zero value insulators in the long run, which will lead string breakage or the power line drop failure and cause huge economic losses to the national economy. Therefore, the detection of zero value insulators is a very important work.A zero value insulator detection method by combining relative temperature distribution characteristics of insulator string with artificial neural network was verified in intelligent artificial climate chamber. This method has numerous advantages such as non-contact, objective, safe, and is worth to be widely used. In the detection process of this method, the insulator infrared image segmentation is the foundation, and it has a significant influence on the result of the method. Co mpared with the insulator infrared image shot in laboratory, the image shot on site has some disruptors such as transmission lines, towers, connection fittings, joint board and these disruptors bring much difficult for image segmentation. At present, an effective insulator string zone extraction method is lack for image shot on site. This paper introduces an insulator string zone extraction method, which has considerated disruptors and makes the above zero value insulator detection method available on site.According to the temperature characteristic of normal insulators, zero value insulators and atmosphere, Otsu double thresholds segmentation method is used. This method has good segmentation result, but it takes too much time, which is bad for the efficiency of field detection. Therefore, the tabu search algorithm is used in image segmentation, and Otsu function is used as the evaluation function. The segmentation result shows that this method can greatly accelerate the speed of segmentation and keep the same image segmentation effect.The binary image obtained by the above method contains several disruptors, while the final research objective is just the insulator string zone, so the insulator string extraction is needed. The height curve of insulator string is similar to sinusoid, it reaches maxima in somewhere of insulator and it reaches minimum in somewhere of steel cap. And insulator string extraction is just based on this characteristic. The steps of insulator string zone extraction are as follow:1) In order to get the height curve of insulator string, it needs to correct angle of the image to make the insulator string axis parallel to the horizontal. Firstly, do skeleton extraction and hough transformation on the binary image to get the angle between insulator string axis and horizontal and the position of insulator string, and rotate the image according to the angel.2) Based on the position of insulator axis, we can specify some part as research object. Then, to get the insulator string height curve and its maxima sequence. In this sequence, the number of the value which are obviously large or small is less, most of them are close to the insulator height, so the median of the maxima sequence is used to indicate the maximum height of insulator string and the research object shrinks.3) Finally, to get the modified height curve and its maxima sequence, proper operation is needed to make one insulator corresponding to one maxima value, combing with minimum sequence of the insulator string height curve, we can process both ends of the insulator string and extract the insulator string zone.The above insulator infrared image segmentation method can effectively remove the interferences, and it lays the foundation for the actual application of this zero value insulator detection method by combining relative temperature distribution characteristics of insulator string with artificial neural network.