Research On Image Segmentation Based On Intelligent Monitoring System

Image segmentation is just to segment an image into different sub-images with different characters and get some interested objects. It is a key step from image process to image analysis, plays an important role in image engineering, and is applied in a lot of fields such as computer vision, pattern recognition, medical image and so on. Based on infrared temperature measurement of intelligent monitoring system, this paper focuses on the image segmentation algorithm.Since the thermal imager get a grayscale image and human’s ability to distinguish color image is several times of gray image, Therefore, it is necessary to change the image to pseudo color image. Due to the grayscale in the infrared image has a narrow range, is uneven and often concentrated in a few areas. The code based on the traditional rainbow use the fixed threshold values when mapping to RGB color values, resulting in uneven color and layering difference. To solve this problem, this article use the multi-threshold method。 Further more, since this system have a higher demand on the speed of the algorithm, a algorithm which is based on fast Otsu method is proposed in this article. The results show that due to algorithm processing, image color has high resolution of the human eye and rich color, and meet the meet the requirements of real-time display of the system.Traditional FCM algorithm requires human to specify the number of clusters, and the algorithm is sensitive to the number of clusters, The results of different cluster num-bers are very different. To solve this problem, this paper proposes an adaptive FCM image segmentation algorithm based on gray level histogram.Firstly, according to the information of the image, the algorithm can select the independent peaks according to certain conditions.Set the initial cluster number to be The number of independent peaks.Use a clustering validity index function to evaluate the advantages and disadvan-tages of clustering results.The number of clusters to obtain the minimum value of the function is regard as the optimal number of clusters. The result of experiments on the transformer infrared image segmentation shows that the algorithm can determine the best number of clusters and fault location is distinct on the segmentation of the trans-former infrared image.Different parts of the equipment needs different level of attention. Some important parts or parts where easy to break down of the site need to be set up separately Which is called ROI (Region Of Interest).Currently in the monitoring system, the ROI is set up by the operator who choose some contour points by using the mouse clicking on the monitor screen. But it will lead to the inaccuracy of the ROI and large temperature analysis error.To solve this problem, we propose algorithm based on GPB Operator which apply to extract ROI contour area.Experiments show that the contour made by the algorithm are clear and close, staff can accurately set the ROI having complex contours.Target profile is determined by the operator easy to bring ROI zoning inaccurate, resulting in large temperature analysis error. To address this problem, we propose ex-traction algorithm based on GPB Operator ROI contour area. Experiments show that the algorithm processing contour boundaries is clear and close, staff can accurately set up the ROI which has a complex contour.