Research Of The Target Detection And Recognition Method For Infrared Image

Infrared imaging system can not only through obstacles such as smoke, dust, fogto detect a target and to realize day and night continuous passive detection, but also toobserve the details of the target, further identification, location and tracking target.Infrared imaging equipment has many advantages of good concealment, detectionability and function of distance etc. So using infrared imaging technology to realize thetarget detection and recognition is the task of research at home and abroad.This article focuses on infrared image segmentation and target recognition andmade the detailed discussion and research. In the infrared image segmentation mainlystudies the clustering method. The numbers of clusters and clustering center have agreat effect on the selection of image segmentation results. The traditional fuzzyc-means algorithm always adopts empirical values as the number of clusters andclustering center. In order to determine the optimal number of clusters, the validitymeasure variable of Global Silhouette Index is employed. The most importantdisadvantage of fuzzy c-means clustering is that it generally does not give properclustering center in the first run. A method which is the minimum-maximum distancebased on the gray value of histogram to compute the original clustering center is putforward. In the infrared image recognition to research this two kinds of characteristicvector of infrared radiation characteristics and based on the moment function andexperimental. The experimental results show that the infrared radiation characteristicsbetter able to describe the characteristics of the target, but to target the translation,rotation and scale transformation it has a certain not sensitivity. Hu invariant distancein continuous meet translation, rotation and scale invariance, but in discrete state thetranslation invariance and rotation invariance also established, and the scale invarianceis not established. Therefore put forward a kind of comprehensive constant distance isused universally for area, boundary, binary images and grayscale images in the discretecase, and conduct theoretical proof and experimental research.