Ship Detection In High Resolution Optical Image

In modernized war, information dominate is the key point in the whole strategy. While the way to capture image and its recognition is no doubt the critical path to get information.Due to the improvement of satellite and the image detection technology, the significance of automatic quick target detection is getting more and more important.This paper mainly does the research on the high resolution optical image. The high resolution optical image plays important role in multi-media, TV guidance. As its wide observation area, short observation period and high space resolution, it also has been widely used in ship detection. The process of the target detection and recognition can be seen as a template matching and recognition process in the optical image. By comparing the features extracted from the image and the real objects, the result will be accepted when the matching confidence are higher than the metrics.The procedure of the processing is as follows. First, do the preprocessing for the image,including enhancing image contrast, uniform light processing and binarization; Then determine the image location by getting the latitude and longitude information and find whether it is in the prior information. If so, we can use this information library to get an accurate separation of the land and sea. If not, we can also do the following processing with our binarized image. After this, we use three basic features?length, area and aspect ratio to make a rough classification to remove some of the false alarms. Then LBP and HU invariant moment are used to make second classification to further remove the false alarms. After these two steps,most false alarms will be removed. Then in order to make a detailed classification for the ships, other features are used to make the final classification.By analyzing the technical indicators, recognition rate, recognition time and false alarm ratio, the paper shows an effective recognition algorithm to the security of national defense.