The Study Of Recognizing Technology Of Aircraft Objects Staying At The Airport In Aerial Images

It is the important research topic that to recognize automatically the targets in aerial image in fields of image processing and pattern recognition. It processes highly academic and actual application value. This paper will build up an intelligent software system based on basic data source from aerial images with the help of mankind's visual system and judging experience.With the development of the vision technology of the computer, especially the development of pattern recognition, it is possible to make the identification of aerial images and photograph into the intelligent direction. Therefore, it seems very urgent to apply high and new technology to image processing, image recognition, image transmission system quickly and accurately, using computer intellectual control method and means, collecting automatically and processing the information of reconnoitering. This can improve the automatic ability to reconnoiter. Because the goals in aerial images are very small and the objects images contact with around complicated background, such as being even and fuzzy and out of shape of illumination of the picture. So it is very difficulty to realize the automatic goal recognition system. It is a complicated integrated software systemï¼›It widely involves image-processing, pattern-recognition, artificial intelligence and computer vision etc. At present it is a hot and difficult problem. We have set up a system of intelligent computer software based on person's vision system and reconnaissance experience to solve some problems in interpretation with slow speed, long periods and without any automatization. In the military field, the plane is the important target, and one of the main attacking targets too. So it is particularly important to utilize Aerospace image to differentiate the type, position and quantity of the plane parked in airport. We select aircrafts which have clear geometry shape, obvious recognized feature as the main goal to set up the automatic recognition system. Through selecting effective features, we find out the object position quickly and reliably from the aerial image and recognize and analyze them. An automatic object recognition and intellectual interpretation system has been set up finally.The study has realized each stages of pattern course, including image input, image preprocessing, object segmentation, feature selecting, object classification and recognition, database inquiring and managing and information obtaining etc. Among them, object segmentation, feature selecting, object classification and recognition are the main point.First, adopting the filter technology of wavelet transform to preprocess the image. It can run out of the noise in image through the filter technology of wavelet transform. It's based on edge detection method. The wavelet transform has offered the method of more resolution. It can provide the information of time and space frequency at the same time. Therefore, when the wavelets smooth the noise, it possess double characteristic in space and frequency. It can overcome the shortcomings of Fourier method and others. On the other hand, noises have remarkable difference. Fully utilizing these characteristics, it can get out of noise in the wavelet. The multiwavelet is made up of scale functions and wavelet functions and has larger freedom. It is easy to construct according to necessary nature. This paper processes the multiwavelet transform coefficients by threshold method. It can denoise and filter. The experiment proves that this method has a better denoising performance to Gauss noises and a better ability of protecting edge at the same time.Second, segmenting and detecting the image according to self-adaptive segmentation and fuzzy difference filter based on local extreme value to go on to picture and goal orientate. Segmenting and detecting the object image is the foundation of feature selecting and recognition. Most research works all are assumed that the objects in the image have already been checked. In order to design an automatic recogni...