A Bayesian Approach To Aircraft Identification By Outline Shape Using Invariant Image Moments
The outline shape is one of the most important attributes in the classification or identification of an object. It is used in human visual perception although the exact processes are yet to be understood. In computer vision system, the object shape is modelled as a mathematical function and characteristic features are then extracted. The features are selected on the basis of invariance to object orientation, rescaling and translation. In this thesis, a set of image moment invariants have been used to represent the outline shape of military aircrafts. The Hu invariant moments have been used due to relative computational ease compared to other equivalent methods and their ability in preserving rotation, scaling and translation invariance. Aircraft satellite images are identified using Bayesian decision theory classification. The classification is based on the statistical properties of a training set. The resulting statistical classifier assigns a measurement to the class which most likely generated the measurement. The classification is supervised for the sake of reduction of identification error rate. The accuracy of the proposed method has been tested by computer simulation experiments using the MATLAB R2009a version. The effect of noise can mask a shape descriptor and render the characteristic feature to be un-usable. The presence of speckle noise in satellite imagery has such effects. In our investigation various satellite images with different levels of noise are filtered using Lee filter, for the extraction of features and the results assessed. The Peak Signal-to-noise Ratio (PSNR) is used to assess the quality of the filtered images. The computer simulation results show that the proposed algorithms are effective in filtering, extraction of features and in the recognition of images. The test images used include 100X100 up to 1024X1024 gray level images.