The Lnfrared Moving Object Recognition Based On Mult--Templates

Infrared moving object recognition is an important problem in image sequent analysis. It is necessary for surveillance application, for guidance of autonomous vehicles, for smart tracking of moving objects, and many other applications. The work of this thesis-- the recognition of infrared moving object based on multi-templates is focused on the extraction of invariant feature under multi-templates. The main content of this thesis is summarized as follows. Chapter 1 gives a brief introduction of the aim of this work as well as the development on the subject of infrared moving object recognition. And for the more, the organization of this thesis is introduced. Chapter 2 introduces the research background. And then provide a review of exiting algorithms of image processing and image recognition. Chapter 3 is focused on the construction of a new feature vector. The characteristics of this approach is that combined an affine invariant moment shape descriptor of the contour of the part brighter image with the infrared image features to recognize the infrared objects based on multi-templates. The moment is scale, rotation, and translation invariant. It is reflection invariant as well. And the others are features from infrared objects based on several JR features. The algorithm is computationally inexpensive and robust even when the object is partially occluded. The experimental results of this algorithm are presented using the infrared data, which is accordance with the conclusions drawn from the theoretical analysis. Chapter 4 investigates the rotation and translation invariant?-D circularly integrated bispectrum (2DCJB). The circularly integrated bispectrum preserves both the phase and scale information of the one- dimension signals. The algorithm is applied to the recognition of infrared moving object based on templates and exhibits superior performance to the Fourier magnitudes in the respect of both normalized correlation and recognition accuracy. Finally, chapter 5 concluded the thesis with a summary of the conclusion drawn from each individual chapters, and suggestions of further research directions based on the work of this thesis.