| Passive millimeter wave security system passively receive object radiating millimeter wave to produce an image. Different objects have different characteristics, so the result of the various types of objects is different when they radiate millimeter wave.As the system doesn’t have electromagnetic radiation to outside it is safe and reliable.Thus it is acceptable by various countries using as security system. In security system,target detection and recognition is the main project, including: image enhancement,target feature extraction and characterization, target detection and recognition and so on.This thesis based on the actual research project, do deeply research on algorithms about passive millimeter wave imaging system of target detection and recognition problem. The main work and conclusion are as follows:1. From passive millimeter wave imaging target detection and recognition system,introduce the system imaging technology. At last, by the radial blur and Gauss blur algorithm, generate virtual passive millimeter wave image. Then we do research on non-linear extrapolation about image enhancement, and verify its feasibility by doing simulation.2. By analyzing and comparing the feature extraction method such as SUSAN operator, Forstner operator, and SIFT operator, do research on feature extraction and description. Then do some pre-digestion and optimization on SIFT operator to solve the large computation to generate feature descriptor. Then to develop the stability in gradient calculation, I do research on the optimization by the steerable filters theory.3. Do research on target detection and recognition in millimeter image, introduce the Gauss mixture model to do estimation of the background. Then introduce the SUSAN descriptor to do feature fusion to improve the detection effect. Then research the target recognition algorithm based on the bag of features model, verify its feasibility in passive millimeter target recognition system, and give an applicable kernel function. |
PDF Full Text Request