Research Of Feature Points Detection And Matching On Digital Image

In the field of digital image processing, the detection of digital image feature points and matching algorithm is becoming an important research topic which combines the knowledge of many fields such as image processing, feature recognition, artificial intelligence, computer science and so on. So it possesses higher practical value. The detection method based on feature is very common in the process of target detection. Image features can generally be divided into three kinds including point feature, line feature, and surface feature. The point feature which is widely used for its small amount of calculation and simple matching. The job we did in the paper includes:In this paper, the multi-analysis had been conducted among Harris arithmetic operators, Moravec arithmetic operators, Plessey arithmetic operators and SUSAN arithmetic operators from detection efficiency, arithmetic operators’ stability, localization accuracy, Anti-noise. The advantage and disadvantage had been compared in detail and the deep reason had also been analyzed profoundly. In addition, the paper introduced a new corner detection way to detect the exact position of the corner based on4-directional edge labeling and corner positioning templates. The specific algorithm and the whole processes had also been described. Experiment results show that the proposed method can detect the exact positions of the real corners, and be useful for the next step.In the aspect of matching algorithm, the paper made some overview of commonly used matching algorithm, moreover, described template based matching algorithm and corner matching algorithm based on singular value decomposition from the algorithm principle and processes. The experiments show the respective characteristics of the two algorithms. At end the paper combined two algorithms to proposed corner matching algorithm based on corner neighborhood pixel gray values.The algorithm combines the grayscale cross-correlation in template based matching algorithm and the corner neighborhood template in corner matching algorithm based on singular value decomposition to match feature points in gray space. Finally, experiments show that the improved algorithm can match the feature points effectively and meet the requirement of real-time.