The Study Of Pattern Recognition Algorithm Based On Complexity Image

In our daily life, it is necessary to perform statistical analysis on passenger-flow in public places. The estimation of passengers flow has been used more and more frequently all over the world because of its economic and social benefits. The main idea of traditional statistical methods for passengers flow is based on infrared imaging and laser detection. These methods have some defects, such as weak anti-interfere capacity and inaccurate identification result. This article takes a research on the passengers flow statistic algorithm based on image information. It is difficult to describe the target in vehicle images because of complex background and dramatic changes of lighting. Meanwhile, the human pattern changes randomly, as a result, the traditional recognition methods can easily cause the confusion of targets. The target recognition algorithm based on clustering algorithm is more efficient in dealing with the target recognition in complex background. The main innovation of this paper including:(1) Novel Clustering-based Algorithm for image feature extractionThis article proposed a novel chain code termed Angle Chain Code and a new method to measure the difference between straight lines'directions. Based on these, a clustering algorithm was proposed to extract the main image edges. This new edge feature extraction method is much faster than the Hough Transform Algorithm.(2) Design of vehicle image classifier based on clonal selection clustering algorithmInspired by the clone selection theory in the natural immune system, a new classifier design method based on clone selection clustering algorithm was proposed, this method can overcome the default of multi-constraint conditions and the local optimization of traditional optimization algorithms. First, the DBSCAN (Density-based Spatial Clustering of Application with Noise) algorithm was used to set up an initial antibody group to reduce the redundant clustering data. Then, combined with the clone selection principle, this article described the immune character of vehicle images. Through the proliferation and variation of antibodies, the affinities between antibodies and antigens were matured, and the antibody diversity was ensured. This new algorithm can get the global optimal solution and obtain the best partition of data sets. The experimental results indicate that the classification accuracy of this new algorithm is above 95%. It is better than the traditional clustering algorithms, and the new algorithm is effective and has practical value.