Multi-objective Harmony Search Algorithm And Its Application In Traffic Image Segmentation

Since the segmentation results of the traffic scene images have direct influence on the detection and recognition of the road, car, lisence plate, traffic sign and so on, the research on traffic scene images segmentation algorithms is of great importance. Most of the existing traffic image segmentation methods usually consider single factor, what's more, the contents of traffic images are complex and varied, therefore, the segmentation results obtained by these algorithms are not satisfactory. The image segmentation method based on multi-objective harmony search algorithm is able to optimize multiple objectives simultaneously, it has better segmentation result and wider application scope when compared with traditional methods. In view of this, applying this method to traffic image segmentation is a good choice. As the accuracy of the cluster centers is directly affected by the performance of the multi-objective harmony search algorithm, the classification result is also affected by this method. However, the existing multi-objective harmony search algorithms have such shortcomings as the parameters are difficult to set and the convergence of these algorithms is poor. Therefore, the research on the multi-objective harmony search algorithm is of great significance. There are three research aspects:In view of the parameters in harmony search need to be set by users according to experience and problem characteristics, a self-adaptive multi-objective harmony search (SAMOHS) algorithm based on harmony memory variance is proposed in this paper. In the SAMOHS algorithm, a modified self-adaptive bandwidth is employed, moreover, the self-adaptive parameter setting based on variation of harmony memory variance is proposed for harmony memory considering rate and pitch adjusting rate. It not only solves the problem of parameter setting, but also can keep a good balance between global and local search. So the global search and local search abilities have been improved. To solve multi-objective optimization problems, the proposed SAMOHS uses non-dominated sorting and truncating procedure to update harmony memory, the non-dominated solutions found during the search process and its diversity are successfully protected.In view of the traditional fuzzy clustering segmentation algorithms usually only consider a single index, the segmentation results obtained by these algorithms are not satisfactory, a self-adaptive multi-objective harmony search based fuzzy clustering for image segmentation is proposed in this paper. This algorithm combines the SAMOHS algorithm with the fuzzy clustering segmentation technique, it encodes the cluster centers in a harmony vector and optimizes two objectives simultaneously. In addition, it uses a cluster validity index to assess the Pareto optimal set obtained by the algorithm, then the classification result is obtained. The segmentation results obtained by the algorithm are superior to the traditional clustering algorithm, and the algorithm has good stability and good anti noise ability.In view of the intelligent vehicle navigation technology based on vision in intelligent transportation need segmentation algorithm can automatically separate the objects in traffic images, a multi-objective harmony search with gaussian mutation based fuzzy clustering for traffic image segmentation is proposed in this paper. A novel muti-objective harmony search with gaussian mutation is applied to the fuzzy clustering segmentation method in the proposed algorithm. It employs dynamic clustering segmentation strategy and optimizes the global fuzzy compact index and fuzzy separation index simultaneously, and it uses a cluster validity index to assess the Pareto optimal set obtained by the algorithm for the purpose of acquiring the optimal cluster centers. The optimal number of clusters is successfully solved by the algorithm for given minimum and maximum number of clusters. What's more, it is able to accurately separate the target from the background in traffic image.