图像智能化的目标检测技术(Ⅲ)--优化聚类与特征提取

探讨了作为自适应图像目标检测技术层次化数据处理流程的中层聚类和末端识别模块,给出了具体应用的几种实现算法.为加快背景抑制中全局优化聚类分割门限的求解速率,利用最优化计算理论,设计了有效平均梯度剪切的快速操作方法.在特征识别中,结合目标的形态结构研究了保持形状特性的多结构元组合滤波算法.

Image Intelligent Technology of Target Detection(Ⅲ)--Optimal Clustering and Character Recognition

Both optimal clustering module and character recognition module in multistage data processing procedure for automatic image target detection are approached mainly, and several application algorithms are derived. In order to accelerate rate of calculating optimal globally segmentation threshold for clutter suppression, a fast algorithm of effective average gradient cutting operation is designed by using optimal computing theory. In recognition procedure, the combination filtering algorithm with multi structuring elements, which is able to preserve geometric information is approached according to shape structures of targets. Experimental results exhibits that target - background expression model and relative algorithm have better generalization.