复杂背景中的目标辨识方法研究

Identification of Objects in a Complex Background:a Case Study

研究了二维熵准则极值的递推算法，以获取图像分割的最优门限，基于知识和多特征复合的思想，提出了一种具有知觉组织特点的目标辨识方法，给出了识别复杂场景中特定目标的实验结果。

The reason why the recognition of an object in a complex background is very difficult inmachine vision application is that the object can hardly be extracted intact from its back-ground.Afast recursive algorithm for calculating the extreme value of the two-dimensionalentropy criterion to determine the optimum threshold of image segmentation is studied.Based on the oUect knowledge and multifeature fusion,a new method for object identifica-tion that features the perceptual organization is proposed.The intact extraction and definiteidentification of the object are achieved almost at the same time. Experimental results of i-dentifying some specific objects in a complicated scene are given.