图象解决是计算机视觉的重要组成部分，它涉及图象处理，分类器设计和逻辑推理等许多领域。针对目前图象解释系统要面对的严重噪声、模糊性和不确定性问题。重点研究了一种基于基因搜索的双向推理技术，该算法分为如下两步：首先通过基于分割区域统计/几何特征的模式分类器来得到初始的分类模糊隶属度，并根据经验（或统计）得到的先验空间位置关系模糊规则来构造一种有效表达图象解释信息的模糊图。然后通过基因搜索算法融合上面的两类信息来得到图象的最佳解释，实验结果表明，该方法对具有单一对象或多个对象的区域均有很好的效果，也是对基于概率、证据和模糊推理等单向推理机制图象解释方法的提高。

Image interpretation is an important part of computer vision, which is related to many fields such as image processing, classifier designing and logic reasoning. In this paper, genetic searching based two directional reasoning is discussed. The algorithm consists of two steps. At first, the fuzzy memberships of classification is obtained by fuzzy classifier based on the statistic/geometric features of segmented regions, and a fuzzy graph used for effectively representing interpretation information is constructed through prior rule base concerning about spatial relations that is acquired from statistics or experience. At second, genetic searching algorithm is used to combine the above two types of information, and the optimistic interpretation is achieved. In order to decrease the computational cost and increase the possibility of getting optimal solution, a new crossover operator of genetic searching is proposed that is based on non random graph partition. The experiments show that genetic searching based fuzzy image interpretation is powerful for the regions that include one or more objects. This method is an improvement over the one directional reasoning method based image interpretation such as probabilistic, evidence and fuzzy reasonings.