中文摘要：

D—S证据理论是一种决策级信息融合方法，能够综合各信息源。为了提高对飞机序列图像进行识别的能力，增加信息可信度，提高系统容错能力，提出一种不变矩与D—S证据理论相结合的方法，对飞机序列图像进行了识别。通过构造飞机图像的模板库，对于待识别的图像提取不变矩特征，提出了两种通过不变矩特征构造基本置信函数，然后运用D—S证据理论进行融合，得到图像的识别结果。通过实验，比较了两种不变矩特征和三种构造函数对识别结果的影响，并将其结果与支持向量机、神经网络的仿真结果相比较。实验结果表明方法是有效且可行的，识别概率优于其它的方法。

英文摘要：

D -S evidence theory, which is a kind of image fusion method on a decision level and capable of combining data of all information sources, can increase information reliability and improve system fault - tolerant ability. This paper combines moment invariant and D - S evidence theory and makes plane sequence image recognition effectively. This paper constructs the template base of plane images, extracts moment invariant characteristics of images to be recognized, and proposes two kinds of methods constructing Basic Belief Assignment, then uses D - S evidence theory to fuse and get the recognition result. This paper compares the effects of two kinds of moment invariant characteristics and three kinds of constructed functions. Comparing with the results of Support Vector Machine and Neural Network, it is obviously that evidence theory method is more effective to 3 - D target recognition.