关键词: 多传感器, 数据融合, 目标识别, 绝对偏差

Abstract: Aimed at the recognition problem of multi-targets with multiple characteristic indexes，a new fusion method for the multi-sensor data is proposed.The method uses the max-min membership function to obtain the index membership matrix.By solving the optimal programming of maximizing the total absolute deviation of the comprehensive membership for all target types，the weights of attributes are derived.Hence，the algorithm of object recognition is given.The method may improve the objectivity and distinguishing degree of target recognition.The example of parts recognition proves that the method is both effective and exercisable.

Key words: multi-sensor, data fusion, object recognition, absolute deviation