针对航空照片中自动目标识别的复杂性．该文提出了一种基于神经网络的自动目标识别算法，

算法包括检测和虚警排除两个阶段．检1删阶段是在整幅区域中i挂行快速搜寻．找出所有可能的目标 虚

警排障阶段是对检测阶段得到的结果进行进一步的验证， 在尽可能保留全部真目标的前提下，将伪目标

排除．实验结果证明了算法的可行性．

关键词 神经网络，航空照片， 自动目标识别-against aerial photographs Automatic Target Recognition of complexity. This paper presents a neural network-based automatic target recognition algorithm, the algorithm including the detection and false alarm ruled out in two stages. Frederick stage is a cut in the whole region i linked to fast-track the search. To identify all possible targets false alarm remove obstacles stage of the testing phase results further verification, reservations at all possible real targets on the premise that goal will be to exclude false. Experimental results demonstrate the feasibility of the algorithm. Keywords neural network, aerial photographs, automatic target recognition