多传感器网络目标检测方法综述

在并行拓扑结构下，从系统级的角度对多传感器网络目标检测方法进行综述，将似然比检测归纳分类为统计量(决策统计量和融合统计量)的确定和门限(决策门限和融合门限)的求解两部分，并分别展开论述。在硬决策融合系统统计量确定方面，分别归纳了理想信道、非理想信道条件下，不同融合统计量构成检测器的检测性能优劣，并通过仿真试验对比分析了不同融合统计量的检测性能；在软决策融合系统统计量确定方面，归纳了软决策融合系统中局部传感器节点性能度量方式，并对比分析了局部传感器节点决策空间划分方法；在门限求解方面，将门限求解方法归纳总结为逼近法、迭代法、蒙特卡罗方法，并分析比较这些方法的适用范围、优缺点等；最后，对多传感器网络目标检测进行了展望。

Target detection with multi-sensor networks:a survey

The decision fusion-based target detection algorithms are reviewed from the technical level for parallel multi-sensor network topologies. The likelihood ratio test method is categorized as determination of the statistics including the local sensor and data fusion center statistics and the thresholds solving including decision thresholds and data fusion center thresholds, and we carry out discussion on this basis. In the aspect of statistics determination with hard decision fusion system, the detection performance of different systems based on different fusion statistics under the ideal channel and the non-ideal channel is summarized and analyzed. The simulation results are also given to illustrate the performance of different fusion statistics. In the aspect of statistics determination with the soft decision fusion system, the performance metric is concluded. Besides, the decision space partition methods of local sensor nodes are compared and analyzed. When comes to the aspect of threshold solving, it can be summarized as approximation, iteration and Monte Carlo simulation. Further, the applications, advantages and disadvantages of these methods are also considered and compared. Finally, further research trends of decision fusion based target detection are proposed.