Abstract: For the underwater multi-targets localization problem of sonar sensor，a new proximate sensor for underwater condition is proposed and the design is based on the analysis and test of ultrasonic traveling and echoing in water.The working principle and components of proximate sensor are introduced in detail.Different from the geometric interpolation method proposed before，a novel multi-targets localization method based on BP neural network is presented，which has the advantage of no limitation on target number.Two kinds of simulation experiments are performed：One is region localization and the other is accurate angle localization.The simulation results show that the method proposed is efficient and has enough accuracy.

Key words: underwater, proximate sensor, BP neural network, localization

摘要： 针对水下声传感器的多目标定位问题，根据水下超声传播与反射原理研制了一种新型接近觉传感器。详细介绍了水下接近觉的工作原理及组成，并对基于此传感器的多目标定位进行了仿真研究；与目前国内外所采用的几何相交定位技术不同，所采用的BP神经网络定位方法突破了目前方法中定位目标数目的限制，通过区域定位和角度精确定位的两类仿真实验表明了所采用方法的有效性及定位精度。

关键词: 水下, 接近觉, BP神经网络, 定位