基于超声相控阵的物体形状识别系统设计与实现

Design and Implementation of Shape Recognition System Based on Ultrasonic Phased Array

针对强辐射、强粉尘、浓雾等特殊环境下光学设备识别效果不理想的问题,提出了一种利用超声换能器阵列实现对物体形状识别的方法。以超声相控阵技术为基础,设计了一种发射阵列与接收阵列上下交错的排布方式,发射阵列有8个阵元组成,接收阵列有2×10个阵元组成。通过提取各个通道回波信号的幅值,结合BP神经网络各层并行性的特性,利用计算统一设备架构(CUDA)并行模型,加速了训练和识别过程。制作了一款基于超声相控阵的物体形状识别系统原型机,通过多次实验测试,实现了2.78倍加速比,并取得良好的识别效果。

This paper presents a new method based on ultrasonic phased array to solve the issue of shape recognition in situations where the optical devices do not operate effectively,such as strong radiation,thick dust,thick fog and so on. The transmitter and receiver transducers are interleaved vertically,and the transmitter and receiver arrays has eight and 2×10 transducers,respectively. The peak values each of receiver output are input into BP neural network,and CUDA parallel computing model is adopted to accelerate the processing of training and recognition. A prototype system for recognition has been built based on ultrasonic phased array. The experimental results show that this system achieved 2.78 speedup ratio and perfect recognition.