基于脉冲耦合神经网络的双核视觉导引控制研究

Study on Visually-Oriented Navigation Control Based on Pulse-Coupled Neural Network and Dual-Core Architecture

针对视觉导航自动导引小车AGV高实时要求,提出了脉冲耦合神经网络结合双核(FPGA芯片和DSP芯片)构架的图像处理系统。并针对双核构架提出了二级处理的理念,结合FPGA芯片的特性设计了并行预处理算法,改进了神经网络的图像分割性能,提高了DSP图像识别的效率。实验证明,提出的设计思想及处理算法有效地提高了AGV视觉导航的准确性和实时性。

In allusion to high real-time requirements of visually-oriented navigation AGV, an originality of embedded image processing system with pulse-coupled neural network and dual-core architecture （FPGA and DSP） was put forward. Also a theory of two-degree treatment for the dual-core architecture was set forth. A parallel gradient method was designed to improve the division of PCNN image segmentation performance according to the characteristics of FPGA. The efficiency of image recognition with DSP was improved. Experimental results show that the design philosophy and processing algorithms can effectively improve the capability of accuracy and real-time ability in control of the AGV navigation.