基于FPGA和神经网络的激光图像处理系统研究

Study on Laser Image Processing System Based on FPGA and Neural Network

针对激光图像数据量大、传输困难的问题,提出了一种基于FPGA和神经网络的激光图像处理系统。研究了系统的硬件组成并对系统中的FPGA模块和存储模块等关键模块进行了分析设计,提出了基于FPGA的神经网络实现方法。实验结果表明:该方法能够实现对激光图像的可靠压缩。其压缩效果与隐含层中神经元的个数有关,神经元个数越多,压缩图像的信噪比越高。当神经元个数为8时,信噪比达到了30d B,能够较好地满足应用要求,因此一般选择隐含层中神经元数为8。

Aiming at the problem of large amount of data and transmission difficulty about laser image,a laser image processing system based on FPGA and neural network is proposed. The hardware components of the system are studied,the key modules of the system such as FPGA module and saving module are analyzed and designed,and the implementation method of the neural network based on FPGA is proposed. Experimental results show that this method can achieve reliable compression of the laser image. The compression effect is related to the number of neurons in the hidden layer,the more of the neurons,the higher of signal to noise ration of the compressed image. When the neuron number is 8,the signal to noise ratio reached 30 d B,and can meet the application requirements well,so can choice 8as the number of neurons in hidden layer.