基于DSP和FPGA的激光图像处理系统研究

Research on Laser Image Processing System Based on DSP and FPGA

为提高激光图像的处理速度和精度,提出了一种基于DSP和FPGA的激光图像处理方法。通过FPGA实现图像数据预处理,大幅缩减DSP运算量,进而提高了DSP的处理速度。基于脉冲耦合神经网络设计了一种激光图像分割方法,根据图像像素的灰度分布特性以及噪声响应特点自适应调整神经元的关键参数;同时可实现噪声位置神经元行为的抑制;利用最大二维Renyi熵准则确定了一种梯度下降法,可用于确定神经元的动态阈值。为验证所述算法的图像处理效果,文中与最大类间方差分割算法进行了对比。对比结果表明:所述算法的CI值可以达到0. 417,图像分割效果比较理想;分割速度和准确性明显优于最大类间方差分割算法,具有较高的实践价值。

In order to improve the speed and precision of laser image processing,a laser image processing method based on DSP and FPGA is proposed.With FPGA,the image data preprocessing is realized,and DSP computation is greatly reduced,so the speed of DSP is improved.A laser image segmentation method based on pulse coupled neural network is proposed.According to the gray distribution characteristics of image pixels and the characteristics of noise response,the key parameters of the neuron are adaptive.At the same time,it can realize the inhibition of the neural activity of the noise location.A gradient descent method is determined by using the maximum two-dimensional Renyi entropy criterion,which can be used to determine the dynamic threshold of neurons.In order to verify the image processing effect of the proposed algorithm,this paper compares the algorithm with the most widely distributed variance segmentation.The comparison results show that the CI value of the proposed algorithm can reach 0.417,and the image segmentation effect is ideal.The segmentation speed and accuracy is obviously better than the most interclass variance segmentation algorithm,which has high practical value.