基于视觉传达的包装样品表面缺陷特征提取系统设计

Design of packaging sample surface defect feature extraction system based on visual communication

为解决多帧包装表面图像复杂背景对缺陷特征提取精度的影响,设计基于视觉传达的包装样品表面缺陷特征提取系统。其中,图像采集模块由光源、CCD线性扫码相机、图像采集模块及A/D模数转换器组成,负责采集包装样品表面图像;FPGA逻辑控制单元产生控制逻辑,采集到的包装样品表面图像传输至DSP数字信号处理器后,利用小波变换方法分解包装样品表面图像内的灰度与细节特征,以包装样品表面图像缺陷前景同背景的分类问题替代包装样品表面图像缺陷特征提取问题,实现视觉传达下包装样品表面缺陷特征提取。实验结果表明,所设计系统对于不同包装表面缺陷类型识别精度基本达到97%以上,平均能耗为0.52 J左右,均显著优于对比系统。

A visual transmission based defect feature extraction system for packaging samples is designed to improve the influence of multi⁃frame packaging surface image complex background on the accuracy of the defect feature extraction.The image acquisition module composed of light source,CCD linear scanning code camera,image acquisition module and A/D analog⁃digital converter collects the packaging samples surface images,and the FPGA logic control unit generates the control logic.The collected surface images of the packaging sample are transmitted to the DSP,in which the gray scale and detail features in the surface image of packaging samples are decomposed by means of the wavelet transform method.The defect feature extraction of packaging sample surface image is replaced by the classification problem of the defect foreground and background of packaging sample surface image to realize the defect feature extraction of packaging sample surface under visual communication.The experimental results show that the identification accuracy of the designed system for the different packaging surface defect types is basically above 97%,and the average energy consumption is about 0.52 J,all of which are significantly better than the comparison system.