摘要
为了解决传统人工检测织物疵点存在视觉疲劳、疵点大小受限、检测精度不高等问题,采用TI的STM320DM6446开发织物疵点自动检测系统。实时采集的图像经过加权中值滤波和顶帽、底帽联合变换降噪处理后,滑动分割每一帧图像。疵点检测系统分析图像灰度分布情况,提取图像纹理特征的熵和能量,通过局部熵和局部能量的自适应阈值分割完成织物的疵点检测。系统的软件主要是在TI的CCS中完成,通过串口完成硬件与主机之间通信。通过与大津法(Qstu)比较,实验验证,该检测系统的疵点检测精度高达93%,检测速度快。
The real-time fabric defect system based on TI＇ s STM320 DM6446 was designed to solve visual fatigue,the size of defect limitted,low precision in intradtional manual fabric defect. The realtime captured image weighted median filter and avoidances hat joint transform to reduce noise,slided every frame image. Defect detection system analysed the image gray level distribution,local entropy and local energy were extracted among the texture feature. Basing on local entropy and local energy threshold segmentation,it finished fabric defect detection. System software was mainly completed in TI＇s CCS,a serial port was to complete communication between hardware with host. Compared to Qstu,experiments verified that defect detection precision of the system was as high as 93 % and the speed was fast.
出处
《毛纺科技》
CAS
北大核心
2016年第11期27-31,共5页
Wool Textile Journal
基金
国家自然科学基金(61301276)
关键词
疵点实时检测
局部熵
局部能量
阈值分割
real-time fabric defect
local entropy
local energy
threshold segmentation
