一种基于分块思想的多纸张缺陷一次提取算法

An Algorithm for Extracting Multiple Paper Defects at Once Based on Blocking Idea

中文关键词:  纸病检测  现场可编程门阵列  连通域标记  包围盒

Key Words:paper defect detection  FPGA  connected domain labeling  bounding box

中文摘要:

      为了解决纸张缺陷检测中多个纸病区域的提取实时性问题，提出了一种采用“先分块、后合并”思想直接对实时图像数据流进行处理的并行化纸病提取算法。该算法充分利用现场可编程门阵列（FPGA）的并行处理特性，在跨时钟域对图像数据进行传输的同时，在另一个分支流程中并行地对分块缺陷连通域进行标记以获取每个缺陷的相对位置坐标，最后以包围盒的形式提取出纸张缺陷区域。该算法基于分块思想，分支处理过程不需要消耗主流程额外的处理时间且硬件资源占用率极小。实验结果表明，该算法能够准确、快速地提取出各种常见纸张缺陷，具有很强的灵活性和实时性，能够满足在线纸病检测的实时性要求。

Abstract:

      In order to solve the problem of real-time extraction of multiple paper defect areas in paper defect detection， a parallel paper defect extraction algorithm based on the concept of “first block， post-merge” was proposed to directly process real-time image data stream. This algorithm made full use of the parallel processing characteristics of FPGA. While translating image data between asynchronous clock domains， it simultaneously labeled the block defects connected components in another branch process to obtain the relative position coordinates of each block defects. Finally， the paper defect area was extracted in the form of a bounding box. The algorithm was based on the blocking idea， and the branch processing process did not consume additional processing time for the main flow， and the hardware resource occupancy rate was minimal. The experimental results showed that the algorithm could extract various common paper defects accurately and quickly， and had strong flexibility and real-time performance. It could meet the real-time requirements of online paper defect detection．