基于卷积滤波的纱筒定位检测算法

Yarn Bobbin Positioning Detection Algorithm Based on Convolution Filtering

摘要: 纱筒定位检测是自动换筒算法中的首要任务，针对纱筒图像背景复杂、分割困难的问题，本文提出一种基于卷积滤波的纱筒定位算法。首先对图像进行直方图均衡化，增强区域特征，再对传统扩散滤波模型进行优化，抑制纹理等高频信息，接着设计圆形卷积核对图像进行卷积滤波，抑制图像背景，最后对卷积图像进行Blob分析，通过区域特征筛选得到内筒轮廓。在自主搭建的检测平台上使用本文方法进行实验，实验结果证明纱线余量的检测精度在1.5%以内，满足实际生产要求，为纺织产业自动化生产提供一定的依据。

Abstract: Yarn bobbin positioning detection is the primary task in the automatic replacement algorithm. Aiming at the problem of complex background and difficult segmentation of yarn bobbin image, this paper proposes a yarn bobbin positioning algorithm based on convolution filtering. Firstly, histo-gram equalization is performed on the image to enhance regional features. Then the traditional diffusion filtering model is optimized to suppress high-frequency information such as texture. Then a circular convolution kernel is designed to filter the image and suppress the image background. Finally, Blob analysis is performed on the convolution image, and the inner bobbin contour is ob-tained by regional feature screening. Experiments are carried out using this method on the self-built detection platform. The experimental results show that the detection accuracy of yarn margin is within 1.5%, which meets the actual production requirements. It provides a certain basis for the automated production of the textile industry.