基于机器视觉的物料检测算法设计

Design of Material Detection Algorithm Based on Machine Vision

摘要: 传统的物料检测方法主要采用人工检测，然而人工检测劳动强度较大，工作效率低。本文采用基于机器视觉的图像处理技术，提出了自动检测通过传送带后下落的物料以及其计数的方法，大大减少人工测量、计数的工作量。首先对图像进行预处理，即采用旋转、滤波去噪等方式将原始图像调整为方面后续处理的图像，接着对感兴趣的区域(ROI)进行图像分割，通过对下落背景的形状以及在图像中占据的主要位置的把握，将其截取出来单独处理。得到截取的下落图像后，分析下落背景以及物料灰度值差异，根据其特征特点(圆形度、面积)将物料作为目标从背景中提取分割出来。将分割后的图像通过腐蚀运算将可能连在一起的两个或两个以上的物料分开，再膨胀使其接近原有面积并其所在位置中心点坐标进行记录，最终实现计数算法。

Abstract: Manual detection is mainly used in traditional material testing. However, the method of manual detection is low in efficiency and high in cost has high labor intensity and low work efficiency. In this paper, a new method of automatically detecting the falling materials passing through the conveyor belt and counting them is proposed. The Image Processing Technology which is based on the machine vision is adopted in this new method. Consequently, the workload of manual measurement and counting can be greatly reduced. First, the image is preprocessed, that is, the original image is adjusted to the image for subsequent processing by rotation, filtering and de-noising. Secondly, the Region of Interest (ROI) is segmented. By grasping the shape of the falling object and the main position it occupies in the image, we extract it for further processing. Thirdly, by analyzing the falling object and the difference of material gray value, we segment from the background as a target according to its characteristics (roundness and area). Fourthly, two or more materials that may be connected together are separated from the segmented image through corrosion operation, which is expanded to make them close to the original area. Finally, the coordinates of the center point of their location are recorded to facilitate the final counting.