基于机器视觉的立木树干边缘检测及边缘直线提取

Edge Detection and Edge Line Extraction of Standing Tree Trunk Based on Machine Vision

以互补金属氧化物半导体(CMOS)获取的立木图像为研究素材,基于图像处理技术和机器视觉技术,通过边缘检测算子进行了立木树干图像边缘特征提取,并对立木树干边缘直线特征进行了研究。结果表明:Prewitt算子和Canny算子虽然对于树干背景的边缘细节保留较多,但对于树干边缘检出率较高,并且树干边缘连续性好; Robert算子和Prewitt算子树干边缘的检测误差较大,Sobel算子、Lo G算子和Canny算子对于树干边缘的检测精度较高;分别将Sigma值为0. 1,0. 3,1. 4和3的高斯滤波的立木图像进行边缘检测,并将直径像素距离提取值与目视判读值进行比较,结果表明Sigma值为0. 3～1. 4时的识别偏差较小。

Based on the image processing and machine vision technology,the edge feature of the tree trunk image was extracted by the edge detection operator,and the linear feature of the tree trunk edge was studied.The results show that:Prewitt operator and Canny operator retain more edge details on the trunk background,but the detection rate of the trunk edge is higher,and the trunk edge continuity is better.The detection error of the edge of the tree between the Robert operator and the Prewitt operator is large.The Sobel operator,LoG operator and Canny operator have higher detection precision for the edge of the trunk.The Gaussian filtered standing images of Sigma 0.1,0.3,1.4 and 3 are edge-detected and the diameter pixel distance extraction value is compared with the visual interpretation value.The results show that the recognition bias of Sigma is 0.3~1.4.