应用Harris角点检测进行图像拼接的超长板材尺寸测量方法

Harris Corner Detection for Image Splicing of Long Plate Size Measurement Method

选择约700 mm(长)×200 mm(宽)的10块柞木板材为研究对象;用相机进行标定得到相机内部参数、外部参数,对采集的板材图像进行镜头畸变修正;对图像进行高斯平滑处理后,采用Harris角点检测方法对采集的先后两帧图像进行图像拼接;采用边缘检测得到板材完整的边界信息,通过建立空间坐标系与像素坐标系的转换关系完成板材尺寸计算。结果表明:相机标定与畸变补偿,提高了测量的精度由5 mm至1 mm;图像拼接方法,解决了长尺寸板材机器视觉测量困难的问题;采用Sobel算子边缘检测方法,能够准确保留边缘信息,速度快、计算简单;测量结果最大标准差为0.582 mm,测量精度小于1 mm。

Ten oak wood panels with a length of about 700 mm(length)×200 mm(width)were selected as the research object.The internal parameters and external parameters of the camera were calibrated with a camera,and the lens distortion of the collected plate image was corrected.After the image was conducted by Gaussian smoothing,Harris corner detection method was used to image the first two images collected.Edge detection was used to obtain the complete boundary information of the plate,and the sheet size was calculated by establishing the conversion relation between the spatial coordinate system and the pixel coordinate system.The results show that the camera calibration and distortion compensation can improve the accuracy of the measurement from 5 mm to 1 mm.The image mosaic method solves the problem of long-dimension machine vision measurement difficulty.Sobel operator edge detection method is with the advantages of accurate edge information,fast velocity,and simple calculation.The measurement results are with the measurement maximum standard deviation of 0.582 mm,and measurement accuracy of less than 1 mm.