Research On The Method Of Close-range Photogrammetry Stumpage Measurement Based On Monocular Vision

Based on the technology of Monocular Vision,Close-range Photogrammetry and Digital Image Processing,this paper proposed a non-contact method which was used to obtain the DBH information.The main tasks of this method included that the threshold segmentation of color stumpage image,the camera calibration of the image acquisition device,the extraction of DBH's terminal coordinates and distortion correction,and the DBH measurement of individual stumpage and neatly arranged stumpages.The main work of this paper is as follows:In order to overcome the shortcomings of traditional 2D Otsu threshold segmentation method such as poor anti-noise capability,ignore the class cohesion,exist much background information in target after segmentation,this paper proposed an improved 2D Otsu algorithm which take into account between-class distance and the within-class distance,it combined the average variance concepts of two categories so that realize the accurate segmentation of color image.After preprocessing the original image,we calculated the grayscale-neighborhood average gradient histogram firstly,then we recursively calculated the probability,mean vector and average variance of the target and background,obtained the optimal threshold vector according to the threshold selection formula.The threshold segmentation of image was carried out by using threshold vector,and the segmentation result was optimized by removing small areas connected region and digital morphological processing.Aiming at the problem in the traditional Zhang's calibration algorithm that had large amount of calculation and only considered the influence of the nonlinear equation on the camera model caused by radial distortion,this paper proposed an improved Zhang's calibration algorithm.First of all,the algorithm reduced the computation complexity by introducing normalized homography matrix,then on the basis of the established internal constraint formula,the internal matrix of camera was calculated by using the least squares.The external matrix of camera was solved by using the relationship between the normalized homography matrix and the interior and exterior matrix of camera,and the two order radial distortion and tangential distortion were introduced at the same time,then we used the mathematical model with nonlinear distortion factor to calculate the distortion factor which in the model of camera.Finally we optimized camera interior and exterior parameters and distortion factor with maximum likelihood estimation to improve calibration accuracy.Since the method of obtain the DBH's terminal coordinates was mainly by using the image processing software which lead to the large change of the coordinate value,this paper proposed a method to automatically obtain the DBH's terminal coordinates.In this method,we calculated the linear equation which used the two feature points of the stumpage's DBH firstly,and then searched for the intersection point of the linear and terminal's binary image so as to find the initial pixel coordinates of DBH's terminal.Finally,the distortion correction of the coordinate points were completed by the distortion coefficient obtained after the camera calibration.On the premise of that the stumpage and the calibration target where was regarded as the same plane,this paper realized the DBH measurement of individual stumpage and neatly arranged stumpages.For the DBH measurement of individual stumpage,after the steps of image segmentation and camera calibration,we used the camera imaging model to backcalculate the three-dimensional spatial coordinates in the case of known the pixel coordinates,and then calculated the stumpage's DBH.But at the same time,the error in this non-coplanar situation will be amplified.So the error correction model is used to make the error correction of the DBH in this paper.For the DBH measurement of neatly arranged stumpages,we used the measurement method of individual stumpage to calculate the DBH of stumpae which hung the calibration target in the first place,and then others DBH was solved according to the DBH and others accounted for the proportion of the pixels.