Research On The Method Of Forest Height Tetrieval Based On Hemispherical Images

In forest resources investigation,the measurement of standing tree height is very important,which directly affects the volume of stand and the accuracy of the extraction of forest biomass.Aiming at the problems of low measuring efficiency,low measuring accuracy and high price of measuring equipment in the existing methods of measuring the height of standing tree,combining with the fish-eye camera imaging principle,image measuring technology and image processing technology,a method for retrieving stand height based on hemispherical image is proposed.Firstly,a mathematical model for measuring the height of standing tree is proposed based on the equidistant projection theory of fish-eye lens.On the premise of guaranteeing the precision of the model,the internal parameters of the model can be obtained quickly.Secondly,the hemispherical image is processed.In this paper,the method of multiple operator fusion is used to detect the edge of the standing tree in the hemispherical image,and the quadtree algorithm is improved to segment the edge of the standing tree.It not only improves the accuracy of edge detection,but also improves the efficiency of tree contour segmentation.According to the characteristic of quadtree algorithm,this paper proposes using Hough transform to detect the straight line and extract the coordinate information of the extreme point of the standing tree contour.Finally,the coordinate information can be substituted into the mathematical model proposed in this paper to calculate the height of the standing tree height.The hardware device for collecting hemispherical image is composed of a mobile phone equipped with a fish-eye lens and a rangefinder.The image acquisition is simple,easy to operate and the device is portable with low cost.In the calibration experiment of the measurement model,when the actual height of the cabinet is 1855mm,the maximum relative error is 1.30%.In the experiment of height estimation of individual standing tree,the maximum relative error of this method is 3.35%,which is higher than that of the ultrasonic range finder widely used at present.At the same time,the method proposed in this paper can extract the extreme point of multiple standing trees in a single image,and calculate the height of multiple standing trees in the image with the relationship of the geometric position of the trees,so as to realize the inversion of stand height.The method proposed in this paper can meet the requirement of height measurement accuracy for national forest resources survey.Compared with other traditional methods,the accuracy is higher and the measurement result is more stable.At the same time,the single image contains a lot of information about the standing tree,which can measure the height of multiple standing trees at the same time and improve the work efficiency.This method can also provide technical support for the measurement of standing tree factors such as DBH and crown width.