Research On Binocular Visual Positioning Algorithm In Forests

As an important part of the terrestrial ecosystem,forests play an important role in maintaining the earth's carbon balance,regulating the climate and soil and water conservation,and are closely related to human survival and development.Accurately obtaining the standing tree parameters such as the number of trees,the DBH and the position of single trees in the forest plots in resource surveys,and monitoring the changes of the forest are of great significance to forest resource management.Compared with the traditional method of acquiring every wood inspection rule used in forest sample surveys,photogrammetry technology improves the work efficiency in forest survey tasks and improves the ability to obtain tree measurement factors.However,due to the problems such as the occlusion of trees in the understory plot and the limitation of the effective range of the camera,the information of the plot obtained from the single frame image is very limited.In order to efficiently acquire the complete data of the sample plot,the experiment adopts a mobile acquisition method to capture the image sequence of the sample plot in all directions and at multiple angles,and uses the visual SLAM(Simultaneous Localization And Mapping)positioning principle to calculate the position and camera of each frame of the image at the time of acquisition in real time.attitude.According to this,the point cloud data recovered from each frame of the stereo image is converted into the starting coordinate system to obtain a complete sample point cloud map.Finally,the standing tree parameters were extracted from the constructed sample map,and the measurement results of the total station were compared to verify the positioning accuracy of the binocular vision algorithm in the understory environment.The main research is as follows:Binocular stereo vision system.Explore the imaging rules of binocular cameras and the reasons for choosing binocular cameras as sensors in forest resource survey tasks.Finally,the internal parameters of the camera are calculated according to Zhang Zhengyou's camera calibration method,and the mapping relationship between the pixel coordinates and the three-dimensional space coordinates in the image is determined.Binocular stereo camera depth measurement.This chapter mainly introduces the experimental data collection work.The plot is composed of three-square experimental plots with a side length of 40 m.The dominant tree species are Mongolian oak,larch and Pinus tabulaeformis.The forest stands are dense,the canopy density is greater than 0.8,and it has the characteristics of forest land environment with large coverage area and obvious relief.According to the image preprocessing,gray difference accumulation window matching and false matching elimination processing flow,the correlation between the pixels to be matched and the candidate pixels in the left and right images is weighed,so as to achieve pixel matching between the left and right stereo images.Finally,according to the camera baseline distance and the principle of similar triangles,calculate the distance from the target point P to the camera imaging plane.Binocular stereo camera pose calculation in real time.According to the order of feature extraction,feature matching,pose estimation and pose optimization,according to the texture information of the surrounding environment such as the ground and trees,the spatial position and attitude of the camera sensor are determined in real time during the movement.Comparing the measurement results of the total station,the root mean square errors of the control point coordinates along the X,Y,and Z axis directions are 0.81 m,0.65 m,and 0.58 m,respectively.A total of 4322 frames of high-resolution stereo images were collected and processed during the experiment.The processing speed of 3948 frames(91%)of the images was 0.06 to 0.08 seconds per frame.The processing speed of the images was higher than the camera acquisition frequency.The algorithm met the requirements of real-time processing.The binocular stereo camera reconstructs the sample location cloud.Combining the single-frame stereo image point cloud reconstruction and the binocular stereo camera real-time calculation results,through continuous transformation between coordinate systems,a complete point cloud map of the plot is reconstructed.It can be seen that the reconstruction effect of the point cloud map covers a wide area,dense quantity,rich texture information,and high recognizability,which can meet the needs of tasks such as breast diameter extraction,tree species identification and single tree positioning in sample survey,especially for field work Forest information that is difficult to measure,such as measuring the diameter of the trunk at any height,the height of the branches,etc.,is more convenient and efficient.