Visual Forest Model Reconstruction And Parameter Extraction Using Close-range Photography Data

Forest resource survey is an important part of the national comprehensive monitoring system of forest resources and ecological conditions.After a forest survey,we can grasp the current status of forest resources,which is helpful for the effective protection and scientific management of forest resources.With the digital development and construction of forest management,detailed investigation of forest resources has become an important way to understand the current status of forest resources.Close-range photography technology is an emerging technical means of forestry resources investigation in precision forestry,and is the main direction of current forestry resources investigation and research.This paper mainly studies the method of visual forest model reconstruction and parameter extraction.Aiming at the significant difference between forest and surrounding environment,we innovate on the forest tree feature detection algorithm,and propose a visual forest construction modeling method and process flow that meet the needs of precision forestry investigation.Subsequently,the three-dimensional reconstruction process suitable for forestry images was engineered into a visual forest model reconstruction and parameter extraction system,with a view to improving the efficiency of forestry resource monitoring and providing data support for forest management.In this paper,the forest scene image obtained by close-range photography technology is used to analyze the color and geometric characteristics of the forest tree.The Multi-Scale Retinex image enhancement of the sequence image is combined with the improved Harris and SURF feature detector for image matching to reconstruct the visual forest model.We propose a visual forest model encryption algorithm based on geometric constraints to encrypt the visual forest model,and the tree positions and breast diameter parameters were extracted from the visual forest model.The sample plots data of Wangyedian Forest Farm in Inner Mongolia and Gaofeng Forest Farm in Nanning,Guangxi were used to carry out precision cross-validation,and to explore the influence of terrain on the extraction accuracy of single tree parameters.Finally,using the development framework Qt combined with Open CV,Open GL and other open source libraries to design and develop the corresponding software system under the Visual Studio 2015 IDE.The visual forest model reconstruction and parameter extraction system tests were carried out on the four tree species of eucalyptus,red spine,larch and Chinese pine in the test area.The results show that Multi-Scale Retinex image enhancement can significantly increase the number of forest feature point pairs.The Harris + SURF feature detection operator can improve the detection efficiency while maintaining the feature detection efficiency,thereby reducing the reconstruction time of the visual forest model.The improved point cloud encryption algorithm can optimize the accuracy of the reconstructed model.The extracted tree positions are basically consistent with the measured values,the error is within 0.2m,and the average relative errors of the DBHs of the four tree species are 6%,5%,4% and 10%,respectively.It proves that the method proposed in this paper and the developed software system have great potential for visual forest model construction and parameter extraction,and can provide effective technical support for forest resource investigation and management.