Research Of Parameter Extraction Of Individual Tree And DBH Inversion Based On UAV Images

Forests can regulate the climate,conserve water,and maintain water,which provides an important support for the balance of the Earth’s ecosystem.Protecting forest resources is an inevitable requirement for mankind to achieve sustainable development.At present,our country has continuously increased the protection and supervision of forests.The structural parameters of individual tree include: DBH,tree height,crown wide,etc.It is the basic work to determinate this parameters for forest resource survey.The determination of DBH parameters are especially important,which is the key to accurately calculate the forest volume and biomass.The traditional individual-tree structure parameters are mainly obtained by manual measurement using tools such as a chest diameter gauge or a belt ruler.With the development of the society,these methods gradually exposed problems such as high labor cost,large measurement error,and data lag.Therefore,how to quickly and accurately obtain the basic parameters of a individual tree has become an important issue in current forest resource management.At the same time,the emergence and popularity of new technologies such as Li DAR and UAV have provided researchers with new ideas.Based on the above background,this study choose Ginkgo as the research object,images are obtained by a UAV equipped with a high-resolution camera,and then extracts parameters such as canopy area,crown wide and tree height.The three-parent diameter inversion model,the binary DBH inversion model and the ternary DBH inversion model were established with these three parameters as independent variables.The main research contents are as follows:(1)Obtaining the image data of the individual ginkgo sample by the UAV,and generating the Digital Orthophoto Map(DOM)and the point cloud data of study area by the Pix4 D software using the motion recovery structure(SFM)method;(2)The multi-scale segmentation algorithm is used to effectively segment the individual ginkgo tree crown by e Cognition Developer Based on the two-dimensional DOM,and the canopy area(CA)and crown wide(CW)values of the tree are extracted;The point cloud data is reconstructed in three dimensions by the ENVI LIDAR plugin,and the tree height(H)value of the single wood ginkgo is extracted by using the sliding box algorithm;(3)After obtaining the basic parameters of crown area(CA),crown wide(CW)and tree height(H),establish a one-dimensional breast diameter inversion model(CA--DBH,CW--DBH,H--DBH),binary DBH inversion model(CA&CW--DBH,CA&H--DBH,CW&H--DBH)and ternary DBH inversion model(CA&CW&H--DBH).The degree of fitting of each model is measured by two factors: coefficient of determination(R2)and root mean square error(RMSE),and the model accuracy is tested by the average error value.The results showed that the CA&CW&H--DBH ternary DBH inversion model had the highest coefficient of determination(R2)of 0.825;the root mean square error(RMSE)was the smallest,only 0.9591;19 groups of samples were tested.The average error of the DBH value of the CA&CW&H--DBH binary DBH inversion model is 0.75 cm,and the error rate is 4.20%,which is less than the allowable error value(5%)of the A-type forest resources.On the whole,the ternary DBH inversion model is the optimal DBH inversion model in this study.It also proves that the parameters of canopy area,crown wide and tree height can be extracted by UAV data,and higher precision of DBH can be calculated.It provides new ideas for automated forest resource survey and monitoring.