Study On Quantitative Understory Vegetation Cover Of Pinus Massoniana Forest By Combined UAV Active And Passive Remote Sensing

As an important parameter to describe the health of forest ecosystems,understorey vegetation cover is an important indicator to detect soil erosion.Quantifying the understorey vegetation cover in the undulating terrain of the southern hilly region is important for accurate erosion management.UAV multi-angle remote sensing systems and Li DAR remote sensing systems are important tools for quantifying understorey vegetation by virtue of their sensitivity to forest stratification.Changting in Fujian used to be the most severely eroded area in southern China,and years of treatment have significantly improved the local soil erosion situation.However,even though the vegetation cover has increased due to years of mountain closure and reforestation,the lack of understorey vegetation in the area has led to the persistence of undergrowth erosion,to the extent that the quality of the stand is still not high.This paper selects Hetian Town,Changting County as the study area,and constructs a method to quantify the undergrowth vegetation cover in Changting soil erosion area by active and passive remote sensing based on multi-angle（0°,±15°,±30°）remote sensing data from UAV and airborne Li DAR data respectively,and verifies the accuracy of active and passive remote sensing inversion results by combining with the undergrowth vegetation cover data from ground field surveys.At the same time,the factors influencing the accuracy of remote sensing quantification of understorey vegetation were analyzed and the accuracy of the results obtained by the two remote sensing systems were compared to construct the adaptation range of the sample site conditions and the remote sensing system,thus completing the revelation of the adaptation mechanism of the active and passive remote sensing quantification of understorey vegetation in the southern soil erosion area.The main conclusions are as follows.（1）The results show that the vegetation morphology classification criteria proposed in this study are more accurate at small scales（R²=0.99,RMSE=0.552）,and the accuracy of the terrain inversion method meets the requirement of separating the canopy from the ground;multi-angle remote sensing can extract the observation images of the target feature from multiple angles within a similar time period,which improves the information abundance of the understory vegetation and significantly increases the accuracy of quantifying the understory vegetation(R²=0.73,RMSE=0.134,R² _multi-angle is 17%higher than R_0°²;15%higher than R_15°²;20%higher than R_30°².RMSE multi-angle is 18%lower than RMSE_0°by18%;9%lower than RMSE_15°;and 51%lower than RMSE_30°.)（2）The Point CNN-based active remote sensing method for quantifying understory vegetation cover proposed in this study can invert understory vegetation cover more accurately（R²=0.79,RMSE=0.117）;For each stand feature,the Point CNN model segmentation accuracy was better than the Cloth filtering algorithm.（3）The factors influencing the quantification of understory vegetation cover by remote sensing were analyzed in the first stage as densities and slopes,and the inversion accuracy of active and passive remote sensing was analyzed under each condition of densities and slopes respectively.It was found that the estimation accuracy of the understory vegetation under medium to low densities was higher than that under high densities;the quantification of understory vegetation under gentle slope conditions was the best,the quantification accuracy of understory vegetation under steep slope conditions was the lowest,and the accuracy of slope was in between.By comparing the accuracy,it was found that the overall accuracy of active remote sensing quantification was higher than that of passive remote sensing quantification,and the scope of active remote sensing quantification of the understory vegetation was closer to the actual measured values of the understory vegetation.At the same time,the active remote sensing quantification accuracy is better in the area of extreme understorey vegetation cover,while the passive remote sensing quantification trend is more suitable to the measured value in the area close to the average of understory vegetation cover.In summary,the joint algorithm of active and passive remote sensing quantification of understory vegetation was constructed by comparing the active and passive remote sensing quantification of understory vegetation in different conditions.The inversion accuracy of the joint algorithm increased significantly compared with the previous results.(R²=0.85,RMSE=0.101.R_union² increased by 28%compared to R² _multi-angle;12%compared to R²_{Li DAR};RMSE_union decreased by 26%compared to RMSE _multi-angle;11%compared to RMSE_{Li DAR}.)...