A Comparative Study of Casuarina equisetifolia Number Extraction Methods Based on UAV Visible Light Remote Sensing Data

Study on Casuarina equisetifolia plantation in Pingtan island of Fujian Province. The methods of plantation extraction based on unmanned aerial vehicle(UAV) visible light remote sensing data were compared and investigated for their applicability in plantation extraction of different ages of plantation forests. Firstly, we acquired the visible photos of the study area by a light-rotor UAV, and generated the Digital Orthophoto Map(DOM) and Canopy Height Model(CHM) through pre-processing; secondly, 6 standard plot areas of mature and young forests were selected respectively, and the number of standard plot plants was extracted using the local maximum method(LMC) based on canopy height model, the local maximum method(LMD) based on orthophoto, and the multi-scale segmentation method(MST) based on orthophoto; finally, we compared the number of plants extracted by the 3 methods with the measured data. The results showed that LMC has the highest overall extraction accuracy among the 3 methods, with an F_score of 0.97, while LMD and MST are significantly lower, with an F_score of 0.90 and 0.78, respectively; the extraction accuracy of LMC method was similar for the stems of young and mature forests, with F_score of 0.97 for both young and mature forests; the accuracy of LMD method was slightly higher in mature forests than in young forests, especially the P-values of the 2 different forest ages, as follows: young forests(RE: 0.97, P: 0.81, F_score: 0.88) and mature forests(RE: 0.94, P: 0.89, F_score: 0.91). The accuracy of MST method was significantly higher in mature forests than in young forests, specifically, young forests(RE: 0.88, P: 0.67, F_score: 0.76) and mature forests(RE: 0.88, P: 0.74, F_score: 0.81). It was concluded that LMC had the highest overall strain extraction accuracy among the 3 methods, and it was suitable for strain extraction of different ages of mosaic plantation forests, which could meet the demand of real-time and rapid extraction of C. equisetifolia plantation strain.