Abstract: [Objective] Identification of forest canopy gap is a prerequisite to quantify the forest gap characteristics (such as size, shape and dynamics), and a basis for further understanding the complex structural forest species regeneration and studying the understory species diversity, in order to study the active and passive remote sensing method for forest gap recognition. This study could provide further technical support for the quantitative analysis of forest gap features.[Method] In this study, true color aerial orthophoto (0.2 m) and airborne LiDAR (3.7 points·m^-2) were used as the active and passive remote sensing data sources, respectively, and northeast typical natural secondary forest-Mao'ershan experimental forest farm Donglin industry zone was selected as the study area for the object oriented gap recognition. Three segmentation schemes (based on aerial image segmentation, LiDAR segmentation, collaboration of aerial image and LiDAR segmentation) were adopted when processing object oriented classification and each segmentation scheme was divided to 10 scales (10-100, step size 10) to find the optimal segmentation scale parameter. Based on the optimal segmentation result, the support vector machine classifier (SVM) with spectral features of aerial image, height features extracted from LiDAR data were used to identify the forest gap.[Result] The optimal scale was 20, and the ED3_modified value of LiDAR data segmentation (0.52±0.11) was always lower than that of aerial image segmentation (0.58±0.07) or collaboration of aerial image and that of LiDAR segmentation (0.58±0.07) over all scales (10-100). Based on LiDAR data segmentation and its optimal segmentation scale of 20, the classification accuracy obtained from integration of spectral (active)+height (passive) characteristic increased 36.71% and 8.17%, respectively, comparing with the classification accuracy obtained from single characteristic of spectral feature and height feature.[Conclusion] Comparing the three segmentation schemes, the result of segmentation based on LiDAR data was the best; the classification accuracy based on integration of active and passive remote sensing was the highest (OA=87.73%, Kappa=0.81).

Key words: forest gap, scale segmentation, classification feature, LiDAR, aerial image, CHM, support vector machine(SVM)