Sunlit Background Visible in Remote Sensing Images Controls Vertical Distributions of Stereoscopic Point Clouds in Forested Area
Apr 20, 2023
Forest three-dimensional structure measurement is a key factor in achieving high-precision estimation of vegetation carbon storage. Progress was made on how to use stereoscopic point cloud for forest three-dimensional structure observation, according to the Aerospace Information Research Institute (AIR), Chinese Academy of Sciences (CAS).
Scientists from the State Key Laboratory of Remote Sensing Science found that the proportion of sunlit background visible in remote sensing images controls the vertical distribution of stereoscopic point clouds. The higher the proportion of visible sunlit background, the stronger the ability to detect the understory terrain.
The results were published in Remote Sensing of Environment, providing new insights for the measurement of forest three-dimensional structure using satellite-borne optical multi-angle stereo observation point cloud.
The team realized the unique value of satellite-borne stereoscopic observations for measuring forest three-dimensional structure and carried out in-depth research. They confirmed that leaf-on imagery can depict the structure of forest canopy top, while leaf-off imagery has the potential for detecting understory ground elevations. They also developed a stereoscopic observation model of forest, which was used in the design of the China's Terrestrial Ecosystem Carbon Inventory Satellite, making it the only satellite in China that is capable of both along-track spectral and multi-angle stereo observations.
Scientists investigated the vertical distribution of optical multi-angle stereo observation point clouds in forests by using data from eight different dates, ranging from August with leaf-on to March with leaf-off. During the period, the Sun elevation angle changes from 55.62° to 15.66° under different terrain and forest density conditions.
The results showed that in non-closed forests, the point clouds in early September are closer to the canopy top, while those in August are closer to the ground. On north-facing slopes, the point clouds are close to the ground when there is no foliage and there is sunlight, while even in leaf-off periods of November or January with lower solar altitude angles and no sunlight, the point clouds are mainly distributed in the tree canopy. The proportion of sunlit background in images controls the ability to detect understory terrains, which can explain the difference between multi-temporal point clouds.
The study reveals the vertical distribution features of stereoscopic point clouds with changes in temporal, terrain, and forest density, providing a theoretical basis for using these point clouds to measure the three-dimensional structure of regional forests.
It also provides a new technical approach for accurately calculating carbon storage in deciduous forest areas.
Contact: luyq@aircas.ac.cn
News & Events