Tree Cover Detection

Study Area (click to enlarge image)

Tree cover detection results (click to enlarge image)

Traditionally, substantial fieldworks that are very laborious are needed to obtain reliable canopy cover. There is a clear need for a quicker but still accurate canopy cover estimation techniques for a large-scale forest inventories such as the NFIs. The increasing availability of data with high spatial resolution has emerged as a potential data source for forest inventory. 

Aims

The aim of this project is automatically obtaining tree/non tree decision in order to be further applicable to automatic and contiguous forest delineation for Swiss NFI.

Method

  a)  DSM generation by NGATE
Next-Generation automatic Terrain Extraction (NGATE) based on stereo matching approaches can be used for high resolution terrain extraction and surface modeling. It represents a quantum leap forward in image processing automation.

  b) Apply image segmentation methods, e.g. mean shift segmentation
Mean-shift method is a non-parametric feature-space analysis technique. The algorithm is a robust feature-space analysis approach which can be applied to discontinuity preserving smoothing and image segmentation problems. It can significantly reduce the number of basic image entities, and due to the good discontinuity preserving filtering characteristic, the salient features of the overall image are retained.

  c)  A Hierarchical Segmentation Approach for Tree/non-Tree Detection
Shadows may occlude smaller crowns, and also the understory may be spectrally similar to the crowns. These effects generally leads to the underestimation of the number of trees observed. To be able to solve the problem mentioned above, a hierarchical segmentation was implemented in the study by taking advantage of the available infrared and NDVI images.

Contact

• Zuyuan Wang