Forest inventory in the digital remote sensing age
Applications of sampling theory together with the technical developments in the field of remote sensing have opened new paths in forest inventory. This paper presents an overview of ongoing research in the field of automatic feature extraction and pattern recognition, which may provide options towards a fully automated forest inventory system. Forest information systems have to provide sufficiently reliable and current information in a suitable form, whilst being both efficient and economical. The methods described, have to be seen as a first step towards an automated forest inventory system. The results presented are satisfying for the specific forest situation, but cannot be generalized for other situations and forest types. More comprehensive research on a variety of stand types, tree species and age classes is required, as different feature extraction approaches may be required on different sensors and different image scales. Given a specific need for information the cost of providing the data should not be higher than the benefit achieved by using modern techniques. In the South African plantation forestry situation with homogenous stands, short rotation and relatively extensive silvicultural operations, stand-oriented image analysis approaches could provide sufficient information at acceptable costs.