A textural modification of the MMAC algorithm for individual tree delineation in forest stand using aerial bitmap images

This paper modifies the multi-level morphological active contour (MMAC) algorithm to process image texture information from aerial color photographs for the delineation of tree crowns. Firstly, the aerial bitmap RGB image is converted to CIELAB. Only the lightness value is used in the BUE (bottom up erosion) process to locate tree seeds in the image. Secondly, these levels are used to detect possible tree locations. Thirdly, false candidates are filtered out based on the local texture by vector quantization. After tree candidates have been verified, individual tree crowns are delineated using the MMAC algorithm. Results showed that the tree crown detection power of the MMAC algorithm was 94% and the omission rate 6%. Aerial color photographs have good spatial resolution of less than 0.5 meters which offers good potential for the accurate delineation of complex tree crowns in mountainous forest stand. By considering textural-based photographic data, the MMAC algorithm was able to more effectively filter out the false-positive recognition of tree crowns and this increases the generalizability of the algorithm.