Object-oriented Forest Information Extraction Research

Forest plays an important role in people's life. It can purify the air, conserve warter, and regulate the ecological banlance. However, as the urbanization process accelerating, the forest resource has been destroyed. It becomes more and more important to monitor the forest resource timely. And in the urban planning, the forest should be protected and dveloped scientifically. This can ensure the development of economy society and environment harmonically.Because the RS is all-day all-weather and timeliness, the RS image has become the important data to obtain the thematic information. At present, only the vegetation is extracted from the RS image based on the visual interpretation or traditional pixel-based approach, including wooldland and grassland. The forest information is not extracted solely most of the time. On the basis of summarizing the status of vegetation information extraction, the paper extracts the forest information using the object-oriented method from three different resolution RS images (ie, ETM, SPOT5, QuickBird image).Firstly, the thesis outlines the history and development of RS, and sumamarizes the information extraction technologies and vegetation information extraction status. Sencondly, it introduces the object-oriented information extraction method, the overview of the study area in Kunming city and the characteristics of the ETM SPOT5 and QuickBird image, and then does ortho generation, fusion and clipping to the three kinds of images. Thirdly, in the study area, it analyses the forest typical characteristics and the best segmentation scales in the three kinds of images, and then studies the forest information extraction used the object-oriented method. Then, it makes the accurancy analysis. Fourthly, it extracts the forest information using the traditional method (maximum likelihood method), and makes the comparison in the forest information extraction results.In the county in Kunming city, using the object-oriented method and the high resolution RS image, the paper extracts the forest information accurately efficiently and economically based on the typical characters and the best segmentation scale. And the forest information can be updated in time. After calculating the forest coverage in QuickBird image, it demarcrates the banned construction area, restricted construction area, and allowed construction area using the forest coverage and the slope. This can assist the urban planning decisions. And it is good for protecting the forest resouse and developing the resouse scientifically.