|
Forest pests is the enemy of forestry development, known as the smokeless fire, its harmfulness self-evident. Pests and diseases is not always a large area into felled occur in the early stages are generally small-scale, decentralized happen. Monitoring for early onset can not be ignored, and if after a long vicious cycle, eventually leading to blockbuster occur, although this time can be monitored, but the damage has been difficult to control. Therefore timely, early monitoring of pests and diseases is important. In this paper, Peony forest farm in Jilin Province as an example, spectroradiometer data, chlorophyll data, remote sensing data, forest map data, GPS data is the main source of information on the subject first disease larch feature spectrum, spectral characteristics and chlorophyll relationships were analyzed and discussed with the measured spectral remote sensing spectroradiometer degree of coupling; Then on this basis, the use of the image data, combined with the GPS data, the larch forest map data to build caducous disease information extracted database, according to the partitions were used to the idea of ??hierarchical classification tree with two different methods of object-oriented caducous disease information extraction; Finally, information extraction accuracy assessment results, and caducous spatial distribution of the disease were analyzed. The main text of the content and conclusions are as follows: a larch disease after by spectral reflectance characteristics change, which is the use of remote sensing techniques for diagnosis and monitoring of forest disease basis. Spectral characteristics of victims and health larch larch spectral differences are mainly in five areas, they are the \absorption band. 2 damaged to different degrees larch reflectance spectroscopy \Three pairs of field measurements corresponding spectral data and SPOT image pixel reflectance values ??for the regression analysis showed that in general they are related, indicating wide-band image reflectivity information not rich and detailed, it reflects the average of a pixel within , can not be completed by a single band information extraction disease early fall, so need for information extraction band synthesis. 4 based on remote sensing data, the feature spectral data, forest maps, GPS data and landscape photo data, build early fall sick comprehensive database of information extraction, integrated database including remote sensing image database, remote sensing knowledge base, supporting information databases and interpretation signs libraries. 5 According to the idea of ??partition hierarchical classification and object-oriented decision tree were used two different methods to achieve caducous disease information extraction, and compared. The results showed that: pixel-based decision tree classification accuracy is low, and object-oriented classification method can not only improve the classification accuracy of remote sensing images, and can effectively avoid the \6 caducous disease by extracting information found to occur early fall sick area of ??2.165 square kilometers, of which 1.6 square kilometers, accounting for slightly injured; caducous disease distribution in the southeast forest into blocks more and distribution, followed by the Northeast direction, the rest are scattered distribution.
|