Abstract: Dangerous pests are severe forest disasters, and they occur every year in large areas and cause great damage to forest ecology and tremendous loss to economy. It has great significance to detect the occurrence of the hazard in time and to monitor its spreading process. The purpose of this study is to probe into the possibility to recognize red-attack pine trees with high resolution imagery, so as to provide a quick and efficient methodology for monitoring, prevention and control of fatal forest pests. The recognition of red-attack pine trees was studied based on the WorldView-2 multi-spectral data using the analysis method of the quadratic classifier. Through the Fisher determination analysis, we found that the red, green and blue bands have rich information of red-attack pine trees. Based on this discovery, we first conducted classification with all the eight bands, then with only the red, green and blue bands, and did comprehensive analysis. Results show that the accuracies of the two classifications are nearly the same, which means that the red, green and blue bands are enough to recognize the red-attack pine trees, so that the dimension of raw imagery can be reduced. This also contributes to the new application areas of GF-2 imagery. Because of the complexity of forest phases, the selection of image resolution is very important in the recognition of individual trees. Researches show that the 2-m resolution remote sensing data can well be used in extracting the red-attack pine trees. Moreover, it is better to choose data in the vegetation growing season, so that the discoloration mixture phenomenon can be reduced and the accuracy and efficiency of monitoring can be increased.