Research On The Application Of Ant Colony Algorithm In The Remote Sensing Image Classification Of Forest

Forest resources are the basic guarantee for human development. The status and changes of forest resources not only affect our environmental, but also influence the sustainable development of social economy. Timely and accurate grasp of information concerning the status and changes of forest resources is an request, and the emergence of remote sensing technology offers the potential for that request. The accurate classification the remote sensing image of forest is the precondition to realize all kinds of application in the field of forest resources, and the selection of useful features is the basis of improving classification accuracy. Therefore, I, with an aim to improve the classification accuracy of the remote sensing image of forest, did some research concerning the key techniques in that realm, and so far have finished the some aspects of work, mainly as follows:1. The research status and development of remote sensing image classification of forest, as well as ACO (ant colony optimization) algorithm, are first introduced. After that, the papers do a detailed analysis about some common features of remote sensing image, and summarize some typical algorithms for feature extraction. Finally, the evaluation system of remote sensing image classification is introduced, and the characteristics of the various bands of Landsat5satellite are analyzed.2. Upon finishing the detailed discussion of the basic principle and characteristics of ACO algorithm, TSP (travelling salesman problem)’s practical application in a specific instance is simulated using the scientific language, Matlab, and its results are briefly analyzed. Taking the disadvantages of ACO algorithm into consideration, and with an aim to improved ACO algorithm, this paper also discusses some of the improved ACO algorithms, and then, with that as basis, comes up with ways to improve them, to make them more suitable for feature selection.3. In order to effectively improve the classification accuracy of remote sensing image of forest, this paper considers the spectrum features and texture features, which are extracted using principal component analysis and gray level co-occurrence matrix, respectively, then puts the improved ACO algorithm into implementation so as to select the optimal feature subset, and also, in combination with SVM (support vector machine), realizes the remote sensing image classification of forest.4. The experimental results show that both adding text features and using ACO algorithm for feature selection can improve the classification accuracy of remote sensing image of forest, which verifies the validity of the algorithm proposed in this paper.