基于蚁群规则挖掘算法的多特征遥感数据分类

蚁群算法作为一种新型的智能优化算法,已经成功应用在许多领域,然而应用蚁群优化算法进行遥感数据处理则是一个新的研究热点。蚁群规则挖掘算法是基于分类规则挖掘进行分类,能够处理多特征的数据。因此,论文将蚁群规则挖掘算法应用到多特征遥感数据分类处理中,并采用北京地区的Landsat TM和 Envisat ASAR数据作为实验数据,对选择的遥感数据进行了多特征分类实验。实验结果分别与最大似然分类法、C4.5方法进行对比,分析表明:1)蚁群规则挖掘算法是一种无参数分类的智能方法,具有很好的鲁棒性,2)能够挖掘较简单的分类规则;3)能够充分利用多源遥感数据等。它可以充分利用多特征数据进行土地覆盖分类,从而能够提高分类的效率。

Study on the multi-feature remote sensing data classification based on ACO rule mining algorithm

Remote sensing data classification is an important source of land cover map, and remote sensing research focusing on image classification has long attracted the attention of the remote sensing community. For several decades the remote sensing data classification technology has gained a great achievement, but with the more multi-source and multi-dimensional data, the conventional remote sensing data classification methods based on statistical theory have some weaknesses. For instance, when the remote sensing data does not obey the pre-assumption of normal distribution, the classification result using Maximum Likelihood Classifier (MLC) will deviate from the actual situation, and the classification accuracy will not be satisfied. So in recent years, many artificial intelligence techniques were applied to remote sensing data classification, aiming to reduce the undesired limitations of the conventional classification methods. Ant colony algorithm as a novel intelligent optimization algorithm has been used successfully in many fields, but its application in remote sensing data processing is a new research topic. Due to the ant colony rule mining algorithm based on classification rule mining, it can process multi-feature data. This paper introduces the theory and flow of application of ant colony rule mining algorithm in multi-feature remote sensing data classification. This paper selected Landsat TM and Envisat ASAR located in Beijing area as experiment data for land cover classification based on ant colony rule mining algorithm, and the classification results are compared with MLC and C4.5. The experimental results indicate that the advantages of ACO used in multi-feature remote sensing data classification can be summarized as follows: (1) It does not assume an implicit assumption for processing dataset; (2) Contextual information can be taken into account; (3) It has strong robustness; (4) It can construct simple classification rules; (5) It can take advantage of multi-source dataset for land cover classification. So the ant colony rule mining algorithm has provided a new approach for multi-feature remote sensing data classification.