Relative performance evaluation of similar supervised classification methods implemented in different geoinformation systems: impacts on the users’ interpretation

In this paper we evaluated the performance of similar algorithms implemented in different geoinformation systems, in order to verify if there are differences among the results of the image classification processes, according to the platform used. We have selected two geoinformation systems commonly used for this purpose: SAGA GIS and ENVI. We have idealized a scenario where a user without formal education in Remote Sensing techniques carried the classification and interpreted the results. Therefore, we have evaluated the impacts of such analyses, which possibly would lead to mistakes. For this reason, we have given equivalent conditions for all algorithms and the systems selected, i.e., we have used the default parameters, for simulating an operator without formal education in Remote Sensing. In fact, the results showed that the same algorithm implemented in different software, generates different results, which impacts directly on the user’s interpretation. Therefore, it was confirmed that there must be attention while selecting a geoinformation system and an algorithm in order to perform an image classification. Otherwise, there will be inequivalent results among the software, which impacts on the user’s capacity of interpretation.