Automatic Scene Classification Based on Gist Feature and Extreme Learning Machine

At present, with the expanding of the image database, scene classification based on machine learning gained much more attention among insiders. But this problem poses many difficult challenges. To overcome these challenges, we have to find a proper feature and an effective classifier. Some research found that Gist feature matches with the action of human visual system perfectly. Some scholars chose Support Vector Machine (SVM) to be the classifier, but when doing multi-class classification, method using SVM tends to be complicated and sensitive to the variation of parameters. Our work proposed a new scene classification method using Gist feature and generating classifier with Extreme Learning Machine. By testing our algorithm and comparing it with method using SVM, we demonstrated that our algorithm is effective and feasible, and have much more stable performance than SVM method.