Multilabel Spatial Image Recognition using Deep Convolutional Neural Network

This exhibits multilabel classification and segmentation of remote sensing satellite images through the deep learning framework. Here, the proposed methodology uses multi labelled Land-Mercede dataset and satellite images to perform the classification. The images obtained through satellite are first preprocessed by perfroming the operations like resizing and spatial blurring. In the next step, it performs the classification to classify each object based on the classes trained and finally segmentation is carried out to detect the changes at a particular place in a different time period. This method has achieved an overall classification accuracy of about 98.58% on a test set and least validation loss of 0.0001468 was also achieved by using a proposed model. The result of this approach can be used for more practical applications like urban planning and also to identify illegal activities that take place in restricted areas, forest, etc.. One of the main applications considered here will help to detect changes that happen in land change over time.