Hybridizing Local and Global Features by Sequential Fusion Technique for Object Detection and Learning
In this research paper, we propose a novel framework to identify the object in first Phase and the learning of object is implemented in second phase. In first phase, we used Haralick texture feature as global feature and CART/LDA classifier to detect the object among three classes such as bike, car and human. If the object is classified as human in phase-1, SIFT features were extracted as Local features from human face in phase-2. Six Emotions in Human face like being sad, happy, disgusted, neutral, surprised and angry were learnt and identified using KNN/RF/LDA in Second phase. Most of the Research works done so far have focused on early or late fusion for hybridization of local and global features to classify images, detect and learn about objects. However, the robustness of the algorithm in detecting and learning objects is not highly substantial. Hence, to achieve maximum robustness, in this research work we implement a sequential hybridization of local and global features to classify and learn objects.
Keywords- Haralick Feature, SIFT feature, Sequential fusion, Hybridization.