Object recognition in images

In the field of image processing, object recognition can be termed as identifying a specific object in a digital image or video. Object recognition have immense of applications in the field of monitoring and surveillance, medical analysis, robot localization and navigation etc. The appearance of an object can be varied due to scene clutter, photometric effects, changes in shape and viewpoints of the object. The the recognition should be invariant to viewpoint changes and object transformations, robust to noise and occlusion. This work aims at formulating a technique that consist of two stages. For the first stage, the query image is categorized using a classifier. For classifier optimization we have implemented two types of classifiers- Support Vector Machine(SVM) classifier that make use of GIST features and k-nearest neighbour(kNN) classifier that make use of Scale Invariant Feature Transform(SIFT). GIST based SVM classification is done using different kernels such as Linear kernel, Polynomial kernel and Gaussian kernel. SIFT features are invariant to affine transformations of the images. SIFT features of the images are extracted and a similarity matrix is formed by matching these SIFT features. Then a k-nearest neighbour(kNN) classifier is implemented on the similarity matrix. GIST feature based SVM classifier with Gaussian kernel showed better classification accuracy than SIFT feature based kNN classifier. The image datasets considered for this work are Coil-20P and Eth80.