Performance analysis of SVM with quadratic kernel and logistic regression in classification of wild animals

In an attempt to develop a system to classify the wild animals using image processing and classification techniques, we study the usage of Haralick textural features are used in wild animal classification which is a computer aided pattern recognition system. The Haralick features from two wild animal classes that include leopard and wildcat are extracted to from the image database. Support Vector Machine (SVM) with quadratic kernel function model and Logistic Regression (LR) model are developed and tested using the created dataset. In each case, the performance of the classifier is measured.We also compare the performances of SVM and LR with and without pre-processing the dataset using Principal Component Analysis (PCA). This study reveals an increment in the accuracy post pre-processing of the dataset.