Satellite imaging is the order of the day. Automatic oil tank detection is one of the important domains in satellite image processing which could be used for disaster monitoring, oil leakage, etc. It has great significance for the military and civilian application to detect oil tanks and locate oil depots automatically from remote satellite images. Automatic oil tank detection in satellite image remains a challenging problem. This paper addresses this problem using enhancement of algorithm for oil tank detection with SURF technique and SVM classifier approach. The proposed approach consists of four stages are pre-processing, segmentation, feature extraction and classification. Initially, the input image is pre-processed using threshold method. Secondly, SURF technique applied on the preprocessed image for the purpose of segment the oil tank. Then, well known features are extracted from the segmented image. Finally, to classify the oil tank and non oil tank support vector machine (SVM) is applied on the extracted feature. To test the proposed method the experiment is carried out the standard benchmark databases are SPOT-5 satellite images, QuickBird satellite images, GeoEye-1 satellite images and Google Earth images. The classification results obtained from the SVM classifier shows the efficiency of proposed method by high oil tank detection rate and low false alarm rate.