Security has become major issue now a day. In order to prevent unauthorized access of confidential data there is a need for accurate and reliable personal identification system. So, biometric based identification system is one of the best solutions. Fingerprint based system is one of oldest biometric identification systems. It is used in many commercial and security applications. Even with advent of technology in fingerprint identification system, the accurate extraction and matching of features from a fingerprint image is a challenging task. The task is much more challenging when fingerprint is affected by non-linear deformations such as rotation and translation. In this paper, fingerprint identification system using improved feature vector based algorithm is presented. In the algorithm Gabor filter is implemented to enhance the fingerprint image. The salient features minutiae (ridge endings) and reference point are extracted from the image. The Euclidian distances between reference point and each minutiae point are calculated and are arranged in ascending order. These are stored in database as feature vectors. The fingerprint matching is done based on the similarity rate between the feature vector of input fingerprint and the feature vectors stored in the database. Algorithms are implemented using Visual Studio 2010 in C++ language using Open CV libraries and tested on the fingerprint database created in the laboratory.