This work presents a vision based object grasping methodology for a robotic manipulator. The method allows the manipulator to grasp a stationary, randomly localized object in any environment. The first part of the work deals with the development of kinematical model viz. forward and inverse kinematics of the robotic manipulator through analytical approach. The second part incorporates Maximally stable extremal regions (MSER) based feature detection method with density based clustering and homography transform to localize the object in real world. The grasping of an object requires detection of its centroid with respect to manipulator base frame. This involves a two-step process, a) determining 3D coordinates with respect to camera frame using intrinsic calibration parameters and pixel values, and b) transformation of camera frame into base frame through extrinsic transformation matrix. The coordinates of centroid are then used with developed inverse kinematics to guide manipulator reach the location of given object. The accuracy and robustness of the algorithm is tested with the robotic manipulator of five degree of freedom. The results obtained prove the effectiveness of grasping algorithm in the real world scenario.