In day-to-day fast growing technology, unmanned aerial vehicle has been the most proposed approach for solving object tracking problems. The challenges of vehicle detection in aerial surveillance include camera motions, panning, tilting and rotation. Airborne platforms at different heights result in different sizes of target objects. The authors design a keypoint merging method for counting number of cars. This paper mainly consists of screening, feature extraction process, keypoint classification, grouping of number of key points, merging the key points. Set of consistent key points can be identified by using feature extraction process.