Selective interception of objects in unknown environment autonomously by UAVs is an interesting problem. In this work, vision based interception is carried out. This problem is a part of challenge 1 of Mohammed Bin Zayed International Robotic Challenge, 2020, where, balloons are kept at five random locations for the UAVs to autonomously explore, detect, approach and intercept. The problem requires a different formulation to execute compared to the normal interception problems in literature. This work details the different aspect of this problem from vision to manipulator design. The frame work is implemented on hardware using Robot Operating System (ROS) communication architecture.