Benda Referensi sebagai Acuan Penyederhanaan untuk Deteksi Benda pada Kondisi Terhalang dengan Metoda Support Machine

Abstract

Object in an unobstructed condition is a very difficult condition to recognize. This is due to the overall size of imperfect objects visible. For robot applications, of course this will be very annoying. This condition will lead to errors in decision-making for robots. On the other hand, the crowd of objects that are too much in the robot work environment will complicate the robot in executing objects desired by the user. Robot will easily recognize the object if the object is in a limited environment and not too much so that the focus of his work is more optimal. On too many objects, the probability of the target object will decrease as the number of other objects also allows becoming the candidate as the target object. For that we need an effort to simplify the work environment of the robot so that the focus can be more focused by dividing the environment that is too crowded into several parts. One of the benchmarks in dividing the environment is a reference object. This research will focus on the discussion and the role of reference objects to simplify the environment that is too complicated for robot work. The reference object will be determined by the robot itself from the features with predefined parameters. An important point in this research is how robots can be as intelligent as humans think in determining what references object criteria are used. As cluster determination for reference object used support vector machine method as guide of robot determine class and feature object. The desired end goal is the formation of a simple system in the robot, Simple in this sense is the maximum number of objects 3 pieces.

Keywords

• object detection
• support vector machine
• reference object