RESEARCH FOR THE APPLICATION OF DEEP LEARNING COMBINED KINECT SENSOR IN OBJECT CLASSIFICATION

Abstract

In this paper, an objective classification system using a Kinect sensor and
deep learning model to process object images is proposed. The deep learning
model is carried out with the data collected from reality to give features such as
shape and color. Besides that, using the Kinect sensor also helps to acquire depth
information of the object easily. Objects of different properties are tested such as
circles, squares, triangles, red, blue, and yellow with different heights. The
proposed system is capable of classifying the characteristics of each object with
high accuracy and low cost. Experimental results show that the objects are
classified with an accuracy of 92% with the average time to recognize each object
being 50ms. Thereby demonstrating the practical applicability of the proposed
system in object classification.