Object Recognition and Classification of 2D-SLAM using Machine Learning and Deep Learning Techniques

Reviewing two-dimensional simultaneous localization and mapping (2D-SLAM) studies in these decades, many researchers focused on the algorithm enhancement for real-time localization and mapping. The related techniques of 2D-SLAM have been investigated deeply. However, most of the researches focus on the SLAM. Less concentration is put on 2D grid map object recognitions and labeling. Therefore, this paper dedicates to integrate recent popular machining learning techniques with 2D-SLAM technology to come out with an application for 2D object segmentation, feature extraction, as well as pattern recognition. Based on a given 2D grid map and a couple of pre-trained patterns, a clustering method and a machining learning based pattern recognition were presented. Experiments show that the proposed process is able to provide satisfactory object identification accuracy.