Road anomaly detection using self-supervised label generator with vertical disparity maps
S. V. S. Ramakrishnam Raju, B. Harikrishna, L. Bhagyalakshmi, Sanjeev Kumar Suman · IET conference proceedings. · 2023
Autonomous vehicles and robotic wheelchairs based real world applications need the automated recognition of roads, route holes, and other irregularities. Traditional techniques of image processing are responsible for the inaccurate identification of abnormalities, which ultimately leads to poor performance. As a result, the primary emphasis of this work is focused on the design of a road anomaly detection system, which is based on a Self-Supervised Label Generator (SSLG) and vertical disparity maps (VDM). Here, VDM is used to identify the regular and irregular boundaries of road by using both depth map image. In addition, an anomaly map generator is used to identify the anomalies presented in the road. Finally, supervised machine learning based SSLG approach is used to label the drivable area, unknown area, and anomaly region. According to the findings of the simulations, the performance of the proposed technique resulted in better performance, when compared to the existing methods for all performance metrics.