Object recognition and detection plays an important role in understanding the surrounding of an autonomous vehicle. In current systems, information from expensive sensors and cameras are used to create a model that will detect surrounding objects. But advancement in camera technologies can provide us huge sets of data through just a high definition camera. In this research, an object detection system based on a dash camera mounted in autonomous vehicles is proposed. The object detection algorithm proposed model is based on the Viola Jones algorithm and is compared to a convolution neural network approach. Based on the findings, the efficiency and accuracy of the proposed method to detect vehicles was found to be superior to the other neural network approaches. The proposed method used in this paper had an accuracy of 98% compared to 94% for the convolution neural network approach. The proposed method took 35 seconds on average as compared to several minutes for the convolution neural network to detect the objects in one image on the same system for training to testing. The proposed method recognizes the object in the image precisely and it takes less time than the other neural network-based algorithms. The proposed model is trained on low-quality images taken from a low-cost camera, which requires very low computing power, eliminating need for costly components. This low-cost implementation will allow existing vehicles to be fitted with autonomous object detection technology with just a dash camera.
Object detection and classification through a dashcam for an autonomous vehicle
2018-12-01
Sonstige
Elektronische Ressource
Englisch
DDC: | 629 |
Fleet dashcam system for autonomous vehicle operation
Europäisches Patentamt | 2024
|Online Contents | 2018
|Towards Anomaly Detection in Dashcam Videos
IEEE | 2020
|Comparative Study of Dashcam-Based Vehicle Incident Detection Techniques
DOAJ | 2021
|