Fusion model based on entropy by using optimized DCNN and iterative seed for multilane detection
- PDF / 3,795,558 Bytes
- 14 Pages / 595.276 x 790.866 pts Page_size
- 75 Downloads / 178 Views
SPECIAL ISSUE
Fusion model based on entropy by using optimized DCNN and iterative seed for multilane detection Suvarna Shirke1 · R. Udayakumar2 Received: 16 February 2020 / Revised: 15 July 2020 / Accepted: 25 August 2020 © Springer-Verlag GmbH Germany, part of Springer Nature 2020
Abstract The recent progress in the DSAs led to the progress of advanced lane detection systems for preventing accidents. The information about the lane is detected, which is used for the vehicle control and provide warning to the drivers. This paper proposes the multi-lanes detection based on an entropy-based fusion approach. The main use of the proposed model for combining the results obtained by the EW-CSA based DCNN and iterative seed based on region for providing efficient multilane detection. Initially, the lanes are detected using a deep learning technique that is trained using an optimization algorithm, EW-CSA. Similarly, the approach namely segmentation based on region is used for the detection of multi-lane. Depends on the results of the two approaches, an fusion model based on entropy is proposed for making the best results, based on a pre-defined threshold. The proposed method performance is evaluated based on the metrics, such as specificity, accuracy, and sensitivity, which outperforms with values 0.887, 0.991, and 0.992, respectively. Keywords Driver assistance systems · Lane detection · Entropy · Deep convolutional neural network · Region based segmentation · Earth worm-crow search algorithm · Fusion model Abbreviations EW-CSA Earth worm-crow search algorithm LDW Lane departure warning ITS Intelligent transportation system DSA Driver assistance system DCNN Deep convolutional neural network ADAS Advanced driver assistance system LIDAR Light detection and ranging GPS Global positioning system EWA Earthworm optimization algorithm CSA Crow search algorithm MBMT Min-between-max thresholding ROI Regions of interest IPM Inverse perspective mapping * Suvarna Shirke [email protected] R. Udayakumar [email protected] 1
Department of Computer Science and Engineering, Bharath Institute of Higher Education and Research, Selaiyur, Chennai 600 073, Tamil Nadu, India
Department of Information Technology, Bharath Institute of Higher Education and Research, Selaiyur, Chennai 600 073, Tamil Nadu, India
2
RANSAC RANdom SAmple Consensus CNN Convolutional neural network DARPA Defense advanced research projects agency FPR False positive rate DVPE Dense vanishing point estimation MMBMT Modified min-between-max thresholding TPR True positive rate
1 Introduction The advancements in LDW system devoted the attention of several researchers over the lane detection for yielding ITS. Several automobile manufacturers devised advanced driver-assistance systems are used for the prevention of unintended lane departure [1]. Thus, researches have been considered for the development of the on-board systems with more intelligent, whose intend is to prevent the accidents for mitigating severity. The ADAS adapts more in
Data Loading...
