Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{198003,
        author = {D.Balaji and Ms. P. Sudha and E. Bhargav and E.Jaya Chandra},
        title = {Vision-Based Adaptive Driver Assistance System With Predictive Lane Guidance Using YOLOv8 and Optical Flow},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {8475-8483},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=198003},
        abstract = {Existing ADAS solutions rely on expensive LiDAR and radar sensors costing tens of thousands of dollars, placing them beyond the reach of most vehicle owners in cost-sensitive markets. This paper addresses that gap by proposing a six-function Vision-Based Adaptive Driver Assistance System that operates entirely on a single consumer-grade monocular camera at a hardware cost of approximately INR 1,500, without any GPU. The proposed system delivers: (1) simultaneous dual-lane boundary detection with transparent overlay, (2) hysteresis-stabilised steering direction prediction with angular precision, (3) real-time multi-class vehicle detection using YOLOv8 nano with ByteTrack identity tracking, (4) metric distance estimation using the pinhole camera model, (5) three-tier collision alerting — DANGER below 8 m, CAUTION at 8–10 m, SAFE at 10 m and above — and (6) dual-method absolute speed estimation combining optical flow and distance-change velocity. A novel three-source edge detection pipeline merging Canny gradients, yellow HSV colour masks, and white brightness masks enables simultaneous detection of solid yellow and dashed white lane markings — a capability absent in all single-source classical approaches. A position-based Hough segment assignment algorithm with four-stage geometric validation completely eliminates the lane-crossing artefact observed in approximately 15% of highway frames under classical slope-sign methods. Experimental evaluation on 2,100 frames of dashcam footage recorded on Indian roads demonstrates 87% dual-lane detection reliability (+18 percentage points over the Canny baseline), 91% vehicle detection accuracy at ranges up to 50 metres, 1.8 m mean absolute distance error, and 98% collision warning precision, all at 15–25 frames per second on standard laptop CPU hardware.},
        keywords = {ADAS, Lane Detection, YOLOv8, Optical Flow, Collision Warning, Hough Transform, Pinhole Camera Model, ByteTrack, Embedded Systems, Real-Time Vision.},
        month = {April},
        }