Mini Project Report on

LANE AND CURVE DETECTION ON ROAD

Submitted in partial fulfilment of the requirement for the award

of the degree of

BACHELOR OF TECHNOLOGY
IN
COMPUTER SCIENCE & ENGINEERING

Submitted by:

Student Name: Shivam Kumar University Roll No. :

2019099

Under the Mentorship of

Mr. Priyank Pandey
ASSISTANT PROFESSOR

Department of Computer Science and

Engineering
Graphic Era (Deemed to be University)
Dehradun, Uttarakhand
July-2024
 CANDIDATE’S DECLARATION

I hereby certify that the work which is being presented in the project report entitled
“Lane and curve detection on road” in partial fulfillment of the requirements for
the award of the Degree of Bachelor of Technology in Computer Science and Engineering of
the Graphic Era (Deemed to be University), Dehradun shall be carried out by the under the
mentorship of Mr.Priyank Pandey, Assistant Professor, Department of Computer Science
and Engineering, Graphic Era (Deemed to be University), Dehradun.

Name : Shivam Kumar

University roll no. : 2019099
 Table of Contents

Chapter No. Description Page No.

Chapter 1 Introduction 1
Chapter 2 Literature Survey 2-4
Chapter 3 Methodology 5-7
Chapter 4 Result and Discussion 8
Chapter 5 Conclusion and Future Work 9
References 10
Chapter 1

Introduction

1.1 Introduction

Driving is an essential part of daily life for many people, and ensuring safety on
the road is a top priority. Lane detection plays a crucial role in making driving
safer, especially for autonomous vehicles and driver-assistance systems. By
accurately identifying lane boundaries, these systems can help keep cars on the
right path, alert drivers to potential lane changes, and navigate curves safely.
This code demonstrates a straightforward approach to lane detection using
OpenCV, a powerful library for image processing and computer vision.

1.2 Problem Statement

Imagine driving on a busy highway or a winding country road. The ability to
clearly see the lanes and understand whether you're going straight,
approaching a curve, or changing lanes is vital for safe driving. However,
detecting lanes in real-time can be tricky due to various factors like lighting
conditions, road quality, and weather.

This project aims to tackle these challenges by:

1. Detecting Lane Markings: Using edge detection techniques and the Hough
Transform, the code identifies the lane lines on the road.
2. Understanding Road Geometry: By analyzing the detected lanes, the code
determines if the road is straight, curving, or if there's a lane change.
3. Real-Time Feedback: Processing video frames in real-time, the system
provides immediate visual feedback, which is essential for practical driving
applications.

By implementing this solution, we can enhance the capabilities of autonomous

vehicles and driver-assistance systems, making roads safer for everyone. This
project is a step towards more advanced and reliable driving technologies.

1
Chapter 2

Literature Survey
The task of lane detection is fundamental in the field of autonomous driving
and driver assistance systems. The development of robust and efficient lane
detection algorithms has been an area of active research, leveraging
advancements in computer vision and machine learning. This literature survey
outlines some key contributions and methodologies in the domain of lane
detection.

2
1. Traditional Methods

(A) Edge Detection and Hough Transform:

One of the earliest and most widely used methods for lane detection involves
edge detection techniques, such as the Canny edge detector, followed by the
Hough Transform to identify straight lines representing lane boundaries.

Hough, P. V. C. (1962). "Method and Means for Recognizing Complex Patterns."

U.S. Patent 3,069,654. This seminal work introduced the Hough Transform,
which has since been a cornerstone in the field of computer vision for
detecting shapes like lines and circles.

Canny, J. (1986). "A Computational Approach to Edge Detection." IEEE

Transactions on Pattern Analysis and Machine Intelligence, 8(6), 679-698. This
paper presents the Canny edge detector, which is crucial for identifying edges
in images before applying the Hough Transform.

(B) Region of Interest (ROI):

To enhance the accuracy and efficiency of lane detection, many systems define
a region of interest in the image, focusing the analysis on the area where lanes
are most likely to appear.

Tsai, Y.-C., et al. (2006). "A Vision-Based Lane Detection System Combining
Hough Transform and Kalman Filter." Proceedings of the 2006 IEEE
International Conference on Systems, Man, and Cybernetics. This work
integrates ROI with the Hough Transform, showing improved performance in
real-world driving conditions.

2. Machine Learning Approaches

(A) Support Vector Machines (SVM):

Some approaches have employed machine learning techniques like Support

Vector Machines to classify lane markings and road boundaries.

Aly, M. (2008). "Real time detection of lane markers in urban streets."

Proceedings of the IEEE Intelligent Vehicles Symposium, 7-12. This study uses
SVM for lane marker detection in urban environments, demonstrating the
effectiveness of machine learning in complex scenarios.

3
(B) Convolutional Neural Networks (CNNs):

The advent of deep learning has significantly advanced lane detection

methods. Convolutional Neural Networks, in particular, have been successful in
capturing complex patterns and features in images.

He, K., Zhang, X., Ren, S., & Sun, J. (2016). "Deep Residual Learning for Image
Recognition." Proceedings of the IEEE Conference on Computer Vision and
Pattern Recognition (CVPR), 770-778. Although not specific to lane detection,
this paper on residual networks (ResNets) has influenced many deep learning-
based lane detection models.

Neven, D., Brabandere, B. D., Proesmans, M., & Van Gool, L. (2018). "Towards
end-to-end lane detection: an instance segmentation approach." Proceedings
of the IEEE Intelligent Vehicles Symposium (IV), 286-291. This work proposes an
end-to-end lane detection model using instance segmentation, showcasing the
power of deep learning in this domain.

(C) Hybrid Approaches:

Combining traditional image processing techniques with machine learning

models has also been explored to leverage the strengths of both approaches.

Asvadi, A., Peixoto, P., Nunes, U., and Pébereau, L. (2017). "3D Lidar-based
static and moving obstacle detection in driving environments: An approach
based on voxels and multi-region ground planes." Robotics and Autonomous
Systems, 83, 299-311. This hybrid approach uses traditional methods to
preprocess data, which is then fed into machine learning models for more
accurate detection.

4
 Chapter 3

Methodology
The methodology for this lane detection project involves several key steps,
combining traditional image processing techniques with real-time video
analysis. The overall approach includes preprocessing the input video,
detecting edges, identifying lane lines using the Hough Transform, and
analyzing these lines to understand lane geometry. Here is a detailed
breakdown of the process:

1. Input Video Capture

The first step is to capture the video input that will be processed. This involves:

 Loading the Video: Using OpenCV to load the video file.

 Frame Extraction: Reading the video frame by frame to process each
image individually.

2. Preprocessing

Preprocessing the video frames is crucial to enhance the accuracy of lane

detection. This involves:

 Grayscale Conversion: Converting the frames to grayscale to simplify the

image and reduce computational complexity.

5
  Edge Detection: Applying the Canny edge detector to highlight the edges
in the image, which are crucial for detecting lane lines.

3. Region of Interest (ROI)

To focus on the part of the image where lanes are likely to be, a region of
interest is defined:

 Mask Creation: Creating a mask to isolate the bottom half of the image
where the road is usually visible. This helps in ignoring irrelevant parts of
the image, such as the sky and surrounding environment.

4. Line Detection

Once the region of interest is isolated, the next step is to detect the lane lines:

 Hough Transform: Using the Hough Transform to detect straight lines in

the edge-detected image. This method is effective in identifying the lines
that represent lane boundaries.

5. Drawing Lines

After detecting the lines, the next step is to visualize them on the original
image:

 Line Drawing: Drawing the detected lines onto a blank image and then
overlaying this on the original frame. This helps in clearly visualizing the
detected lanes.

6. Analyzing Lane Geometry

To understand the driving path, the slopes of the detected lines are analyzed:

 Slope Calculation: Calculating the slope of each detected line to

determine its orientation.
 Lane Determination: Classifying the detected lines into left and right
lanes based on their slopes. Lines with negative slopes are typically part
of the left lane, and those with positive slopes are part of the right lane.

6
7. Lane Change and Curve Detection

Based on the analyzed slopes, the system determines the current driving
condition:

 Straight Path: If both left and right lanes are detected with appropriate
slopes, it indicates a straight path.
 Curve Detection: If only left or right lanes are detected, it suggests the
vehicle is on a curve (left or right).
 Lane Change: If the slopes of detected lines do not fit typical left or right
lane patterns, it may indicate a lane change.

8. Real-Time Feedback

The final processed frames are displayed in real-time:

 Frame Display: Showing the processed video frames with the detected
lanes and annotated driving conditions, such as straight, left curve, right
curve, or lane change. This real-time feedback is crucial for practical
driving applications.

7
Chapter 4
Result and Discussion

Result:
The lane detection algorithm effectively processes video frames to identify and
superimpose lane markings using fundamental computer vision techniques. It
begins by converting frames to grayscale, applying Canny edge detection, and
isolating regions of interest. Through the Hough transform, it identifies straight
lines and curves, enhancing them with visual overlays on the original video
feed. By analyzing slopes derived from detected lines, the algorithm categorizes
lanes as straight, left curve, right curve, or undergoing a lane change, providing
immediate feedback on road conditions.

Discussion:
While capable in controlled environments, this approach may struggle with
challenges like varying lighting conditions, road surface textures, and sharp
curves where lane markings are ambiguous or absent. Improvements could
encompass adaptive parameter tuning, employing advanced curve fitting
algorithms for smoother lane delineation, and integrating machine learning
models for enhanced robustness across diverse road conditions. Furthermore,
optimizing for real-time performance on different hardware configurations
would expand its applicability in autonomous driving and assistive navigation
systems, ensuring reliable and accurate lane detection in real-world scenarios.

8
Chapter 5

Conclusion and Future Work

Conclusion:

The implemented lane detection algorithm demonstrates effective real-time

processing of video frames to detect and highlight lane markings using basic
computer vision techniques. It successfully applies edge detection, region
masking, and Hough line transformation to identify straight lines and curves,
categorizing lanes based on slope analysis. This approach provides immediate
visual feedback on lane status, supporting applications in driver assistance
systems.

Future Work:

To enhance its utility, future work could focus on improving robustness in

challenging conditions such as varying lighting and road textures.
Implementing adaptive algorithms for parameter tuning and integrating
advanced curve fitting techniques would facilitate more accurate and reliable
lane tracking, especially around sharp curves or ambiguous markings.
Additionally, exploring machine learning approaches for lane detection could
further enhance performance across diverse environments. Optimizing
computational efficiency for deployment on embedded systems and exploring
real-time performance enhancements on different hardware platforms would
also be crucial for broader application in autonomous driving and navigation
systems.

9
References
[1] N. K. Kanhere and S. T. Birchfied, “Real-time incremental segmentation
and tracking of vehicles at low camera angles using stable features,” IEEE
Trans. Intell. Transp. Syst., vol. 9, no. 1, pp.148160, March 2008
(Example : Journal papers)
[2] K. Onoguchi, “Moving object detection using a cross correlation between a
short accumulated histogram and a long accumulated histogram”, Proc.
18th Int. Conf. on Pattern Recognition, Hong Kong, August 20 - 24, 2006,
vol. 4, pp. 896 – 899 (Example : Conference papers)
[3] T. H. Cormen, C. E. Leiserson, R. L. Rivest and C. Stein, “Introduction to
Algorithms”, 2nd ed., The MIT Press, McGraw-Hill Book Company, 2001
(Example : Text Book/ Magazine)
[4] Open Source Computer Vision (OpanCV) [Online]. Accessed on 21st April
2022:
http://opencv.willowgarage.com/wiki/ (Example : Website)

10