Vehicle Detection and Tracking

An image processing pipeline to detect and track vehicles

INTRODUCTION

The goal of this project is to write a software pipeline that detects vehicles in an image and tracks them across frames of a video captured from a front-facing camera on a car

A demonstration of the pipeline is show in this video

This project was undertaken as part of the Udacity Self-Driving Car NanoDegree.

Pipeline summary

• Extract features from a labelled training set of images and train a Linear SVM classifier
  • Features: histogram of oriented gradients, colour histograms and spatial binning
• Implement a sliding-window across an image and use a trained classifier to search for vehicles
• Create a heat map of recurring detections frame by frame to reject outliers and track detected vehicles
• Estimate a bounding box for each tracked vehicle

HOW TO USE

Project dependencies

You can follow the guide explained here to setup a working environment.

• https://github.com/udacity/CarND-Term1-Starter-Kit

Jupyter Notebooks

The included notebooks demonstrate how to use the project code.

• vehicle_detection.ipynb
  • Demonstrates the end-to-end process, from the dataset, to training the classifier and processing images and video
• notebooks/pipeline_breakdown.ipynb
  • Breaks down the image processing pipeline into stages to visualise the process
• notebooks/helper.ipynb
  • A collection of other useful snippets of code used along the way

RELEVANT LINKS

Project writeup

• writeup.md

Original project repo

• https://github.com/udacity/CarND-Vehicle-Detection

RELEVANT FILES

• vehicle_tracking.py, including:
  • A Vehicle class to represent tracked vehicle objects
  • A VehicleTracking class that implements the full processing pipeline
• helper.py, including independent implementations of various stages of the pipeline and other miscellaneous helper functions