Vehicle Detection

• Vehicle Detection + Machine Learning
• OpenCV + Python + scikits-learn + scikit-image

---

    

---

Lane Detection | Full Video | Full Report

Overview

Machine Learning Classifier

• I trained a linear SVM using LinearSVC from sklearn

    from sklearn.svm import LinearSVC

• To go through all the training images, my classifier takes 0.6 seconds to achieve 97.6% accuracy.
• Here is a sample batch of my training images provided by GTI Vehicle Image Database
• From this awesome vehicle image database, there are two types of data: car images & non car images

Computer Vision Pipeline

+ Define Window Grid to extract images

+ Extract Image HOG (Histogram of Oriented Gradients) Features

+ Feed through Pre-train Classifier to detect vehicles

+ Apply heatmap to identify all vehicle detections

+ Apply threshold to remove false positives (3 overlapps = vehicle)

+ Draw boxes on Original Image

+ Hooray