Project Milestone Report

Car detection on mobile phones

Tianyu Wang

Milestone Achievements

1. Merge different car front view image data set and preprocess the image according to the phone camera quality.
2. Train a car front view detection model by Adaboostalgorithm using OpenCVlibrary.
3. Evaluate the car detector performance offline.

Preprocess training images

1. Training datasets:
MIT CBCL CAR DATABASE #1: 516 128x128 PPM format images


Caltech Cars 2001 (Rear): 526 images of rear view cars


Negative images: Google Street View images with no car 250x150 large images


Preprocess:
1. Crop the car regions and form 40x40 grayscale images. Remain as little background as possible in the positive samples to get better results.
2. Equalize color histogram in the whole image.

Experiments

True Positive on training sets:


True Positive on test sets: INRIA Annotations for Graz-02 (IG02)


False Positive on test sets: INRIA Annotations for Graz-02 (IG02)


What's next

1. Evaluate the detection performance thoroughly
2. Modify the training dataset and retrain the car detector
3. Modify the training parameters for Adaboostalgorithm
4. Reduce false positive results
5. Collect car/noncarimages by Nexus One and evaluate the car detector on this test dataset 6. Port the car detection model to Nexus One and evaluate the accuracy and realtime performance

References

  1. Jack Nasar et al, "Mobile telephones, distracted attention, and pedestrian safety", Accident Analysis and Prevention 40 (2008) 69-75
  2. MIT cbcl Car Database, http://cbcl.mit.edu/software-datasets/CarData.html
  3. The PASCAL object recognition database collection. http://www.pascal-network.org/challenges/VOC/
  4. Lee, D. “Boosted Classifier for Car Detection.” Carnegie Mellon University.
  5. Caltech Rear View Car 2001 dataset http://www.vision.caltech.edu/html-files/archive.html
  6. Tutorial: OpenCV haartraining (Rapid Object Detection With A Cascade of Boosted Classifiers Based on Haar-like Features) http://note.sonots.com/SciSoftware/haartraining.html