Classification of vehicle manufacturer
Project Summary
Overview
The task is to find suitable and accurate way to determine the truck manufacturer from given image
data. Part of the solution should be a comparison of different approaches of image classification
(SVM, RandomForest etc.) and evaluation of OCR tools usability for this problem. Data are
collected from two camera sources from different places and the data were selected from various
parts of the day to provide picture variability.
Data
Given image data of trucks displays front part of the vehicle (the image has been already preprocessed,
including equalization of histogram). Given data sets have three main classes - DAF, MAN
and Mercedes. The numbers of collected samples are shown in following table.
- Number of samples
DAF 202
MAN 267
Mercedes 242
According to the fact that data are collected from real devices, there are many difficulties with
processing this type of data. The data sets contain various weather conditions, vehicle damages
and of course some outliers samples as you can see on following images.
(a) Outlier (b) Vehicle damages (e.g. missing letters)
(c) Bad weather conditions
Figure 1: Difficulties with processing data
B–1
Goal
The goal of the project is to compare various methods of machine learning for vehicle manufacturer
recognition. For feature extraction, you can use given application for histogram of oriented gradient
extraction [1,2]. Application is able to generate ARFF file with extracted features which were
selected by parameters or you can choose other feature extraction of your choice. Part of solution
is to compare usability of OCR tools like Tesseract.
Literature
[1] Dalal, Navneet and Triggs, Bill, ”Histograms of oriented gradients for human detection,” 2005
IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR’05), San
Diego, CA, USA, 2005, pp. 886-893 vol. 1. doi: 10.1109/CVPR.2005.177
[2] Shah, Mubarak. Histograms of Oriented Gradients. [online]. UCF Computer Vision Video Lectures
2012 [cit. 14.11.2016]. Available from https://www.youtube.com/watch?v=0Zib1YEE4LU
B–2