Proposed method

In Chapter 2 we saw that we can expect a classification rate between 80-97%. We will try to classify vehicles according to car make like Skoda Fabia, Skoda Felicie, Daf, etc.

The recognition process can be described as follows (figure 3.1 on page ):

• Extraction and Geometrical Normalization: Vehicle samples are extracted from images in order to provide learning classifiers for vehicle car makes. The Region of Interest (RoI) is extracted from every image sample and it is normalized to MxN resolution.
• Feature Extraction: Every RoI is processed by feature extraction algorithm. Feature vectors are projected into feature space.
• Learning Classifiers: We used k-NN, and FLD as classifiers and two different metrics for k-NN

Figure 3.1: Proposed Classification Process