Program flow

On figure A.2 on page we can see in more detail how the car recognition process is implemented. It consists of three basic modules:

• feature extraction module
• error function module
• classification module

We will focus on these modules later.

In the beginning the input test car image is loaded. Features are extracted by Feature Extraction Module. If the car image does not have cached feature data the cache is being created. Later this cached is preferred since it is many times faster than to compute the features again. The next step is to compare error in Error Function Module between every train car and test car. Once error distances are computed the classification performed by Classification module can begin. It can consists from selecting the closest car or by dimension reduction, etc. The classification can also utilize the steps above. The last step is to display the closest image from the train database.

Figure A.2: Program Flow