Research Teem Meeting
Here we discussed our previous research thus far into our implementation method. Various technologies were still flouting about in our heads, from techniques such as trying to implement neural networks to extract key features from the image of a vehicle, and then more sensibly, using Support Vector Machines (SVM) algorithms. We were still quite lost as to how we should tackle our task, so we sought help/guidance from Jane.
Jane shared our confusion as to how we should tackle this problem, and did mention that the project idea was stripped from a final year project student who implemented the project using machine learning and neural networks to categorise each vehicle, but thought it would be interesting to see how this could be implemented using pure image processing.
We bounced our ideas off Jane, and she agreed we should look at trying to extract key features from vehicles. Furthermore, we should look as trying to measure the size of the vehicle in the image, and use the wheels as a bases as to how we'll get a measurement. This measurement will either be:
- a pixel count
- metric length representation
- a ratio of lengths.
Using the chart above as a starting point, we can see that if we can isolate the size of the wheel in an image, we can make an approximation of the size of the vehicle as either one of the measurement mentioned above.
Team minutes can be found here hosted on Google Drive (DIT signin required) Team Minutes