The TRR1 emphasizes the communication of the microcontroller, camera and database. Thus, based on feedback received by prof. Dovis, an approach to actually measure how the system is working can go this way:

  1. Long-run simulation: the system should be able to work for a whole ‘journey’ in the context of a university let’s consider between 07.00 and 20.00. Let’s also consider taking a picture every 5 minutes, which translates into 156 pictures per day. A priori: no 2 consectuives fails. How many of these pictures are successfully sent from the Raspberry Pi to the database service?
  2. MSE: Mean-square error may help us to know the distortion between the 2 pictures, the one taken directly from the camera and the one which was encrypted and stored in the database. Use OpenCV. Maybe not need it if not compression.

Also, as suggested by prof. Dovis: perform tests on aspects like how much does daylight illumination affects the system? and what if people pass by the rack while taking the pictures?

Check exposition parameter in the camera. Automatic.

However, these kinds of tests fit into the second branch of the architecture, the processing of the images which we are going to elaborate on during TRR2.

Test bike detections with both alternatives for 5 days:

  1. Deal with the daylight and time: monitoring the working state of the system during different hours of the day, TBD yet, however, early morning, noon and evening probably should be emphasized. We need to know if our bike recognition algorithm can deal with these issues.

  2. People and objects: Same as above, but with physical tangible constraints, such as people leaving their bikes or picking them up, persons passing by, etc.

    Algorithm detect people, so have exception loops in the code.