In order to further expand this project, consider the following assumptions and findings :
Assumptions :
- The dimensions of the object we are trying to detect are known.
- The setup is an ideal situation, although we have now expanded our results for a more arbitrary setup.
- Object recognition is sensitive to brightness and intensity --> change in color sometimes make object unrecognizable
- For stereo video settings, on timing the following process:
- Take an image, switch camera, take another image, load to OpenCV is 4.3 seconds
- Switching between cameras(left camera capture followed by right camera capture): 4ms
- Take an image switch camera, take another image, load to OpenCV is 1.05 seconds on Downsampled Image.
- Conclusion: Detecting moving object with this setup is not possible, restricted to stationary scenes, live stream needed
- Please find the link for another document where we have documented the steps(give you an indication as to ways to go about reproducing and hopefully building on our work) and progress :
- https://docs.google.com/document/d/18NVdjQQEMhK3DtA_-ApHYeXf_j6ZUWTe8pYkBAoidsc/edit#