I divided the project into two phases -- first one to identify automatically the objects inside the closed container and second to establish the ethylene secretion levels inside the container.
For the first step, i.e. the recognition of the fruits/vegetables, I did image segmentation to optimize the detection mechanism. I deployed it using raspberry pi and opencv with contours detection. I initially took a default image to check the results :-
source : Google images
I ran the image segmentation techniques for such a condition with the following resulting contours :-
You can find the code on the github link that I have shared.
The image segmentation is dependent on a lot of things with light/brightness of the original image of the major issue.
I will be modifying the algorithm to take into consideration low light conditions inside a closed chamber with the raspberry pi mounted inside the top.
Also, my next steps will include machine learning to identify the fruits and isolating them individually.