This week is the second week of the year 2 project. We started working on the algorithm to improve the resolution of the picture and combine two kind of pictures together.
In this week, we have done a lot of researching releated to our projects. First, the team working on the hardware found the way to connect the thermal camera to Raspberry Pi. And it's now ready to take pictures/info from both cameras. In the next stage, they needed to find a way to fix the two cameras in place for futher image alignment.
In addition, our software team has a big progress too. First, we have found the way to improve the resolution of the picture, and it works perfectly for existing picture. This cost a large amount of time to write and improve the code based on the SRCNN(deep learning). In addition, we spent 2days to treaning the data to make it has a better performance. Third, we found the way for picture alignment using Bayesian methods. But insufficient measurements of local motion estimation make these algorithms error-prone. Therefor, we will try to find a new way to overcome it in next week hopefully.
However, the problems comes out too. We found the thermal camera could't output the image we are expecting. It only output the thermal infomation which means we have to work out a solution to reconstruct those information together and form a 32 * 32 image. This requires more time than we planned before. Beyond that, we could't do any image fusion releated tasks at the moment coursed by the incorrect thermal image output, which means we have to spend much more time on image fusion algorithm in the following week.
Overall, this week's task completion is still within the expected range. But we have to speed up and spend more time to work out a feasible program sample in the following week.
Problem found:
1. The thermal camera couldn't output the correct picture we needed.
2. Picture alignment not working in some certain circumstance.
Tasks for next week:
1. Reconstruct the thermal infomation from thermal camera and form a 32*32 image. 2. Find out the way to combine thermal and visible image together. 3. Working out a feasible program sample.
Progress check:
✔ Experiment Material Preparation
✔ Raspberry Pi Built
⃞ Combine thermal and visible picture together
✔ Improve the picture resolution
⃞ Poster making
The top picture is the result we wanted. Our aim is to using the second and third image to obtain the first image.
Comments