This week is the last week of the year 2 project. In the last week, we basically finished the requirements we initial planned. And our goal is to improving our current product.
In the last week, we have more focused on the software part of our project. In specific, we worked out the imagine fusion function. And it improves the original thermal image's resolution.
To achieve this, we have been searching for a possible algorithm for a long time. However, most of the solution is not suitable for us. Therefore, we tried to combine them together and provide the image fusion function. But in practice, we realised the our thermal camera's resolution is not higher enough which means it couldn't provide the enough temperature information for further fusion. In other word, in RGB color model, our current algorithm is not good enough to provide a clear 'real-time' fusion image. But after the discussion and a large amoung of experiments, we finally found it works in grayscale (Figure[1], Figure[2] and Figure[3] shows the result in greyscale models, thermal image and pre-processed visible image respectively) color model and it provides some ideas for us. The temperature information provided by the grayscale image is not very obvious. But it inspired us to find a Image Fusion in RGB model with a new direction.
Figure [1] : Final result - Fused image
Figure [2] : Thermal Image - PreProcessed
Figure [3] : Visible Image - PreProcessed
After the tireless efforts of our team, we finally found a suitable method for RGB model image fusion. In this model, we have applied edge detection into the fusion function to provide a more accurate fused image. It provides us a new image with a good thermal and visible image fusion function. It can clearly matching the thermal image's info to visible image with a good edge detection and matching.
So far, our project can be said to be initially completed, but in the existing software version, we still feel that the final fusion effect is not obvious, so we decided to further improve it to improve the intuitiveness of the final fusion image. Therefore, we have developed another version of our software which give us more color information on the fused image. And the screenshot is the final result of our project with image fusion and edge detection and alignment.
Summary of week’s activities:
Finished the image fusion function.
Plant the software in Raspberry Pi.
Poster preparation.
Problem found:
1. The resolution improments in RGB color model is not enough.
2. The hot/cool point detection function is not working.
Tasks for next week:
1. Finish the poster.
2. Prepare for the presentation.
3. Try to improve the resolution in RGB color model.
4. Fix the hot/cool point detection.
Progress check:
✔ Experiment Material Preparation
✔ Raspberry Pi Built
✔ Combine thermal and visible picture together
✔ Improve the picture resolution
✔ Poster making
Commenti