20 July 2018

My true geekiness is shining through

After months of hiatus from science, I finally got to play with some data. Although this data is from biology, I still can apply my skills. And that really brought me pure joy.
As I said in the last post, the main goal of the project is detecting which images have animals in them. This week I did not have much time to full around with the data. I had only one afternoon.
Most of the time I spent exploring the data. Because, at first, the problem sounds simple, all one has to do is detect the movement in the images.
Well, that’s what is the camera supposed to do anyway. It supposed to trigger only when there is a movement in front.
But this particular camera kept on triggering regardless and produced several tens of thousand images.
Initially, I applied the simple method, calculating the entropy of the images and trying to establish the threshold.
It sort of worked. My rough algorithm found images of some animals my frustrated husband missed. See when a human is inspecting so many images showing the exact same scene, the procedure turns very quickly into torture.
Here are a few images of the animals, few coyotes and few targets of my hubby’s research, deer. Coyotes seem cool and mostly ignoring cameras.

Screen Shot 2018-07-15 at 18.48.54Screen Shot 2018-07-15 at 18.49.44

And nice photo of a deer during daytime

Screen Shot 2018-07-15 at 18.47.41

This one deer actually detected the camera, it emits infrared flash. My hubby thinks that deer thought the flash is a glint of the moonlight reflecting in lion’s eyes. therefore such a panicky jump.

Screen Shot 2018-07-15 at 17.51.48

Here is one orange UFO which is most likely monarch butterfly.

Screen Shot 2018-07-15 at 18.50.02

And of course some homo sapiens sneaked in as well:

Screen Shot 2018-07-15 at 17.52.06
But my algorithm detected countless images of wind moving plants, clouds passing overhead, and the sun setting. All of it produces the same entropy as the movement of the animal.

Screen Shot 2018-07-15 at 18.00.56

So the next step will be to develop the method that will make a difference between false positives, i.e. wind and change in lighting levels that are declared as a movement by the simple algorithm I developed.
Another interesting thought is that my hubby thinks that carefully developed software capable to identify these animals in photos, might find a market. So, just in case, I will not share too many details about how I’m solving problems in the images, nor I will share the code. 


No comments:

Post a Comment