Image processing: Find similar object in multiple images and determine its height

Hello everybody,

I have the following problem:

I have several thousands of images of a camera which is located outdoor and has a fixed position. The camera takes photos during the winter months and each image contains a fixed gauge (levelling staff) which can be used to measure the snow depth. I simply could read the snow depth manually from each image, but due to the number of pictures I wanted to create an automatized approach which is able to read and store the snow height of each picture. Maybe someone can give me a small hint what might be the best approach? The images are full of disruptive objects such as trees. The color of the gauges are red and black striped (each stripe has a length of 10 cm which allows for the determination of the absolute snow depth). My approaches so far:

• bwlabelling of the corresponding binary image and the boundingbox determination via regionprops (it detects thousands of objects but not the gauge)
• detectSURFFeatures / extractFeatures / matchFeatures after cropped the gauge (but the gauge feature always has 0 points)
• edge detection (also did not lead to the expected results)

Hopefully someone can provide me some guidance or help :)?

An example file is attached!

Greetings

[Matt J]

I suggest you attach a representative image or two for us to play with.

[Timo]

Hey Matt, you are totally right :)! Thank you!

[Timo]

Especially the gauge in the foreground is important, the second one in the background could be neglected :). Thanks in advance!