Hello Jakub Horák.

So, you want to track beetle(s) on a video sequence, right?

In the computer vision field, object(s) tracking can be applied to solve your problem.

Depending on the number of target objects (whether single or multiple targets), it can be divided to 1. object tracking and 2. multiple object tracking. And both of them can be implemented by using opencv and python environments.

In a broad sense, object tracking can be explained as localizing bounding boxes (bboxes) of target and measuring associations of bboxes over time (known as tracking-by-detection). Another general method is tracking by searching. The former is a typical approach for multiple objects tracking and the latter is for single object tracking. You can easily search more details from Google.

1 and 2 are similar in a way, but also many things are different. Therefore, picking a method which you will apply to your problem from top-category (e.g., multiple or single) to bottom-category is first you do.

BW.

Hi,

Is openCV/Python OK for this task?

Yes they are. If the camera is placed in a controlled environment (lighting does not change, the pot or whatever it is, is fixed, etc.) then frame differencing could be a simple yet fast option. Else, machine learning could help in tracking the beetle, but you need a labeled training data set. The YOLO — You only look once- real time object detection is written in Python and is open source, it could be of help. But I see that your vision problem is not that complicated.

Hope that helps!

Regards

Thank you guys!!

I have video sequence (400+ hours) with just 1 moving object in one arena.

Environment changed sometimes in the sense of shade intensity (day/night), but object is always visible and different in environment.

Thanks a lot!!!

have a nice day

JH

Our pleasure to have discussed this. Would be nice to hear what you ended up using.

Hi Jakub Horák,

If the bucket never moves and the camera stays in the same position you could consider a very simple approach:

1) Use masks to ignore the outside bucket contents. You should only take into account the area inside the bucket you are interested in.

2) Use a median filter to remove sand texture. (OpenCV brings you this functionality, just take a look at its tutorials)

3) Use a segmentation technique as simple as a threshold and then check for contours with the right properties. Those properties would be as simple as area, aspect ratio or the parameters of the fitting ellipsoid. (again OpenCV brings you the functionality needed to accomplish , just take a look at the OpenCV official tutorials)

4) Create a classifier to classify detect the beetles.

Thank you a lot!

As Abbas Cheddad said, It would be nice to hear what solution you end up using and what results do you get. It might be quite educational for other people trying to learn computer vision.

@ Sang-Il Oh : In the computer vision field, object(s) tracking can be applied to solve your problem.

Yes, definitely. By triangulating (or tessellating) the frames in a video and applying shape detection, analysis and classification techniques, it is possible to organized video frames into clusters containing objects (e.g., ground beetles) with similar behavior. For more about this, see

Foundations of Computer Vision

• March 2017
• Intelligent Systems Reference Library 124
• DOI:
• 10.1007/978-3-319-52483-2
• Edition: 1st
• Publisher: Springer International Publishing Switzerland
• Editor: Janusz Kacprzyk and Lakhmi C. Jain
• ISBN: ISBN 978-3-319-30260-7, ISBN 978-3-319-30262-1 (eBook)
• Projects:
• Computer Vision and Image Analysis, Understanding and Processing
• Shapes in Visual Scenes, Digital images and Videos

  Book Foundations of Computer Vision

Hi, did you solve your problem?

If Matlab works for you, you can send me an example of your video and I'll help you to create a basic solution quite fast.

General my recommendations will be to start from "optical flow" algorithm and post-processing filters: threshold, blob analysis.

optical flow link: https://uk.mathworks.com/discovery/optical-flow.html

Michael Ashkerov: No I still did not solve my problem. It would be great!! I Will send you personal message. Thank you a lot!!