how to solve occlusion problem in multiple object?

asked 2014-10-09 07:29:39 -0500

coco gravatar image

updated 2014-10-19 02:31:45 -0500

I am trying to run code of occlusion for multiple object tracking and following I don't know much about occlusion please suggest some code or solution regarding multiple object tracking

edit retag flag offensive close merge delete



Kalman Filter?

thdrksdfthmn gravatar imagethdrksdfthmn ( 2014-10-09 08:28:55 -0500 )edit

thanx Abhishek Kumar Annamraju. I got knowledge from this . Can you tell me how to use kalman filter in occlusion i found some reference code but don't know how to use it. how to relate kalman filter to occlusion.

coco gravatar imagecoco ( 2014-10-16 01:37:43 -0500 )edit

@thdrksdfthmn Yeah Kalman filter. Which is actually use for occlusion.

But don't know how to use it.

and how to include on my own code

coco gravatar imagecoco ( 2014-10-16 01:43:59 -0500 )edit

I think you should start with this for better understanding, then this for having an idea of code. The idea is to always enforce the prediction with a detection (and search for the object in a smaller area, around the prediction); if there is no detection for a few frames than you should loose the tracking (but this is when testing and tuning the filter)

thdrksdfthmn gravatar imagethdrksdfthmn ( 2014-10-16 02:15:28 -0500 )edit

@thdrksdfthmn thanx I will try and Let you know if I will Find Any Problem. Meanwhile Can you tell me how to train features of any object. And practical example for it.

Than in Advance. :)

coco gravatar imagecoco ( 2014-10-16 06:57:51 -0500 )edit

I do not know what "to train features" means, but I know that features can be found using FeatureDetector. You shall also see the DescriptorExtractor that may be used for training.

thdrksdfthmn gravatar imagethdrksdfthmn ( 2014-10-16 08:26:31 -0500 )edit

@thdrksdfthmn Thanx for the reply. Train feature means...train own algorithm like left eye , right eye and all other face feature have own xml file. so i want specific shape base feature for that I have to train them and make xml file so that I can include directly in to my code . So do you know exact procedure to train feature.

coco gravatar imagecoco ( 2014-10-17 00:09:14 -0500 )edit

yes, it is traincascade in opencv. These are the docs, here is a related question, so good luck!

thdrksdfthmn gravatar imagethdrksdfthmn ( 2014-10-17 02:18:54 -0500 )edit

@coco and @thdrksdfthmn Make sure you fetch all the other possibilities you have other than traincascade, try going up for advanced feature extractors like Weber Descriptors,SIFT, etc. Traincascade is a very tedious process and unless combined with parallel computing it is a very long job, even the results are uncertain as you need to play with the training parameters a lot. Look for SVM training also if you have to go up for machine learning. Anyway here's a tutorial from my blog for traincascade :

Abhishek Kumar Annamraju gravatar imageAbhishek Kumar Annamraju ( 2014-10-17 02:39:39 -0500 )edit

I know, that is what I have done a lot of time for a better face detector, and it was exhausting and not a very big improvement (except the speed; I have trained a LBP). @coco You can try to search for another type of detector, like adaptive threshold or segmentation, morphology, colours, saturation, or I do not know what. And then a classifier for validating the detection.

thdrksdfthmn gravatar imagethdrksdfthmn ( 2014-10-17 03:32:57 -0500 )edit

@AbhishekKumarAnnamraju: Have you trained 2 detectors, or it is the same for lateral and behind?

thdrksdfthmn gravatar imagethdrksdfthmn ( 2014-10-17 03:46:23 -0500 )edit

@thdrksdfthmn I have trained many xml files for the same object/dataset using variations in parameters and then tried to use the best ones to detect the cars in most efficient way.

Abhishek Kumar Annamraju gravatar imageAbhishek Kumar Annamraju ( 2014-10-17 06:44:40 -0500 )edit

@Abhishek Kumar Annamraju@thdrksdfthmn thank you for reply. I am trying all methods. It will take some time. Letter i will inform you about it. Thanks again

coco gravatar imagecoco ( 2014-10-19 01:53:22 -0500 )edit