morganpackard's profile - activity

Hello CV'ers. I'm tearing my hair out with an issue and hoping for some help.

I'm doing some simple background subtraction and am getting garbage data in my processed images. The garbage data is present or absent depending on the dimensions of my original image. Some image dimensions work fine, others are very glitchy. Generally, it seems like the issue appears as one of the sides of the image approaches 100px.

I'm adding two screen shots. Both contain a sequence of four images. The images are created with the following processes (left to right):

1. Raw image
2. Running Average of raw input images (cv.RunningAvg in python)
3. difference of raw image and running average
4. binarized difference image

This is image is big (300px wide or so)

This image is small (100px or so wide)

I'm seeing the issue on my mac, and on a raspberry pi.

So, the short version of the question is: why am I getting crazy glitches with certain image dimensions, but not others. And how can I use small images without getting these glitches?

Based on a previous question asked here, it looks to me like OpenCV discards the alpha channel when reading frames from video using VideoCapture. Is this correct? Is VideoCapture not capable of reading the alpha channel of a video? I just want to make sure so I'm not tempted to keep trying. I've tried messing around with the "channel" argument in VideoCapture::retrieve, without any success so far.

I'd also really like to be able to be able to read frames from a video file without losing the alpha channel.

Confirmed. The image is findable. Any other ideas?

I've seen several references to an issue in which imread fails silently when the wrong libs are used. It's working properly for me in release mode, but not in debug mode. I've triple-checked that my configuration is the same across the two with the exception of adding a "d" at the end of every lib's name, for "debug". So instead of including opencv_highgui248e.lib, I include opencv_highgui248d.lib. I'm using visual studio 2012, building for x64. My library path is C:\sdf3\opencv_2_4_8\build\x64\vc11\lib (though I've also tried vc12).

Any ideas what else I can try?

@berak -- right, so I suppose if I wanted to use HOG I'd need to do some of my own training. More reason to avoid HOG if I can get away with something less sophisticated.

I need to read a video stream from a camera mounted about 10 feet in the air, and pointed at a steep downward toward a sidewalk. I simply need to detect if ANY person-sized object has moved into the camera's view, and remained there for a few seconds.

I know this is fairly trivial if I have a controlled environment and a relatively static background. I can remove the background, then run a contour detector. However, if it's an outdoor camera, I'll need to be constantly grabbing new background shots, waiting until the sidewalk is empty before grabbing, which seems far from foolproof.

I've experimented with HOG:

HOGDescriptor hog;
hog.setSVMDetector(HOGDescriptor::getDefaultPeopleDetector());

This seems to give some false positives, and may be more complex than I need (I don't care if an object is a person or not, I just care if it's present and big).

What approach would you all take for simply detecting whether there's a big thing on a sidewalk or not, across a wide range of weather and light conditions?

**** EDIT *******

This looks like it may be most of what I want.

http://docs.opencv.org/trunk/doc/tutorials/video/background_subtraction/background_subtraction.html