prakharmohan's profile - activity

@Ana M. Were you able to solve this?

How to do that same thing using Python and OpenCV?

@Andries it might be that OpenCV is not compatible, you might want to ping the contributors, ask on the irc channel and visit the mailing list. Tell me if it helps.

I can retag it but this is a suggestion, please retag this question with the proper tags. It'll help you, you know? And PLEASE go through the FAQ before randomly posting the questions.

Accept my answer if it helped you. You can do that by clicking the tick mark below my answer.

Yes try that. If not tell me, I'll try looking at it over the weekend.

The tick near the answer?

I don't know the original process but the links given below might give you some idea.

http://docs.opencv.org/doc/user_guide/ug_traincascade.html

http://coding-robin.de/2013/07/22/train-your-own-opencv-haar-classifier.html

The link provided here might help you. Note that you can also edit the threshold according tot he shade of green and the "purity" according to your use.

Accept the answer if it helps :)

Okay. This is completely new. I have never attempted this. So you can follow the guide here and see is that works out for you.

If it is too complicated a crude algorithm will be to:

1. Get images from the web cam instead of the video.
2. Apply face detection algorithm to the image.
3. Stream the video.

For face detection on images follow this guide.

For streaming you can use the SO question here.

Hope this helps. :)

Anything else, feel free to comment!

What do you want to do with the video? I mean, what processing?

Yup, that is ffmpeg problem. Are you using Linux for development? Because in that case you can specify at the time of compilation that you want to use ffmpeg. If you are on Linux, give it a try.

Refer to: http://stackoverflow.com/questions/21378091/stream-record-program-vlc Also: http://stackoverflow.com/questions/21399999/record-stream-username-and-password-using-vlc And: http://stackoverflow.com/questions/21324785/ip-camera-access-using-opencv

And no, not an inconvenience. Helping helps :)

You are using C++? You can always open the exe files using the system() method. And if you just want to open the stream you can use cvCaptureFromFile() as I have mentioned in my answer. It works provided you provide the correct link and have all the codecs. :)

Some preprocessing techniques are:

1. Normalization of the images. You can read more about it here.
2. Also, you can read about smoothing the images and averaging.
3. After this you can also look up noise reduction.

All of these methods will help in the recognition process. Hope this helps :)

Well in that case you can use VLC to store the stream and then use VideoCapture::Open() to open the video. Tell me if this helps then I can edit my answer and you can accept it. :)

There might be a problem with the ffmpeg dll that is being used. Which build are you following? x86 or x86_64?

Follow the links given below for installation.

http://docs.opencv.org/doc/tutorials/introduction/android_binary_package/O4A_SDK.html

http://docs.opencv.org/doc/tutorials/introduction/android_binary_package/dev_with_OCV_on_Android.html

The link given here explains color detection and object tracking. The thing is, for object tracking the most basic algorithm is to compare two successive frames and on the basis of the comparison, find the difference in the position of the object, Now that you know the difference in the position you can track the object.

Hope this helps. :)

Add the code please.

Considering the question regarding identification of the ball, have you thought of edge detection as a possible way of detecting the ball?

Also, as your background is completely white, why not go for contour detection? It might pose a problem because you are using a basketball and it has dots on the surface for gripping the ball, but I think you can decide the size of the contours too, in that case you can use Contour detection as another method for detecting the ball. (Not too sure about this)

And Lastly, color detection, since your ball is orange and the color is not similar to any other color in the image you can use color detection, and once you have detected/identified the color you can apply edge detection for better detection of the ball. I came up with these three methods. Please not that the methods are computationally expensive.

Hope this helps :)

Okay, assuming that you have nothing currently on your system I am giving you the installation steps. Also not that the IDE that I use is eclipse, thus I am including the installation of eclipse in the procedure. Just follow the steps below and it should work.

1. Download python.
2. Download eclipse.
3. In eclipse go to Help->Install New Software.
4. There click on Add. The link that you are to provide is http://pydev.org/updates and the Name should be PyDev (I named it that, you can name it whatever you want to).
5. Select ONLY PyDev and install it.

Now follow the guide given here. It should work.

OpenCV team is working on a port for WindowsRT. It can be found here. Also, more details and info can be found here.

Hope this helps :)

Can you post your code please?

I have never done this but I think the function cvCaptureFromFile() can be used to do it. Don't know its current equivalent according to the newer documentation but you can use this. It works with some Network camera so ideally it should work.

Also, you can use VLC to stream and capture a stream too. Hope this helps :)

If you think the answer helps you then please accept it too :)

If you think the answer helps you then please accept it too :)

Which IDE are you using? And have you installed Python already?

Here is a blog post on that which might help you.

For shape detection you can use simple things like extracting features like edges, regions, textures, contours etc.

For color detection you can access the Scalar information of the image and then use it to detect whichever color you want and even decide the threshold.

And yes, as far as I know, both can be one simultaneously although the final program might be computationally expensive. Hope this helps :)

First of all:

Matching signatures is one of THE most difficult tasks. Why? Because each signature is different. Even if you compare the same signatures, the probability of them being the same is very less. So in your case, you will have to decide the threshold value.

About the method:

A very simple and crude algorithm might comprise of the following steps:

1.Divide the image into strips. Both the images, i.e. both the inputs.

2.Match the corresponding strips and find the similarity.

On the basis of the threshold decide if the images are same or not.

How does the camera send images? Do you have any idea about that?

You need to install the ffmpeg codec on your system. You can find it here

On installation and inclusion it should work.

You can apply a filter and use the RGB values to detect the level of the primary colors that are present in an image. After that you can edit the levels so as to reduce the yellow component in the image.

You can get the scalar values using cvGet2D (don't know what the name is in the current documentation) and use this to find out the levels. Then you can create your own Scalar from this and convert it into an image using cvSet2D (again check documentation for current name).

Hope this helps :)

Have you checked your CPU and RAM usage during the execution of the program? If so then what is the output? How busy is your system? Both of the above will give you some idea about what you are doing and where you are going wrong.

+1 Official IRC Channel NEEDED!

Given below is the code for accessing an IP camera using OpenCV. On running the program I get the error

Camera not found.

Error in opening cap_ffmpeg_impl.hpp

#include <opencv\cv.h>
#include <opencv2\highgui\highgui.hpp>
#include <opencv2\imgproc\imgproc.hpp>
#include <iostream>
#include <stdio.h>

using namespace std;
using namespace cv;

int main()
{
    Mat frame;
    VideoCapture cap("http://169.254.159.110/mjpg/1/video.mjpg");
    if(!cap.isOpened())
    {
        cout<<"Camera not found"<<endl;
        getchar();
        return -1;
    }
    namedWindow("video", 1);
    while (cap.isOpened())
    {
        cap >> frame;
        if(frame.empty()) break;
        imshow("video", frame);
        if(waitKey(30) >= 0) break;
    }
    return 0;
}

I have installed the ffmpeg codec using this guide.

I have also used iSpy software to check the correct link for connecting to the ip camera. It is still not working. I am using VS 2012 and Opencv 2.4.6.

Where am I going wrong?

I have downloaded the ANdroid SDK and Eclipse (ADT Bundle) from

http://developer.android.com/

Now I want to create an android app that uses OpenCV. I want to complete the setup so that I can use OpenCV functions in ANdroid apps. What do I do now?