Installing OpenCV videoio library and build from source failed

Robotics & Edge Computing Jetson & Embedded Systems Jetson TK1
1

Using Ubuntu and Jetpack 3 and the TK1 along with the OV5640 camera, I am developing a machine vision project. From /home/ubuntu/OpenCV4Tegra I ran the shell script to install OpenCV4Tegra.
I have linked up necessary libraries and build paths in Nsight and whatever OpenCV libraries I have work on the machine. I need to use cv::VideoCapture along with gstreamer to capture images and process them. I already have built my gstreamer pipeline and gotten it working.

I have already tried the following:

  1. following this link - https://elinux.org/Jetson/Installing_OpenCV but option 1 (using prebuilt libraries) does not work. running
sudo dpkg -i libopencv4tegra-repo_l4t-r21_2.4.10.1_armhf.deb

returns no such file or directory.
2. Building from source from both the elinux site and from the openCV documentation. I git cloned both the openCV source files and the OpenCV_contrib repository, then I updated cmake to the latest version and everytime I try to run Cmake with the any sort of flags and options I get the same error message.

Build output check failed:
    Regex: 'command line option .* is valid for .* but not for C\+\+'

I am out of ideas and I need this to be working as soon as possible. All I need is the videoio .so library installed correctly but I am not sure how to do that without completely installing openCV from source. If there is a way to just link that library please let me know. Otherwise I need to get OpenCV built from source on this machine.
Thanks

2

I should point out that this all started because I tried to run this code written in Nsight

VideoCapture cap(gst_pipeline, CAP_GSTREAMER); // open the pipeline
    // View video
    *cv::Mat frame;
    while (1) {
        cap >> frame;  // Get a new frame from camera

        // Display frame
        imshow("Display window", frame);
        cv::waitKey(1); //needed to show frame
    }

and I got this error:

../src/test.cu(110): error: identifier "CAP_GSTREAMER" is undefined

../src/test.cu(110): error: no instance of constructor "cv::VideoCapture::VideoCapture" matches the argument list
            argument types are: (const std::string, <error-type>)

So if OpenCV4Tegra is not the problem then maybe I did something wrong here but from what I saw on the OpenCV documentation the CAP_GSTREAMER is enumerated in the

videoio.hpp

file

3

Hi,

Do you install opencv4tegra with JetPack installer?
It looks like you want to build it from source.

Could you try to reflash the system and install the OpenCV package from JetPack?
[url]https://developer.nvidia.com/embedded/jetpack-2_3_1[/url]

Thanks.

4

I had initially installed OpenCV4Tegra with the JetPack installer only to realize that the videio library is missing from the installation. I need the videio library for my application which is why I was trying to build from source

5

As far as I remember OpenCV4Tegra hadn’t gstreamer support, so you would have to build and install your own version. I’d suggest to use at least a 3.2 version so that you get some carotene optimizations…
I haven’t tried opencv4 on TK1, but if you want to give it a try be aware that since opencv4 CUDA stuff is now in opencv_contrib.

For the compilation error, with opencv-2.4 it might be CV_CAP_GSTREAMER, but not included in Opencv4tegra. Since opencv3, you may try to use cv namespace instead: cv::CAP_GSTREAMER. That being said, you don’t need to specify this, it should automatically detect a gstreamer pipeline.

You may also face some issues if you keep the opencv4tegra install. You may check this topic if you want to keep it.