Fix infinite loop on ArUco apriltag refinement

[buq2]

Software entered infinite loop when image height
was smaller than 10*cv::getNumThreads(). With high
core count machines this could happen with very
reasonable image sizes.

Fix is to ensure that chunksize is at least 1.

The bug can be triggered with following test program:

#include <opencv2/aruco.hpp>
#include <opencv2/core/utility.hpp>
#include <iostream>
#include <opencv2/imgcodecs.hpp>

int main(int argc, char *argv[]) {
    // Bug will happen even with 1 thread, but then the image
    // will be too small to detect corners.
    // cv::setNumThreads(1);
    const auto num_th = cv::getNumThreads();
    std::cout << "Number of threads: " << num_th << std::endl;

    // If 'freeze' is set to true, bug is triggered
    constexpr bool freeze = true;
    std::cout << "Should freeze: " << freeze << std::endl;

    // When triggering the bug, use image size that is
    // 1 pix smaller than 10*number of threads
    const auto height_img = freeze ? num_th*10-1 : num_th*10;
    std::cout << "Image height: " << height_img << std::endl;

    // Just to get nice white boarder
    const int shift = height_img > 10 ? 5 : 1;
    const auto height_marker = height_img-2*shift;

    // Create marker template
    cv::Mat img_marker;
    cv::Ptr<cv::aruco::Dictionary> dictionary = cv::aruco::getPredefinedDictionary(cv::aruco::DICT_4X4_50);
    cv::aruco::drawMarker(dictionary, 23, height_marker, img_marker, 1);

    // Copy to bigger image
    cv::Mat img(height_img, height_img, CV_8UC1);
    img.setTo(255);
    img_marker.copyTo(img(cv::Rect(shift, shift, height_marker, height_marker)));

    // Just for visualizing
    cv::imwrite("marker.png", img);

    // Bug only in apriltag refinement
    auto params = cv::aruco::DetectorParameters::create();
    params->cornerRefinementMethod = cv::aruco::CORNER_REFINE_APRILTAG;

    std::vector<std::vector<cv::Point2f>> corners;
    std::vector<int> ids;

    // If freeze=true, this will never return
    cv::aruco::detectMarkers(img, dictionary, corners, ids, params);

    // Ensure we get reasonable corners 
    // (not with this default test, white boarder needs to be bigger)
    for (const auto &marker : corners) {
        for (const auto &p : marker) {
            std::cout << p << std::endl;
        }
    }
}

Pull Request Readiness Checklist

See details at https://github.com/opencv/opencv/wiki/How_to_contribute#making-a-good-pull-request

• I agree to contribute to the project under Apache 2 License.
• To the best of my knowledge, the proposed patch is not based on a code under GPL or another license that is incompatible with OpenCV
• The PR is proposed to the proper branch
• There is a reference to the original bug report and related work
• There is accuracy test, performance test and test data in opencv_extra repository, if applicable
  Patch to opencv_extra has the same branch name.
• The feature is well documented and sample code can be built with the project CMake

[asmorkalov]

@buq2 Thanks for the contribution. The patch looks reasonable for me. Could you add your reproducer with setNumThreads to the PR as test?

[buq2]

@asmorkalov Do you mean adding the main function code from the initial PR description to (for example) https://github.com/opencv/opencv_contrib/blob/3.4/modules/aruco/test/test_arucodetection.cpp ? Sure! The "test" is not a very good test case as if it fails, it hangs indefinitely, but maybe it is better than nothing. I'll add it soonish and make it such that it will test also the other refinement methods with different initial number of threads with small aruco images.

[asmorkalov]

@buq2 Thanks for the test. I refactored it a bit, to make it parameterized (obvious test report in case of failure) and extracted the iteration with num_threads=1 as ground truth to make it obvious for reader.

[AleksandrPanov]

LGTM, thx for fix and tests

[buq2]

Thanks for the reviews and refactoring! Next one will be less painful for everyone :)

[asmorkalov]

@alalek I propose to merge the patch.

[AleksandrPanov]

warning: comparison between signed and unsigned integer expressions

sorry, I overlooked

[AleksandrPanov]

warning: comparison between signed and unsigned integer expressions

sorry, I overlooked