seisgo / ellipsefit Goto Github PK

大佬您好，我刚刚接触C++不久，在配置环境这里一直是懵懵的。想借鉴你的代码实现瞳孔检测中的椭圆拟合，但是遇到了 无法找到源文件algo_global.h 的问题，同时提示无法打开源文件QVector，希望你能指点我一下。
环境为vs2017，QT5.12.1，clapack库已经配置好。

您好，我在Vs2019上配置了clapack3.2.1，运行您的程序时出现类似：
错误 C2065 “complex”: 未声明的标识符 (编译源文件 DirectEllipseFit.cpp) QtWeld C:\Program Files\CLAPACK\include\clapack.h 78
错误 C2065 “incx”: 未声明的标识符 (编译源文件 DirectEllipseFit.cpp) QtWeld C:\Program Files\CLAPACK\include\clapack.h 81
请问是我clapack的配置出问题导致的吗？

RANSAC filtering on each iteration is supposed to choose a minimal number of points to fit a model of ellipse and measure which other points agree with this model (inliers). In the current implementation, on each iteration, 3 datapoints are chosen, and two of them are treated as focal points of an ellipse. That is incorrect, because the goal is to identify the ellipse parameters from the sampled points, including the focal points.
In a general case, one needs at least 5 points to define an ellipse, less if there is info on the center:
https://math.stackexchange.com/questions/880258/minimal-number-of-points-to-define-a-rotated-ellipse
A possible solution is to change the function to sample 4 points (without replacement) from dataset. Use them to estimate the two foci, e.g. using one of the funcEllipseFit functions. That can be used to measure the distance from each of the sample points to the foci. All datapoints which distance to the foci is close to the average sample distance can be treated as 'inliers'.

大神，我有个实验用的是std::vector,不想再导入QT库了。所以想直接用std::vector。
还有一点，作者的clapack库里都用了什么函数？
如果只是一个“integer”类型，那么我不用clapack,直接改“integer”为“int”是否可行？

Hi @seisgo, thanks for the nice open source code contained in this repository!

It would be great if you could document under which software license (see https://choosealicense.com/ and https://help.github.com/en/articles/licensing-a-repository) this code is released.