找出y最小的顶点(如果有平手，则x最大)。假设顶点是A，列表中的前一个顶点是B，列表中的下一个顶点是c。现在计算AB和AC的叉乘的符号。

引用:

如何确定一个简单多边形的方向?在 常见问题:计算机。图形。算法。 维基百科的曲线定位。

下面是基于这个答案的一个简单的Python 3实现(反过来，它是基于已接受答案中提出的解决方案)

def is_clockwise(points):
    # points is your list (or array) of 2d points.
    assert len(points) > 0
    s = 0.0
    for p1, p2 in zip(points, points[1:] + [points[0]]):
        s += (p2[0] - p1[0]) * (p2[1] + p1[1])
    return s > 0.0

正如这篇维基百科文章中所解释的曲线方向，给定平面上的3个点p, q和r(即x和y坐标)，您可以计算以下行列式的符号

如果行列式为负(即定向(p, q, r) < 0)，则多边形是顺时针方向(CW)。如果行列式为正(即定向(p, q, r) > 0)，则多边形是逆时针方向(CCW)。如果点p, q和r共线，行列式为零(即定向(p, q, r) == 0)。

在上面的公式中，由于我们使用的是齐次坐标，我们将1放在p, q和r的坐标前面。

对于那些不想“重新发明轮子”的人，我认为值得一提的是，这个检查是在一个名为Shapely (github)的漂亮的Python包中实现的(它基于GEOS C/ c++库):

Shapely is a BSD-licensed Python package for manipulation and analysis of planar geometric objects. It is using the widely deployed open-source geometry library GEOS (the engine of PostGIS, and a port of JTS). Shapely wraps GEOS geometries and operations to provide both a feature rich Geometry interface for singular (scalar) geometries and higher-performance NumPy ufuncs for operations using arrays of geometries. Shapely is not primarily focused on data serialization formats or coordinate systems, but can be readily integrated with packages that are.

来源:https://shapely.readthedocs.io/en/stable/

一个给出OP坐标的小例子:

import numpy as np
from shapely.geometry import Polygon

points = np.array([
    (5,0),
    (6,4),
    (4,5),
    (1,5),
    (1,0)
])

P = Polygon(points)

这是新构造的多边形:

import matplotlib.pyplot as plt

x,y = P.exterior.coords.xy
plt.plot(x,y)
plt.axis('equal')
plt.grid()
plt.show()

你可以直接使用LinearRing的is_ccw属性来检查多边形是CW还是CCW:

type(P.exterior)
>: shapely.geometry.polygon.LinearRing

P.exterior.is_ccw
>: True

如果颠倒:

points = np.flipud(points)
points
>: 
array([[1, 0],
       [1, 5],
       [4, 5],
       [6, 4],
       [5, 0]])


P1 = Polygon(points)

P1.exterior.is_ccw
>: True

进一步阅读的文档和参考资料:

shaely is_ccw (github): https://github.com/shapely/shapely/blob/eba985c6e0170ecdd90c83592fd0afa7ae793cb8/shapely/predicates.py#L72-L108 Libgeos (github): https://github.com/libgeos/geos GEOS API参考:https://libgeos.org/doxygen/classgeos_1_1algorithm_1_1Orientation.html#a5af93795969b80f97d7997195974d7c8 GEOS实现(github): https://github.com/libgeos/geos/blob/ab0ce6dafdf7f75ec6d234b6c65bb209037dda17/src/algorithm/Orientation.cpp#L43-L133

下面是一个基于@Beta答案的算法的简单c#实现。

让我们假设我们有一个Vector类型，它的X和Y属性为double类型。

public bool IsClockwise(IList<Vector> vertices)
{
    double sum = 0.0;
    for (int i = 0; i < vertices.Count; i++) {
        Vector v1 = vertices[i];
        Vector v2 = vertices[(i + 1) % vertices.Count];
        sum += (v2.X - v1.X) * (v2.Y + v1.Y);
    }
    return sum > 0.0;
}

%是执行模运算的模运算符或余数运算符，该运算符(根据维基百科)在一个数除以另一个数后求余数。

根据@MichelRouzic评论的优化版本:

double sum = 0.0;
Vector v1 = vertices[vertices.Count - 1]; // or vertices[^1] with
                                          // C# 8.0+ and .NET Core
for (int i = 0; i < vertices.Count; i++) {
    Vector v2 = vertices[i];
    sum += (v2.X - v1.X) * (v2.Y + v1.Y);
    v1 = v2;
}
return sum > 0.0;

这不仅节省了模运算%，还节省了数组索引。

测试(参见与@WDUK的讨论)

public static bool IsClockwise(IList<(double X, double Y)> vertices)
{
    double sum = 0.0;
    var v1 = vertices[^1];
    for (int i = 0; i < vertices.Count; i++) {
        var v2 = vertices[i];
        sum += (v2.X - v1.X) * (v2.Y + v1.Y);
        Console.WriteLine($"(({v2.X,2}) - ({v1.X,2})) * (({v2.Y,2}) + ({v1.Y,2})) = {(v2.X - v1.X) * (v2.Y + v1.Y)}");
        v1 = v2;
    }
    Console.WriteLine(sum);
    return sum > 0.0;
}

public static void Test()
{
    Console.WriteLine(IsClockwise(new[] { (-5.0, -5.0), (-5.0, 5.0), (5.0, 5.0), (5.0, -5.0) }));

    // infinity Symbol
    //Console.WriteLine(IsClockwise(new[] { (-5.0, -5.0), (-5.0, 5.0), (5.0, -5.0), (5.0, 5.0) }));
}

虽然这些答案是正确的，但它们在数学上的强度比必要的要大。假设地图坐标，其中最北的点是地图上的最高点。找到最北的点，如果两个点相等，它是最北的，然后是最东的(这是lhf在他的答案中使用的点)。在你的观点中，

点[0]= (5,0)

点[1]= (6,4)

点[2]= (4,5)

点[3]= (1,5)

点[4]= (1,0)

If we assume that P2 is the most north then east point either the previous or next point determine clockwise, CW, or CCW. Since the most north point is on the north face, if the P1 (previous) to P2 moves east the direction is CW. In this case, it moves west, so the direction is CCW as the accepted answer says. If the previous point has no horizontal movement, then the same system applies to the next point, P3. If P3 is west of P2, it is, then the movement is CCW. If the P2 to P3 movement is east, it's west in this case, the movement is CW. Assume that nte, P2 in your data, is the most north then east point and the prv is the previous point, P1 in your data, and nxt is the next point, P3 in your data, and [0] is horizontal or east/west where west is less than east, and [1] is vertical.

if (nte[0] >= prv[0] && nxt[0] >= nte[0]) return(CW);
if (nte[0] <= prv[0] && nxt[0] <= nte[0]) return(CCW);
// Okay, it's not easy-peasy, so now, do the math
if (nte[0] * nxt[1] - nte[1] * nxt[0] - prv[0] * (nxt[1] - crt[1]) + prv[1] * (nxt[0] - nte[0]) >= 0) return(CCW); // For quadrant 3 return(CW)
return(CW) // For quadrant 3 return (CCW)