凸包算法——Graham Scan、Jarvis March（Gift Wrapping）实现

凸包算法的目的是从一个点集计算出最小的凸多边形，包围所有点。

1. Graham Scan

• 步骤:

  1. 选择基准点: 选择 y 坐标最小的点（如果有多个，选择 x 坐标最小的点）。
  2. 按极角排序: 计算所有点相对于基准点的极角，并按极角排序。
  3. 构建凸包: 使用栈来维护凸包的顶点。逐一检查点，保证凸包是凸的，弹出不符合条件的点，最终栈中的点即为凸包的顶点。

• 时间复杂度: O(n log n)（排序步骤占主导）

代码：

import numpy as np

def graham_scan(points):
    points = sorted(points,key = lambda p:(p[0],p[1]))
    def orientation(p,q,r):
        return (q[1] - p[1]) * (r[0] - q[0]) - (q[0] - p[0]) * (r[1] -q[1])
    
    def build_hull(points):
        hull = []
        for p in points:
            while len(hull) >= 2 and orientation(hull[-2],hull[-1],p) <= 0:
                hull.pop()
            hull.append(p)
        return hull

    lower = build_hull(points)
    upper = build_hull(reversed(points))

    return np.array(lower[:-1] + upper[:-1])


points = np.array([[-5.90257, -0.22237], [1.03258, -0.533936], [0.907549, -0.134126], 
                   [0.717333, 0.471533], [0.713893, 0.484054], [0.708021, 0.510585], 
                   [0.668892, 0.637087], [0.5562, -0.748886], [0.552133, 1.01539], 
                   [-1.36528, 1.202], [-1.49895, 1.6176], [-1.50235, 1.62983], 
                   [-1.50702, 1.64434], [-1.51095, 1.65558], [-1.51498, 1.66897], 
                   [-1.51726, 1.67879], [-1.5189, 1.68445], [-1.54057, 1.75716], 
                   [-1.5986, -1.40478], [-2.28137, -1.61885], [-2.43108, -1.66877], 
                   [-2.49418, 1.55916], [-4.34462, -2.27342], [-4.35825, 0.963766], 
                   [-4.43426, -2.3003], [-5.8972, -0.250361]])
hull_points = graham_scan(points)

2. Jarvis March (Gift Wrapping)

• 步骤:

  1. 选择起点: 选择最左边的点作为起点。
  2. 寻找下一个点: 从当前点开始，选择相对于当前点所有其他点中极角最小的点作为下一个凸包点。
  3. 重复直到封闭: 继续选择极角最小的点，直到回到起点形成闭环。

• 时间复杂度: O(nh)（h 为凸包的顶点数，最坏情况下为 O(n^2)）

代码：

import numpy as np

def jarvis_march(points):
    hull = []
    leftmost = np.argmin(points[:, 0])
    point_on_hull = leftmost
    while True:
        hull.append(point_on_hull)
        endpoint = (point_on_hull + 1) % len(points)
        for i in range(len(points)):
            if np.cross(points[endpoint] - points[point_on_hull], points[i] - points[point_on_hull]) < 0:
                endpoint = i
        point_on_hull = endpoint
        if endpoint == leftmost:
            break
    return np.array(hull)

points = np.array([[-5.90257, -0.22237], [1.03258, -0.533936], [0.907549, -0.134126], 
                   [0.717333, 0.471533], [0.713893, 0.484054], [0.708021, 0.510585], 
                   [0.668892, 0.637087], [0.5562, -0.748886], [0.552133, 1.01539], 
                   [-1.36528, 1.202], [-1.49895, 1.6176], [-1.50235, 1.62983], 
                   [-1.50702, 1.64434], [-1.51095, 1.65558], [-1.51498, 1.66897], 
                   [-1.51726, 1.67879], [-1.5189, 1.68445], [-1.54057, 1.75716], 
                   [-1.5986, -1.40478], [-2.28137, -1.61885], [-2.43108, -1.66877], 
                   [-2.49418, 1.55916], [-4.34462, -2.27342], [-4.35825, 0.963766], 
                   [-4.43426, -2.3003], [-5.8972, -0.250361]])

hull_indices = jarvis_march(points)
hull_points = points[hull_indices]

print("Convex Hull points:", hull_points)