一、算法步骤
- 首先找到点集中y坐标最小的点作为初始点 p 0 p_0 p0,若y坐标相同,选取x坐标最小的一个点作为 p 0 p_0 p0
- 以p0为原点,对点集进行极角排序得到集合
{
p
0
,
p
1
.
.
.
p
n
}
\left\{p_0,p_1...p_n\right\}
{p0,p1...pn}
- 令i=1,构建一个存储凸包边界的栈stack,栈内初始存储元素 p 0 p_0 p0与 p 1 p_1 p1
- 判断集合下一个元素
p
i
+
1
p_{i+1}
pi+1在栈顶两元素构成的向量的哪一侧:
若在左侧,则点 p i + 1 p_{i+1} pi+1进栈, i = i + 1 ; i=i+1; i=i+1;重复执行步骤3,直至 i > n i>n i>n
若在右侧,则栈顶元素出栈,重复执行步骤3
二、算法示意图
三、代码
1.导入python库
from math import atan2
import math
import matplotlib.pyplot as plt
import numpy as np
2.构造工具函数
def atan(point,y,x):
'''返回点(x,y)相对于point的角度'''
x = x - point[0]
y = y - point[1]
if x==0 and y == 0:
return 0
point = (5,0) #表示x轴的向量,随便取
cos = (point[0]*x+point[1]*y)/(math.sqrt(point[0]**2+point[1]**2)*math.sqrt(x**2+y**2))
return np.arccos(cos)*(180/math.pi)
def angle_sort(p0,points):
'''对点集从小到大进行极角排序'''
dic = {}
for point in points:
angle = atan(p0,point[1],point[0])
dic[point] = angle
points = [k[0] for k in sorted(dic.items(),key=lambda x:x[1])] #,reverse=True
return points
def cross_product(a,b,c):
'''判断点c在由点a,b构成的向量的那一侧'''
result = a[0]*b[1]-a[1]*b[0] + b[0]*c[1]-b[1]*c[0] + c[0]*a[1]-c[1]*a[0]
if result < 0:
return False #点c在向量ab右边 返回False
else:
return True #点c在向量ab左边 返回True
3.初始化点集
集合的高度由vertical决定,宽度由level决定
key=3
num=120
level = 10
vertical = 5
seed = np.random.RandomState(key)
seed2 = np.random.RandomState(key+1)
Z1 = seed.rand(num,1)*level #生成点集
Z2 = seed2.rand(num,1)*vertical
Z = np.concatenate([Z1,Z2],axis = 1)
points = [tuple(i) for i in Z]
三、计算凸包
#起点为y坐标最小的点
ymin = min(points,key=lambda x:x[0])[1]
start = min([i for i in points if i[1] == ymin],key = lambda x:x[0])
stack=[] #定义栈
points = angle_sort(start,points)
stack.append(points[0])
stack.append(points[1])
i=2
while len(stack)!=0 and i != len(points):
if cross_product(stack[len(stack)-2],stack[len(stack)-1],points[i]):
stack.append(points[i])
i+=1
else:
stack.pop()
if len(stack)<2:
stack.append(points[i])
i+=1
continue
四、画出凸包
stack.append(stack[0])
x = [i[0] for i in points]
y = [i[1] for i in points]
x0 = [i[0] for i in stack]
y0 = [i[1] for i in stack]
plt.figure(figsize=(10,10))
plt.scatter(x,y)
plt.plot(x0,y0)
plt.xlim(-1,11)
plt.ylim(-1,6)
i=0
for a,b in zip(x,y):
plt.text(a, b, i)
i+=1
该算法参考自算法导论第33章凸包的生成