DBSCAN 聚类 和 gmm 聚类 测试

测试代码

import math
import numpy as np
import matplotlib.pyplot as plt
from sklearn.cluster import DBSCAN
from sklearn.mixture import GaussianMixture

# 给定的线段数据（x1, y1, x2, y2）
lines = [
    # (265, 106, 268, 1),
    (265, 105, 268, 0),  # 点1: (265, 105), (268, 0)
    (246, 62, 247, 7),   # 点2: (246, 62), (247, 7)
    (246, 60, 246, 5)    # 点3: (246, 60), (246, 5)
]
lines_polar=[]
for line in lines:
    x1, y1, x2, y2 = line
    dx = x2 - x1
    dy = y2 - y1
    if dx == 0:
        x_value_when_y_0 = x1
        if dy != 0:
            angle_radians= math.pi / 2  # 对应90度，转换为弧度制为pi/2
        else:
            angle_radians= 0
    else:
        slope = (y2 - y1) / (x2 - x1)
        if slope == 0:
            x_value_when_y_0 = x1
        else:
            x_value_when_y_0 = x1 - y1 / slope
            m = dy / dx
            angle_radians = math.atan(m)
    lines_polar.append((x_value_when_y_0, angle_radians*2))

# 提取所有点 (x1, y1) 和 (x2, y2)

# 将点转换为 NumPy 数组
points = np.array(lines)

# DBSCAN 聚类
def dbscan_cluster(points, eps=6, min_samples=2):
    dbscan = DBSCAN(eps=eps, min_samples=min_samples)
    dbscan.fit(points)
    return dbscan.labels_

# GMM 聚类
def gmm_cluster(points, n_components=2):
    gmm = GaussianMixture(n_components=n_components)
    gmm.fit(points)
    labels = gmm.predict(points)
    return labels

# 比较 DBSCAN 和 GMM 聚类
def compare_clustering(points):
    # 使用 DBSCAN 聚类
    dbscan_labels = dbscan_cluster(points)
    
    # 使用 GMM 聚类
    gmm_labels = gmm_cluster(points)
    
    # 打印聚类标签
    print("DBSCAN 聚类标签:", dbscan_labels)
    print("GMM 聚类标签:", gmm_labels)
   
# 主函数
if __name__ == "__main__":
    compare_clustering(points)

  经过测试发现  ，当需要聚类的点数小于4个时，DBSCAN 聚类 效果很差，gmm效果较为不错。