聚类+分类器python代码_聚类分类python代码

from sklearn.cluster import DBSCAN
import matplotlib.pyplot as plt
from matplotlib.patches import Rectangle
import open3d as o3d
import numpy as np
 
# Reading data
inputPath = r'E:\迅雷下载\kitti-velodyne-viewer-matlab-main\training\velodyne\000020.bin'
num = np.fromfile(inputPath, dtype='float32', count=-1, sep='', offset=0)
new = np.asarray(num).reshape(-1, 4)
 
# Assigning data to different variables
X = num[0::4]
Y = num[1::4]
Z = num[2::4]
W = num[3::4]
 
# Creating point cloud
xyz = np.zeros((np.size(X), 3))
xyz[:, 0] = X
xyz[:, 1] = Y
xyz[:, 2] = Z
pc = o3d.geometry.PointCloud()
pc.points = o3d.utility.Vector3dVector(xyz)
 
# Downsampling
downpcd = pc.voxel_down_sample(voxel_size=0.4)  # Adjust voxel size
 
# Segmentation
plane_model, inliers = downpcd.segment_plane(distance_threshold=1, ransac_n=3, num_iterations=1000)  # Adjust distance_threshold
[a, b, c, d] = plane_model
print(f"Plane equation: {a:.2f}x + {b:.2f}y + {c:.2f}z + {d:.2f} = 0")
inlier_cloud = downpcd.select_by_index(inliers)
outlier_cloud = downpcd.select_by_index(inliers, invert=True)
 
# Clustering using DBSCAN
clustering = DBSCAN(eps=2, min_samples=10).fit(np.asarray(outlier_cloud.points))
labels = clustering.labels_
 
max_label = labels.max()
print(f"point cloud has {max_label + 1} clusters")
 
# Visualization with bounding boxes
fig = plt.figure()
ax = fig.add_subplot(111)
 
car_cluster_labels = []
for cluster_label in range(max_label + 1):
    cluster_points = np.asarray(outlier_cloud.points)[labels == cluster_label]
    if len(cluster_points) > 0:
        min_x, min_y, min_z = np.min(cluster_points, axis=0)
        max_x, max_y, max_z = np.max(cluster_points, axis=0)
        width = max_x - min_x
        height = max_y - min_y
        depth = max_z - min_z
 
        volume = width * height * depth
        # Assume a car's volume is between 2 and 10 m^3 and z range between -1.5 and -0.5
        if 2 < volume < 10 and -1.5 < min_z < -0.5:
            car_cluster_labels.append(cluster_label)
            rect = Rectangle((min_x, min_y), width, height, linewidth=1, edgecolor='r', facecolor='none')
            ax.add_patch(rect)
            ax.text(min_x + width / 2, min_y + height / 2, 'car', color='blue', ha='center', va='center')
    # Plot points with different colors for each cluster
    ax.scatter(cluster_points[:, 0], cluster_points[:, 1], s=5, label=f"Cluster {cluster_label} - Car" if cluster_label in car_cluster_labels else f"Cluster {cluster_label}")
print(f"Found {len(car_cluster_labels)} car clusters")
# Show the legend to differentiate clusters
ax.legend()
# Show the 2D plot with bounding boxes
ax.set_aspect('equal', adjustable='box')
plt.show()
# Show the 3D point cloud with bounding boxes
o3d.visualization.draw_geometries([outlier_cloud])