图聚类 - 地图建模

图论的基础是三元组，点集，边集，权重函数。然而在实际的地图中，我们只知道地图（row，col）的值，可能表示海拔，也可能表示坡度，形式上就是一个二元矩阵。

形如此图：

表示的矩阵：

在这个矩阵中，255表示普通节点（不是障碍物），0表示障碍物。

转换为邻接矩阵，并画出点线图（代码）：

import cv2
import numpy as np
# from MCL import markovCluster
import matplotlib.pyplot as py

obstacle = 0

unObstacle = 255


# 读取地图数据
def read_map_from_pic(file_path):
    pic = cv2.imread(file_path, cv2.IMREAD_GRAYSCALE)
    return pic


# 判断两个节点是否有相连的边
def is_neighbour(position1, position2):
    # 检查两点元素是否是障碍物
    if map[position1.row][position1.col] == obstacle or map[position2.row][position2.col] == obstacle:
        return False
    if abs(position1.row-position2.row) < 2 and abs(position1.col - position2.col) < 2:
        return True
    else:
        return False


def getNodeSetByMap(map):
    nodeSet = []
    for row in range(map.shape[0]):
        for col in range(map.shape[1]):
            if map[row][col] == unObstacle:
                nodeSet.append(Node(Position(row, col)))
    return nodeSet



def getAdjacencyMatrix(nodeSet):
    edgeList = []
    adjacencyMatrix = np.zeros((len(nodeSet), len(nodeSet)))
    for row in range(len(nodeSet)):
        for col in range(len(nodeSet)):
            adjacencyMatrix[row][col] = 1 if is_neighbour(nodeSet[row].position, nodeSet[col].position) else 0
            if adjacencyMatrix[row][col] == 1:
                xList = [nodeSet[row].position.row, nodeSet[col].position.row]
                yList = [nodeSet[row].position.col, nodeSet[col].position.col]
                edgeList.append([xList, yList])
    return adjacencyMatrix, edgeList


def paintGraph(map, edgeList):
    rowList = []
    colList = []
    for row in range(map.shape[0]):
        for col in range(map.shape[1]):
            if map[row][col] == unObstacle:
               rowList.append(row)
               colList.append(col)
    py.scatter(rowList, colList, color='silver')
    for i in range(len(edgeList)):
        py.plot(edgeList[i][0], edgeList[i][1], color='cornflowerblue', linewidth=1, alpha=0.08)

    py.show()

# 三元组存储稀疏矩阵（领接矩阵）
class Triple:

    def __init__(self, position, value) -> None:
        super().__init__()
        self.position = position
        self.value = value


class Position:

    def __init__(self, row, col) -> None:
        super().__init__()
        self.row = row
        self.col = col


class Node:

    def __init__(self, position) -> None:
        super().__init__()
        self.position = position

# MCL聚类


if __name__ == "__main__":
    file_path = r"E:\python_code\pytorch_learn\data\tsp\pse_cvt1.png"
    map = read_map_from_pic(file_path)
    position1 = Position(4, 1)
    position2 = Position(3, 0)
    print(is_neighbour(position1, position2))
    # 图论点集
    nodeSet = getNodeSetByMap(map)
    # 构造邻接矩阵
    adjacencyMat, edgeList = getAdjacencyMatrix(nodeSet)
    print(adjacencyMat.shape)
    # 画图
    paintGraph(map, edgeList)
    #dimension = 4
    #numIter = 10
    #inflateMat = markovCluster(adjacencyMat, dimension, numIter)
    #print(inflateMat)

输出：

谱聚类

sci_learn

马尔可夫链