feat(All):Initial
This commit is contained in:
@@ -0,0 +1,336 @@
|
||||
/**
|
||||
* @author ErSan
|
||||
* @email mlt131220@163.com
|
||||
* @date 2025/07/03 19:38
|
||||
* @description 点云语义重建
|
||||
*/
|
||||
import * as THREE from 'three';
|
||||
import {ConvexGeometry} from "three/addons/geometries/ConvexGeometry.js";
|
||||
|
||||
class PointCloudReconstructor {
|
||||
// 点云数据存储
|
||||
points: THREE.Vector3[] = [];
|
||||
colors: THREE.Color[] = [];
|
||||
// 语义颜色分组
|
||||
colorGroups:Map<string,THREE.Vector3[]> = new Map();
|
||||
|
||||
colorTolerance: number;
|
||||
distanceThreshold: number;
|
||||
minClusterSize: number;
|
||||
downsampleResolution: number;
|
||||
|
||||
progressCallback: Function | null;
|
||||
|
||||
constructor() {
|
||||
// RGB颜色容差
|
||||
this.colorTolerance = 5;
|
||||
// 空间聚类距离阈值
|
||||
this.distanceThreshold = 0.3;
|
||||
// 最小聚类点数
|
||||
this.minClusterSize = 10;
|
||||
// 点云抽稀系数
|
||||
this.downsampleResolution = 0.05;
|
||||
|
||||
// 进度回调
|
||||
this.progressCallback = null;
|
||||
}
|
||||
|
||||
/**
|
||||
* 处理几何数据，提取点和颜色
|
||||
* @param {THREE.BufferGeometry} geometry - 点云几何数据
|
||||
*/
|
||||
processGeometry(geometry) {
|
||||
this.points = [];
|
||||
this.colors = [];
|
||||
|
||||
const vertices = geometry.getAttribute('position').array;
|
||||
const colors = geometry.getAttribute('color')?.array || [];
|
||||
const hasColors = colors.length > 0;
|
||||
|
||||
for (let i = 0; i < vertices.length; i += 3) {
|
||||
this.points.push(new THREE.Vector3(
|
||||
vertices[i],
|
||||
vertices[i + 1],
|
||||
vertices[i + 2]
|
||||
));
|
||||
|
||||
if (hasColors) {
|
||||
this.colors.push(new THREE.Color(
|
||||
colors[i],
|
||||
colors[i + 1],
|
||||
colors[i + 2]
|
||||
));
|
||||
} else {
|
||||
this.colors.push(new THREE.Color(0xffffff));
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// 点云抽稀算法
|
||||
downsamplePoints(points, resolution) {
|
||||
const grid = new Map();
|
||||
const downsampled:THREE.Vector3[] = [];
|
||||
|
||||
points.forEach(point => {
|
||||
const gridX = Math.floor(point.x / resolution);
|
||||
const gridY = Math.floor(point.y / resolution);
|
||||
const gridZ = Math.floor(point.z / resolution);
|
||||
const gridKey = `${gridX},${gridY},${gridZ}`;
|
||||
|
||||
if (!grid.has(gridKey)) {
|
||||
grid.set(gridKey, true);
|
||||
downsampled.push(point.clone());
|
||||
}
|
||||
});
|
||||
|
||||
return downsampled;
|
||||
}
|
||||
|
||||
/**
|
||||
* 按语义颜色分组点云
|
||||
*/
|
||||
groupBySemanticColor() {
|
||||
this.colorGroups.clear();
|
||||
|
||||
for (let i = 0; i < this.points.length; i++) {
|
||||
const color = this.colors[i];
|
||||
const colorKey = this.getColorKey(color);
|
||||
|
||||
if (!this.colorGroups.has(colorKey)) {
|
||||
this.colorGroups.set(colorKey, []);
|
||||
}
|
||||
|
||||
(this.colorGroups.get(colorKey) as THREE.Vector3[]).push(this.points[i].clone());
|
||||
|
||||
// 更新进度（分组阶段占20%）
|
||||
if (this.progressCallback && i % 100 === 0) {
|
||||
this.progressCallback(20 * i / this.points.length, `分组点云: ${i}/${this.points.length}`);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
/**
|
||||
* 生成颜色分类键值（考虑容差）
|
||||
* @param {THREE.Color} color - 输入颜色
|
||||
* @returns {string} 颜色分类键
|
||||
*/
|
||||
getColorKey(color) {
|
||||
const r = Math.round(color.r * 255 / this.colorTolerance) * this.colorTolerance;
|
||||
const g = Math.round(color.g * 255 / this.colorTolerance) * this.colorTolerance;
|
||||
const b = Math.round(color.b * 255 / this.colorTolerance) * this.colorTolerance;
|
||||
return `${r},${g},${b}`;
|
||||
}
|
||||
|
||||
/**
|
||||
* 空间聚类算法（使用网格加速）
|
||||
* @param {THREE.Vector3[]} points - 输入点集
|
||||
* @returns {THREE.Vector3[][]} 聚类结果
|
||||
*/
|
||||
spatialClustering(points:THREE.Vector3[]) {
|
||||
const clusters:THREE.Vector3[][] = [];
|
||||
const visited = new Set();
|
||||
const grid = new Map();
|
||||
const gridSize = this.distanceThreshold * 1.5;
|
||||
|
||||
// 创建空间网格
|
||||
for (let i = 0; i < points.length; i++) {
|
||||
const point = points[i];
|
||||
const gridKey = this.getGridKey(point, gridSize);
|
||||
|
||||
if (!grid.has(gridKey)) {
|
||||
grid.set(gridKey, []);
|
||||
}
|
||||
|
||||
grid.get(gridKey).push(i);
|
||||
}
|
||||
|
||||
let processed = 0;
|
||||
const totalPoints = points.length;
|
||||
|
||||
for (let i = 0; i < points.length; i++) {
|
||||
if (visited.has(i)) continue;
|
||||
|
||||
visited.add(i);
|
||||
const cluster = [points[i]];
|
||||
const queue = [i];
|
||||
|
||||
while (queue.length > 0) {
|
||||
const currentIndex = queue.shift() as number;
|
||||
const currentPoint = points[currentIndex];
|
||||
|
||||
// 获取当前点所在的网格及其相邻网格
|
||||
const neighborCells = this.getNeighborCells(currentPoint, gridSize);
|
||||
|
||||
for (const cell of neighborCells) {
|
||||
if (!grid.has(cell)) continue;
|
||||
|
||||
const cellPoints = grid.get(cell);
|
||||
for (const neighborIndex of cellPoints) {
|
||||
if (visited.has(neighborIndex)) continue;
|
||||
|
||||
const neighbor = points[neighborIndex];
|
||||
if (currentPoint.distanceTo(neighbor) < this.distanceThreshold) {
|
||||
visited.add(neighborIndex);
|
||||
cluster.push(neighbor);
|
||||
queue.push(neighborIndex);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if (cluster.length >= this.minClusterSize) {
|
||||
clusters.push(cluster);
|
||||
}
|
||||
|
||||
// 更新进度（聚类阶段占30%）
|
||||
processed++;
|
||||
if (this.progressCallback && processed % 10 === 0) {
|
||||
const progress = 20 + 30 * processed / totalPoints;
|
||||
this.progressCallback(progress, `空间聚类: ${processed}/${totalPoints}`);
|
||||
}
|
||||
}
|
||||
|
||||
return clusters;
|
||||
}
|
||||
|
||||
// 获取网格键值
|
||||
getGridKey(point, gridSize) {
|
||||
const x = Math.floor(point.x / gridSize);
|
||||
const y = Math.floor(point.y / gridSize);
|
||||
const z = Math.floor(point.z / gridSize);
|
||||
return `${x},${y},${z}`;
|
||||
}
|
||||
|
||||
// 获取相邻网格
|
||||
getNeighborCells(point, gridSize) {
|
||||
const cells:string[] = [];
|
||||
const x = Math.floor(point.x / gridSize);
|
||||
const y = Math.floor(point.y / gridSize);
|
||||
const z = Math.floor(point.z / gridSize);
|
||||
|
||||
for (let dx = -1; dx <= 1; dx++) {
|
||||
for (let dy = -1; dy <= 1; dy++) {
|
||||
for (let dz = -1; dz <= 1; dz++) {
|
||||
cells.push(`${x + dx},${y + dy},${z + dz}`);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
return cells;
|
||||
}
|
||||
|
||||
/**
|
||||
* 重建三角面模型（分帧处理）
|
||||
* @returns {THREE.Group} 包含所有重建物体的场景组
|
||||
*/
|
||||
reconstruct(onProgress, onComplete) {
|
||||
// 更新进度
|
||||
let progress = 0;
|
||||
onProgress(progress, `分组点云...`);
|
||||
|
||||
// 第一阶段：分组点云
|
||||
this.groupBySemanticColor();
|
||||
const objectGroup = new THREE.Group();
|
||||
const colorKeys = Array.from(this.colorGroups.keys());
|
||||
let totalObjects = 0;
|
||||
|
||||
// 创建一个任务队列
|
||||
const tasks: {colorKey:string,cluster:THREE.Vector3[]}[] = [];
|
||||
|
||||
progress = 10;
|
||||
onProgress(progress, `分组点云完毕`);
|
||||
|
||||
const progressItem = Math.floor(40 / colorKeys.length);
|
||||
|
||||
for (const colorKey of colorKeys) {
|
||||
const points = this.colorGroups.get(colorKey);
|
||||
|
||||
// 对点云进行抽稀
|
||||
const downsampledPoints = this.downsamplePoints(points, this.downsampleResolution);
|
||||
|
||||
// 空间聚类
|
||||
const clusters = this.spatialClustering(downsampledPoints);
|
||||
|
||||
for (const cluster of clusters) {
|
||||
tasks.push({
|
||||
colorKey,
|
||||
cluster
|
||||
});
|
||||
}
|
||||
|
||||
progress += progressItem;
|
||||
onProgress(progress, `空间聚类...`);
|
||||
}
|
||||
|
||||
let taskIndex = 0;
|
||||
const totalTasks = tasks.length;
|
||||
let lastUpdateTime = performance.now();
|
||||
|
||||
// 更新进度
|
||||
progress = 50;
|
||||
onProgress(progress, `重建对象: ${taskIndex}/${totalTasks}`);
|
||||
|
||||
// 分帧处理函数
|
||||
const processNextTask = () => {
|
||||
const now = performance.now();
|
||||
const elapsed = now - lastUpdateTime;
|
||||
|
||||
// 控制处理速度（每帧最多处理10个任务）
|
||||
const maxTasksPerFrame = Math.min(10, Math.max(1, Math.floor(elapsed / 5)));
|
||||
let processedThisFrame = 0;
|
||||
|
||||
while (taskIndex < totalTasks && processedThisFrame < maxTasksPerFrame) {
|
||||
const task = tasks[taskIndex];
|
||||
const { colorKey, cluster } = task;
|
||||
|
||||
try {
|
||||
// 使用凸包算法生成表面
|
||||
const geometry = new ConvexGeometry(cluster);
|
||||
|
||||
// 从颜色键解析原始颜色
|
||||
const [r, g, b] = colorKey.split(',').map(Number);
|
||||
const material = new THREE.MeshStandardMaterial({
|
||||
color: new THREE.Color(r / 255, g / 255, b / 255),
|
||||
flatShading: true,
|
||||
side: THREE.DoubleSide
|
||||
});
|
||||
|
||||
const mesh = new THREE.Mesh(geometry, material);
|
||||
objectGroup.add(mesh);
|
||||
totalObjects++;
|
||||
} catch (e) {
|
||||
console.warn("凸包生成失败:", e);
|
||||
}
|
||||
|
||||
taskIndex++;
|
||||
processedThisFrame++;
|
||||
|
||||
// 更新进度（重建阶段占50%）
|
||||
const status = `重建对象: ${taskIndex}/${totalTasks}`;
|
||||
onProgress(progress + 50 * taskIndex / totalTasks, status);
|
||||
}
|
||||
|
||||
// 所有任务完成
|
||||
if (taskIndex >= totalTasks) {
|
||||
onComplete(objectGroup, totalObjects);
|
||||
return;
|
||||
}
|
||||
|
||||
// 下一帧继续
|
||||
lastUpdateTime = performance.now();
|
||||
requestAnimationFrame(processNextTask);
|
||||
};
|
||||
|
||||
// 开始重建任务
|
||||
processNextTask();
|
||||
}
|
||||
|
||||
dispose(){
|
||||
this.points = [];
|
||||
this.colors = [];
|
||||
this.colorGroups = new Map();
|
||||
this.progressCallback = null;
|
||||
}
|
||||
}
|
||||
|
||||
export {PointCloudReconstructor};
|
||||
@@ -0,0 +1 @@
|
||||
export {PointCloudReconstructor} from "./PointCloudReconstructor";
|
||||
Reference in New Issue
Block a user