wasm-spatial-core

Browser-side point cloud & terrain preprocessing — ingest, edit geometry, emit 3D Tiles / glTF. Zero server, zero upload. CRS: WGS84, Web Mercator, UTM, China offsets (not arbitrary EPSG).

What you get from npm install

The published package is a prebuilt WASM binary (point-cloud + geotiff):

| Included | Requires custom build | |----------|----------------------| | LAS, PLY, OBJ, PCD | LAZ / COPC (laz-support) | | Octree + 3D Tiles (pnts) | E57 (e57-support) | | GeoTIFF → quantized-mesh | Terrain deformation (terrain-edit) | | Coordinates, GeoJSON, MVT | Spatial IR + GLB ingest (mesh-ingest) | | | Mesh QEM / clip / OBB split (mesh-edit, needs mesh-ingest) | | | WebGPU compute kernels (webgpu) |

Formats: 10+ in the default npm build · 15+ with optional format features (LAZ/COPC, E57, GLB ingest).

Check at runtime: supportsLaz(), supportsGeotiff(), lazStatus(), supportsWebGpu(), supportsMeshEdit().

Full matrix: Feature flags in the repo README.

🚀 Quick Start

npm install wasm-spatial-core

import { loadSpatialCore } from "wasm-spatial-core";

const core = await loadSpatialCore();

// Coordinate conversion
const wgs84 = new Float64Array([116.404, 39.915, 121.474, 31.230]);
const gcj02 = core.batchWgs84ToGcj02(wgs84);

// GeoJSON parsing
const coords = core.parseGeoJsonCoords(geojsonStr);

☁️ Point Cloud → 3D Tiles

Drag a LAS file into the browser, get a Cesium-ready 3D Tiles tileset — no server needed.

LAZ / COPC: not included in the default npm binary. Build with --features laz-support or check supportsLaz() before calling parsePointCloudAuto on compressed files.

import { loadSpatialCore } from "wasm-spatial-core";

const core = await loadSpatialCore();

// 1. Parse point cloud (LAS — default npm build)
const lasBuffer = await fetch("scan.las").then(r => r.arrayBuffer());
const points = core.parseLasPoints(new Uint8Array(lasBuffer));

// 2. Decimate if needed (voxel grid → uniform density)
const decimated = core.decimateVoxelGrid(
  points.positions(),
  points.colors(),
  1.0  // 1-meter grid
);

// 3. Build spatial index (octree)
const octree = core.buildOctree(decimated.positions, 50000, 10);

// 4. Generate 3D Tiles tileset
const tileset = core.generateTileset(
  decimated.positions,
  50000,           // max points per tile
  10,              // max tree depth
  decimated.colors
);

// 5. Use with Cesium
console.log(tileset.tilesetJson());       // tileset.json
console.log(tileset.tileCount());        // number of .pnts tiles
const tile0 = tileset.tile(0);           // Uint8Array of first tile
const bounds0 = tileset.tileBounds(0);  // Float64Array [minX..maxZ]

// 6. LOD: get visible tiles for current camera
const fov = Math.PI / 3; // 60° vertical FOV
const visible = core.getVisibleTiles(
  decimated.positions,
  camera.x, camera.y, camera.z,
  fov, 1920, 1080
);
// → Uint32Array of node indices to load

📋 API Reference

Coordinate Projection

| Function | Description | |----------|-------------| | batchWgs84ToGcj02(coords) | WGS-84 → GCJ-02 | | batchGcj02ToWgs84(coords) | GCJ-02 → WGS-84 | | batchWgs84ToBd09(coords) | WGS-84 → BD-09 | | batchBd09ToWgs84(coords) | BD-09 → WGS-84 | | batchWgs84ToMercator(coords) | WGS-84 → EPSG:3857 | | batchMercatorToWgs84(coords) | EPSG:3857 → WGS-84 | | batchWgs84ToGcj02Mercator(coords) | WGS-84 → GCJ-02 → Mercator | | *InPlace variants | Zero-copy in-place for all above | | wgs84ToUtm(lng, lat) | WGS-84 → UTM | | utmToWgs84(zone, e, n, isN) | UTM → WGS-84 |

GeoJSON

| Function | Description | |----------|-------------| | parseGeoJsonCoords(input) | Extract coordinates → Float64Array | | countGeoJsonFeatures(input) | Count features | | parseGeoJsonStream(input, size, cb) | Chunked parser (full JSON parse, batched coord output) | | parseGeoJsonLazy(input) | One-feature-at-a-time iterator (input string required) | | geoJsonFromCoords(coords, type) | Generate GeoJSON | | filterGeoJsonByProperty(input, k, v) | Filter features | | filterGeoJsonByBBox(input, ...) | Spatial filter |

Point Cloud (LAS — default npm)

| Function | Description | |----------|-------------| | parseLasHeader(bytes) | Parse LAS header | | parseLasPoints(bytes) | Parse all points | | parseLasPointsWithProgress(bytes, cb) | Parse with progress | | parsePointCloudAuto(bytes) | Auto-detect format (LAZ/COPC only with laz-support build) | | decimateVoxelGrid(pos, col, size) | Voxel decimation | | decimateRandom(pos, col, count) | Random sampling | | colorizeByHeight(pos, minZ, maxZ) | Height-based coloring | | estimateNormals(pos, k) | kNN normal estimation |

Point Cloud — LAZ / COPC (custom build only)

| Function | Description | |----------|-------------| | new PointCloudStreamer(url) | Create streamer | | .parseHeader() | Parse header | | .readPoints(offset, count) | Read points by offset | | .readRegion(min, max) | Spatial range read | | computeRegionByteRange(...) | Compute byte range | | supportsLaz() | false in default npm; true with laz-support | | lazStatus() | Runtime LAZ capability string |

Octree

| Function | Description | |----------|-------------| | buildOctree(positions, maxPts?, maxDepth?) | Build spatial octree | | Octree | Octree class | | .nodeCount() / .depth() / .totalPoints() | Tree stats | | .rootBounds() / .nodeBounds(i) | Bounding boxes | | .leafCount() | Number of leaf nodes | | octreeMemoryUsage(n, internal, pts) | Memory estimate |

3D Tiles (pnts)

| Function | Description | |----------|-------------| | encodePntsTile(pos, cx, cy, cz, colors?) | Encode pnts binary | | generateTileset(pos, maxPts?, maxDepth?, colors?) | Full tileset | | TilesetResult | Tileset class | | .tilesetJson() / .tileCount() / .tile(i) | Access tiles | | .tileBounds(i) / .tileUri(i) | Tile metadata |

LOD

| Function | Description | |----------|-------------| | computeScreenSpaceError(geoErr, dist, fov, h) | SSE in pixels | | getVisibleTiles(pos, cam, fov, w, h, ...) | Visible tile indices |

Spatial Analysis

| Function | Description | |----------|-------------| | haversineDistance(lng1, lat1, lng2, lat2) | Great-circle distance | | bearing() / destination() / midpoint() | Geodesic | | bufferPoint() / bufferLineString() | Buffer geometry | | polygonArea() / polylineLength() | Measurements | | simplifyDouglasPeucker(coords, tol) | Line simplification | | polygonIntersection() / polygonUnion() | Boolean ops |

Cesium Integration

| Function | Description | |----------|-------------| | batchWgs84ToCartesian3(coords) | WGS84 → ECEF | | generateCesiumGeometry(geojson, h) | Triangulate → mesh | | generate3DTile(geojson, h) | Build b3dm tile |

Memory

| Function | Description | |----------|-------------| | memoryInfo() | WASM memory → MemoryInfo | | getAllocatedBytes() | Peak allocation | | setInputSizeLimit(bytes) | Set max input size |

🖥️ Node.js Batch Processing

Server-side pipelines via the nodejs WASM target — no browser or COOP/COEP headers required.

import { loadSpatialCoreNode, batchPointCloudToTileset } from "wasm-spatial-core/node";
import { readFileSync, writeFileSync, mkdirSync } from "node:fs";
import { join } from "node:path";

const core = await loadSpatialCoreNode();
const las = readFileSync("scan.las");
const result = await batchPointCloudToTileset(core, las);

mkdirSync("output/tiles", { recursive: true });
writeFileSync("output/tileset.json", result.tilesetJson);
result.tiles.forEach((data, i) => {
  writeFileSync(join("output/tiles", result.tileUris[i]), data);
});

Build the Node.js WASM package: npm run build:wasm:node (outputs to npm/pkg-node/).

🌐 Live Demo

https://reed-soul.github.io/wasm-spatial-core/examples/index.html

📄 License

MIT © 2026 Zhiqi Weilai