grabbit
10fe6f95dd
Performance optimizations: - Frame accumulator: iterator-based loops, hot/cold path separation - Morphology: dirty rect tracking for Zhang-Suen thinning - Meteor detector: precomputed thresholds, sqrt elimination in centroid - Async frame prefetch with producer-consumer pattern - 2MB BufReader buffer to reduce syscalls FFmpeg improvements: - Hardware decoding with auto-fallback (VideoToolbox/VAAPI) - 3-second timeout probe to detect unsupported codecs - Automatic CPU fallback when hardware decode fails Detection visualization: - Draw fireball trajectories in red on maxpixel - Draw meteor trajectories in green - Save detection overlay as _detections.png Test results (1024 frames @ 1080p): - Detection processing: ~420ms/block (was ~987ms) - AV1 video: auto-fallback to CPU decode - H.264 video: uses hardware acceleration when available Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Meteor Fullstack - Distributed Monitoring Network
Overview
Meteor Fullstack 是一个覆盖边缘设备、云端 API、数据处理和 Web 前端的完整流星监测网络。采用 npm workspace 管理前后端，通过 Go、Rust 等多语言组件连接实时观测、事件验证和结果展示。
Key Features
- Vida Detection Algorithm: 基于 Vida et al. 2016 论文的科学级流星检测
- Advanced Memory Management: 零拷贝架构、分层帧池、自适应内存管理
- Real-time Processing: ~100ms/block 处理延迟
- Full Stack: 边缘客户端 → 后端 API → 前端展示完整链路
System Architecture
┌─────────────────┐ ┌─────────────────┐ ┌─────────────────┐
│ Edge Client │────▶│ Web Backend │────▶│ Frontend │
│ (Rust) │ │ (NestJS) │ │ (Next.js) │
└────────┬────────┘ └────────┬────────┘ └─────────────────┘
│ │
│ ▼
│ ┌─────────────────┐
│ │ Compute Service │
│ │ (Go) │
│ └────────┬────────┘
│ │
▼ ▼
┌─────────────────┐ ┌─────────────────┐
│ Camera/Video │ │ PostgreSQL │
│ Input │ │ + AWS (S3/SQS)│
└─────────────────┘ └─────────────────┘
Project Structure
| 目录 | 技术栈 | 说明 |
|---|---|---|
meteor-edge-client/ |
Rust, Tokio | 边缘客户端，摄像头采集 + Vida 检测 |
meteor-web-backend/ |
NestJS, TypeORM | Web API，认证、设备、事件管理 |
meteor-frontend/ |
Next.js 15, React 19 | 前端应用，仪表盘、图库、分析 |
meteor-compute-service/ |
Go 1.24 | 事件处理服务，SQS 消费者 |
infrastructure/ |
Terraform | AWS 基础设施 (RDS, S3, SQS) |
Quick Start
# 安装依赖
npm install
# 启动开发服务器 (前端 + 后端)
npm run dev
# 构建边缘客户端
cd meteor-edge-client && cargo build --release
# 测试 Vida 流星检测
./target/release/meteor-edge-client test-vida video.mp4
Vida Detection Algorithm
边缘客户端实现了 Vida 流星检测算法（与 RMS/克罗地亚流星网络兼容）：
- FTP 压缩: 256帧 → maxpixel, avepixel, stdpixel, maxframe
- 双路径检测:
- FireballDetector (K1=4): 检测明亮火球
- MeteorDetector (K1=1.5): 检测普通流星
- 形态学处理: 清理、桥接、闭合、Zhang-Suen 细化
- Hough 变换: 2D 直线检测 + 3D 点云分析
详见 CLAUDE.md 中的完整文档。
Development Commands
Full Stack
npm run dev # 启动前端和后端
npm run build # 构建所有服务
npm run test:fullstack # 运行完整测试套件
npm run lint # 代码检查
Edge Client (Rust)
cd meteor-edge-client
cargo build --release
cargo test
./target/release/meteor-edge-client run --camera device:0 # 使用摄像头
./target/release/meteor-edge-client run --camera file:v.mp4 # 使用视频文件
./target/release/meteor-edge-client test-vida video.mp4 # 测试检测
Backend (NestJS)
cd meteor-web-backend
npm run start:dev # 开发模式
npm run migrate:up # 运行迁移
npm run test # 单元测试
npm run test:integration # 集成测试
Frontend (Next.js)
cd meteor-frontend
npm run dev # 开发服务器
npm run build # 生产构建
npm run test # Jest 测试
npm run test:e2e # Playwright E2E
Testing
| 类型 | 命令 | 说明 |
|---|---|---|
| 单元测试 | npm run test |
Jest (前端+后端) |
| 集成测试 | npm run test:integration |
后端 API + 数据库 |
| E2E 测试 | npm run test:e2e |
Playwright |
| 边缘客户端 | cargo test |
Rust 单元测试 |
| 全栈测试 | npm run test:fullstack |
完整测试套件 |
测试环境需要 Docker: ./test-setup.sh
Documentation
CLAUDE.md- 完整技术文档和 Claude Code 指引TESTING.md- 详细测试矩阵docs/EDGE_DEVICE_REGISTRATION_COMPLETE.md- 设备注册规范infrastructure/README.md- 基础设施部署
License
MIT