30分钟搞定！YOLOv8+OpenClaw+腾讯云，手把手教你打造工业缺陷检测AI员工

# 配置飞书应用凭证
openclaw config set channels.feishu.appId "你的_APP_ID"
openclaw config set channels.feishu.appSecret "你的_APP_SECRET"
openclaw config set channels.feishu.enabled true
openclaw config set channels.feishu.connectionMode websocket
openclaw gateway restart

# 安装YOLO依赖
pip install ultralytics opencv-python

# 下载YOLOv8轻量模型
mkdir -p /root/yolo_models
cd /root/yolo_models
wget https://github.com/ultralytics/assets/releases/download/v8.2.0/yolov8n.pt

mkdir -p /root/openclaw-skills/defect-detection
cd /root/openclaw-skills/defect-detection

#!/usr/bin/env python3
"""
缺陷检测Skill - 基于YOLOv8
支持：单图检测、批量统计、异常告警
作者：AI小怪兽
"""

import os
import sys
import json
import cv2
import numpy as np
from datetime import datetime
from pathlib import Path
from ultralytics import YOLO

# 全局配置
MODEL_PATH = "/root/yolo_models/yolov8n.pt"
DEFECT_CLASSES = ["scratch", "dent", "color_error"]  # 划痕、凹坑、色差
STATS_FILE = "/tmp/defect_stats.json"

class DefectDetector:
    def __init__(self):
        """初始化YOLO模型"""
        self.model = YOLO(MODEL_PATH)
        self.stats = self.load_stats()
    
    def load_stats(self):
        """加载统计数据"""
        if os.path.exists(STATS_FILE):
            with open(STATS_FILE, 'r') as f:
                return json.load(f)
        return {
            "total": 0,
            "defects": 0,
            "types": {cls: 0 for cls in DEFECT_CLASSES},
            "daily": {}
        }
    
    def save_stats(self):
        """保存统计数据"""
        with open(STATS_FILE, 'w') as f:
            json.dump(self.stats, f, indent=2)
    
    def detect_single(self, image_path):
        """
        单张图片检测
        返回：检测结果和标注图像
        """
        # 执行推理
        results = self.model(image_path, conf=0.25)[0]  # 置信度阈值0.25 
        
        # 解析检测结果
        defects = []
        for box in results.boxes:
            cls_id = int(box.cls)
            conf = float(box.conf)
            x1, y1, x2, y2 = box.xyxy[0].tolist()
            
            # 此处需根据实际训练类别映射
            # 演示代码使用简化映射
            if cls_id < len(DEFECT_CLASSES):
                defect_type = DEFECT_CLASSES[cls_id]
            else:
                defect_type = f"unknown_{cls_id}"
            
            defects.append({
                "type": defect_type,
                "confidence": round(conf, 2),
                "bbox": [int(x1), int(y1), int(x2), int(y2)]
            })
        
        # 更新统计
        self.stats["total"] += 1
        if defects:
            self.stats["defects"] += 1
            for d in defects:
                if d["type"] in self.stats["types"]:
                    self.stats["types"][d["type"]] += 1
        
        # 记录日统计
        today = datetime.now().strftime("%Y-%m-%d")
        if today not in self.stats["daily"]:
            self.stats["daily"][today] = {"total": 0, "defects": 0}
        self.stats["daily"][today]["total"] += 1
        if defects:
            self.stats["daily"][today]["defects"] += 1
        
        self.save_stats()
        
        # 生成标注图像
        img = cv2.imread(image_path)
        for d in defects:
            x1, y1, x2, y2 = d["bbox"]
            # 画框
            cv2.rectangle(img, (x1, y1), (x2, y2), (0, 0, 255), 2)
            # 添加标签
            label = f"{d['type']}:{d['confidence']}"
            cv2.putText(img, label, (x1, y1-10), 
                       cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 2)
        
        # 保存标注结果
        output_path = f"/tmp/detected_{Path(image_path).name}"
        cv2.imwrite(output_path, img)
        
        return {
            "defects": defects,
            "count": len(defects),
            "image_path": output_path
        }
    
    def get_today_report(self):
        """生成今日统计报表"""
        today = datetime.now().strftime("%Y-%m-%d")
        daily = self.stats["daily"].get(today, {"total": 0, "defects": 0})
        
        if daily["total"] == 0:
            return "今日暂无检测数据"
        
        report = f"""📊 今日缺陷统计报告 ({today})
总检测数：{daily['total']}件
缺陷数：{daily['defects']}件
不良率：{(daily['defects']/daily['total']*100):.2f}%

缺陷类型分布："""
        
        for defect_type, count in self.stats["types"].items():
            report += f"\n- {defect_type}：{count}次"
        
        # 添加建议
        if self.stats["defects"] > 0:
            max_type = max(self.stats["types"].items(), key=lambda x: x[1])
            report += f"\n\n建议重点关注{max_type[0]}问题，占比{(max_type[1]/self.stats['defects']*100):.1f}%"
        
        return report
    
    def check_alert(self):
        """检查是否需要告警（缺陷率>10%）"""
        today = datetime.now().strftime("%Y-%m-%d")
        daily = self.stats["daily"].get(today, {"total": 0, "defects": 0})
        
        if daily["total"] > 0:
            defect_rate = daily["defects"] / daily["total"]
            if defect_rate > 0.1:  # 10%告警阈值
                return {
                    "alert": True,
                    "rate": defect_rate,
                    "message": f"⚠️ 告警：今日缺陷率{defect_rate:.2%}，超过10%阈值"
                }
        return {"alert": False}

def main():
    """OpenClaw Skill入口函数"""
    detector = DefectDetector()
    
    # 解析输入参数（由OpenClaw传入）
    if len(sys.argv) > 1:
        command = sys.argv[1]
        
        if command == "detect" and len(sys.argv) > 2:
            # 检测图片
            image_path = sys.argv[2]
            result = detector.detect_single(image_path)
            print(f"🔍 缺陷检测结果")
            if result["defects"]:
                for d in result["defects"]:
                    print(f"- {d['type']} (置信度{d['confidence']})，位置：{d['bbox']}")
                print(f"⚠️ 发现缺陷，建议复检！")
            else:
                print(f"✅ 未检测到缺陷")
            
        elif command == "report":
            # 统计报表
            print(detector.get_today_report())
            
        elif command == "alert":
            # 告警检查
            alert = detector.check_alert()
            if alert["alert"]:
                print(alert["message"])
            else:
                print(f"✅ 当前缺陷率正常")
    
    else:
        print("Usage: defect_detector.py [detect|report|alert] [image_path]")

if __name__ == "__main__":
    main()

检测这张图片里的缺陷：https://example.com/product-001.jpg

🔍 缺陷检测结果
检测到2处缺陷：
- 划痕 (置信度0.92)，位置：[120,45,300,180]
- 凹坑 (置信度0.76)，位置：[450,60,620,190]
⚠️ 发现缺陷，建议复检！

统计今天的产品缺陷情况

📊 今日缺陷统计报告 (2026-03-13)
总检测数：1245件
缺陷数：87件
不良率：6.99%

缺陷类型分布：
- 划痕：45次
- 凹坑：28次
- 色差：14次

建议重点关注划痕问题，占比51.7%

# 每小时执行一次检测，假设最新图片路径为 /data/latest.jpg
0 * * * * cd /root/openclaw-skills/defect-detection && python3 defect_detector.py detect /data/latest.jpg >> /tmp/detect.log 2>&1
# 每小时检查一次告警
5 * * * * cd /root/openclaw-skills/defect-detection && python3 defect_detector.py alert | grep "告警" && /root/send_alert.sh

#!/bin/bash
# send_feishu_alert.sh
# 用法：./send_feishu_alert.sh "告警消息" /path/to/image.jpg

MESSAGE=$1
IMAGE_PATH=$2

# 从OpenClaw配置中读取App Secret
APP_SECRET=$(python3 -c "import json; c=json.load(open('/root/.openclaw/openclaw.json')); print(c['channels']['feishu']['appSecret'])")

# 获取tenant_access_token
TOKEN=$(curl -s -X POST 'https://open.feishu.cn/open-apis/auth/v3/tenant_access_token/internal' \
  -H 'Content-Type: application/json' \
  -d "{\"app_id\":\"你的APP_ID\",\"app_secret\":\"$APP_SECRET\"}" \
  | python3 -c "import json,sys; print(json.load(sys.stdin)['tenant_access_token'])")

# 上传图片获取image_key
IMAGE_KEY=$(curl -s -X POST 'https://open.feishu.cn/open-apis/im/v1/images' \
  -H "Authorization: Bearer $TOKEN" \
  -F "image_type=message" \
  -F "image=@$IMAGE_PATH" \
  | python3 -c "import json,sys; print(json.load(sys.stdin)['data']['image_key'])")

# 发送文本消息（可扩展为同时发送图片）
curl -s -X POST 'https://open.feishu.cn/open-apis/im/v1/messages?receive_id_type=open_id' \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d "{\"receive_id\":\"管理员_OPEN_ID\",\"msg_type\":\"text\",\"content\":\"{\\\"text\\\":\\\"$MESSAGE\\\"}\"}"