基于Python与OpenCV的学校食堂后厨穿戴识别系统开发：从算法逻辑到工程落地

pip install ultralytics opencv-python

import cv2
from ultralytics import YOLO

# 加载自定义训练的后厨检测模型
model = YOLO('kitchen_yolov8n.pt') 

def detect_frame(frame):
    # 进行推理，conf设置置信度阈值，实验室实测0.45能平衡误报与漏报
    results = model(frame, conf=0.45, iou=0.5)
    return results

def check_compliance(results, frame):
    boxes = results[0].boxes
    detections = []
    
    # 提取所有检测到的对象
    for box in boxes:
        cls = int(box.cls[0])
        conf = float(box.conf[0])
        x1, y1, x2, y2 = map(int, box.xyxy[0])
        label = model.names[cls]
        detections.append({'class': label, 'box': (x1, y1, x2, y2), 'conf': conf})

    # 逻辑判定：人员是否佩戴帽子
    persons = [d for d in detections if d['class'] == 'person']
    hats = [d for d in detections if d['class'] == 'chef_hat']
    
    for person in persons:
        px1, py1, px2, py2 = person['box']
        # 定义头部区域（通常为人体框的上半部分）
        head_region = (px1, py1, px2, py1 + int((py2 - py1) * 0.4))
        
        is_wearing_hat = False
        for hat in hats:
            hx1, hy1, hx2, hy2 = hat['box']
            # 简单的中心点包含判定
            hat_cx = (hx1 + hx2) / 2
            hat_cy = (hy1 + hy2) / 2
            if (head_region[0] < hat_cx < head_region[2]) and (head_region[1] < hat_cy < head_region[3]):
                is_wearing_hat = True
                break
                
        if not is_wearing_hat:
            # 触发报警逻辑
            cv2.rectangle(frame, (px1, py1), (px2, py2), (0, 0, 255), 2)
            cv2.putText(frame, "No Hat Detected", (px1, py1-10), cv2.FONT_HERSHEY_SIMPLEX, 0.9, (0,0,255), 2)
            
    return frame

# 鼠患识别逻辑片段
def detect_rodent(results, frame):
    for box in results[0].boxes:
        cls = int(box.cls[0])
        if model.names[cls] == 'mouse':
            x1, y1, x2, y2 = map(int, box.xyxy[0])
            # 绘制红色警示框
            cv2.rectangle(frame, (x1, y1), (x2, y2), (0, 0, 255), 3)
            cv2.putText(frame, "Rodent Alert!", (x1, y1-10), cv2.FONT_HERSHEY_SIMPLEX, 1, (0,0,255), 2)
            # 此处可添加截图上传云端或触发蜂鸣器代码
    return frame

# 简单的区域入侵/溢满判定逻辑
import cv2

# 定义垃圾桶区域 (示例坐标)
trash_bin_roi = np.array(, np.int32)

def check_trash_overflow(results):
    for box in results[0].boxes:
        if model.names[int(box.cls[0])] == 'garbage':
            x1, y1, x2, y2 = map(int, box.xyxy[0])
            # 计算垃圾框中心点
            cx, cy = (x1 + x2) // 2, (y1 + y2) // 2
            # 判断中心点是否在ROI内
            if cv2.pointPolygonTest(trash_bin_roi, (cx, cy), False) >= 0:
                return True
    return False