从OpenCV到YOLOv5，手把手整合完整视觉项目完整代码

importcv2importtorchimportnumpyasnpimporttime# -------------------------- 请修改这3个参数（必改）--------------------------model_path="best.pt"# YOLOv5模型路径（与代码同一目录填"best.pt"）video_path=0# 0=默认摄像头，可替换为视频路径（如："test.mp4"）save_video=True# 是否保存检测结果视频（True=保存，False=不保存）# --------------------------------------------------------------------------# -------------------------- 可选参数（按需修改）--------------------------conf_threshold=0.4# 置信度阈值（3.31优化技巧）track_distance=50# 追踪距离阈值（避免重复计数）alarm_line=(200,350,1080,350)# 越线报警线（x1,y1,x2,y2）# --------------------------------------------------------------------------# 1. 初始化所有模块（整合本周所有核心功能）# 1.1 加载YOLOv5模型（优化后，兼容所有版本）model=torch.hub.load('ultralytics/yolov5','custom',path=model_path,force_reload=True)model.conf=conf_threshold# 应用优化技巧，过滤无效框class_names=['cat','dog']# 类别的，与数据集一致colors=[(255,0,0),(0,0,255)]# 颜色（猫=蓝，狗=红）# 1.2 初始化视频读取与保存（4.2进阶功能）cap=cv2.VideoCapture(video_path)cap.set(cv2.CAP_PROP_FRAME_WIDTH,1280)cap.set(cv2.CAP_PROP_FRAME_HEIGHT,720)# 初始化视频写入器fourcc=cv2.VideoWriter_fourcc(*'mp4v')fps=cap.get(cv2.CAP_PROP_FPS)frame_width=int(cap.get(cv2.CAP_PROP_FRAME_WIDTH))frame_height=int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT))out=Noneifsave_video:out=cv2.VideoWriter('complete_project_result.mp4',fourcc,fps,(frame_width,frame_height))# 1.3 初始化计数、追踪、报警相关变量（4.1+4.2功能整合）total_cat=0# 总猫数total_dog=0# 总狗数current_cat=0# 当前帧猫数current_dog=0# 当前帧狗数tracks={}# 目标追踪字典track_id=0# 目标唯一IDcrossed_ids=set()# 已越线目标ID（避免重复报警）alarm_flag=False# 报警标志# 2. 项目核心循环（整合所有功能，一步到位）whilecap.isOpened():ret,frame=cap.read()ifnotret:break# 视频读取完毕，退出循环# 重置当前帧计数current_cat=0current_dog=0# 3. 模块1：OpenCV图像预处理（3.27轮廓检测+画面优化）# 灰度处理+轮廓提取（可选，可按需开启）gray=cv2.cvtColor(frame,cv2.COLOR_BGR2GRAY)blur=cv2.GaussianBlur(gray,(5,5),0)# 模糊去噪ret,thresh=cv2.threshold(blur,127,255,cv2.THRESH_BINARY)contours,_=cv2.findContours(thresh,cv2.RETR_EXTERNAL,cv2.CHAIN_APPROX_SIMPLE)# 绘制轮廓（可选，可视化预处理效果）cv2.drawContours(frame,contours,-1,(0,255,0),1)# 4. 模块2：YOLOv5目标检测（3.30+3.31优化功能）results=model(frame)detections=results.pandas().xyxy[0].values# 提取检测结果# 5. 模块3：目标追踪与计数（4.1基础功能）current_tracks=[]fordetindetections:xmin,ymin,xmax,ymax,conf,cls,name=det center_x=int((xmin+xmax)/2)center_y=int((ymin+ymax)/2)cls=int(cls)# 当前帧计数ifcls==0:current_cat+=1else:current_dog+=1# 目标追踪逻辑（优化后，更稳定）matched=Falsefortrack_id_exist,(track_center_x,track_center_y,track_cls)intracks.items():distance=np.sqrt((center_x-track_center_x)**2+(center_y-track_center_y)**2)ifdistance<track_distanceandtrack_cls==cls:tracks[track_id_exist]=(center_x,center_y,cls)current_tracks.append(track_id_exist)matched=True# 模块4：越线报警（4.2进阶功能）line_y=alarm_line[1]iftrack_center_y>line_yandtrack_id_existnotincrossed_ids:crossed_ids.add(track_id_exist)alarm_flag=True# 报警可视化+控制台提示cv2.putText(frame,f"ALARM:{name}crossed line!",(500,40),cv2.FONT_HERSHEY_SIMPLEX,1,(0,0,255),3)print(f"【报警】{name}越过报警线，时间：{time.strftime('%H:%M:%S')}")break# 新目标计数ifnotmatched:tracks[track_id]=(center_x,center_y,cls)current_tracks.append(track_id)ifcls==0:total_cat+=1else:total_dog+=1track_id+=1# 删除消失的目标轨迹tracks={k:vfork,vintracks.items()ifkincurrent_tracks}# 6. 模块5：可视化展示（整合所有可视化元素）# 绘制报警线cv2.line(frame,(alarm_line[0],alarm_line[1]),(alarm_line[2],alarm_line[3]),(0,255,0),2)cv2.putText(frame,"Alarm Line",(alarm_line[0],alarm_line[1]-10),cv2.FONT_HERSHEY_SIMPLEX,0.5,(0,255,0),2)# 绘制检测框、类别、置信度fordetindetections:xmin,ymin,xmax,ymax,conf,cls,name=det xmin,ymin,xmax,ymax=int(xmin),int(ymin),int(xmax),int(ymax)cls=int(cls)cv2.rectangle(frame,(xmin,ymin),(xmax,ymax),colors[cls],2)label=f"{name}{conf:.2f}"cv2.putText(frame,label,(xmin,ymin-10),cv2.FONT_HERSHEY_SIMPLEX,0.5,colors[cls],2)# 绘制目标轨迹fortrack_id_exist,(center_x,center_y,cls)intracks.items():cv2.circle(frame,(center_x,center_y),5,colors[cls],-1)iftrack_id_exist>0:prev_center=tracks.get(track_id_exist-1,None)ifprev_centerandprev_center[2]==cls:cv2.line(frame,(prev_center[0],prev_center[1]),(center_x,center_y),colors[cls],2)# 绘制计数信息（总计数+当前帧计数）cv2.putText(frame,f"Total Cat:{total_cat}",(20,40),cv2.FONT_HERSHEY_SIMPLEX,1,(255,0,0),2)cv2.putText(frame,f"Total Dog:{total_dog}",(20,80),cv2.FONT_HERSHEY_SIMPLEX,1,(0,0,255),2)cv2.putText(frame,f"Current Cat:{current_cat}",(1000,40),cv2.FONT_HERSHEY_SIMPLEX,1,(255,0,0),2)cv2.putText(frame,f"Current Dog:{current_dog}",(1000,80),cv2.FONT_HERSHEY_SIMPLEX,1,(0,0,255),2)# 绘制报警提示（持续3秒）ifalarm_flag:cv2.putText(frame,"ALARM! Target Crossed Line",(400,700),cv2.FONT_HERSHEY_SIMPLEX,1.5,(0,0,255),3)alarm_flag=False# 保存检测结果视频（模块4功能）ifsave_videoandoutisnotNone:out.write(frame)# 显示完整项目画面cv2.imshow("Complete Vision Project (OpenCV + YOLOv5)",frame)# 按「q」键退出ifcv2.waitKey(1)&0xFF==ord('q'):break# 释放所有资源（避免内存泄漏）cap.release()ifsave_videoandoutisnotNone:out.release()cv2.destroyAllWindows()# 打印最终项目统计结果（便于汇报、整理数据）print("="*60)print("完整视觉项目 - 最终统计结果")print("="*60)print(f"总检测到猫：{total_cat}只")print(f"总检测到狗：{total_dog}只")print(f"越线目标数量：{len(crossed_ids)}个")print(f"检测结果视频：{'已保存（complete_project_result.mp4）'ifsave_videoelse'未保存'}")print(f"项目运行状态：成功完成")print("="*60)