【图像算法 - 13】基于 YOLO12 与 OpenCV 的实时目标点击跟踪系统（系统介绍 + 源码详细）

基于 YOLO12 与 OpenCV 的实时点击目标跟踪系统

在计算机视觉领域，目标检测与跟踪是两个核心任务。本文将介绍一个结合 YOLO 目标检测模型与 OpenCV 跟踪算法的实时目标跟踪系统，该系统允许用户通过鼠标交互选择特定目标进行持续跟踪，支持多种跟踪算法切换，适用于视频监控、行为分析等场景。

【图像算法 - 13】基于 YOLO12 与 OpenCV 的实时目标点击跟踪系统

系统功能概述

该系统主要实现以下功能：

使用 YOLO 模型对视频帧进行目标检测，识别出画面中的各类物体
支持用户通过左键点击选择特定目标进行跟踪
提供 MOSSE、CSRT、KCF 三种经典跟踪算法供选择
实时显示跟踪状态、目标类别及 FPS 等信息
支持右键点击停止跟踪，回到目标检测模式

点击跟踪

技术原理

系统采用 “检测 + 跟踪” 的混合架构：

首先利用 YOLO 模型进行目标检测，获取画面中所有目标的边界框和类别信息
当用户选择特定目标后，启动选定的跟踪算法对该目标进行持续跟踪
跟踪过程中实时更新目标位置，若跟踪失败则提示 “Lost”
整个过程通过可视化界面展示，支持用户交互操作

这种架构结合了 YOLO 检测精度高和传统跟踪算法速度快的优点，在保证一定精度的同时兼顾了实时性。

代码解析

核心依赖库

import cv2          # 用于视频处理和跟踪算法
import numpy as np  # 用于数值计算
import argparse     # 用于命令行参数解析
from ultralytics import YOLO  # 用于YOLO目标检测

全局变量与交互设计

定义全局变量存储跟踪状态和用户交互信息：

selected_box = None  # 选中的目标边界框
tracking = False     # 跟踪状态标志
target_class = None  # 目标类别
click_x, click_y = -1, -1  # 鼠标点击坐标
mouse_clicked = False      # 鼠标点击标志
debug = False              # 调试模式标志

鼠标回调函数处理用户交互：

def mouse_callback(event, x, y, flags, param):global click_x, click_y, mouse_clicked, trackingif event == cv2.EVENT_LBUTTONDOWN:  # 左键点击选择目标click_x, click_y = x, ymouse_clicked = Trueif debug:print(f"Left click at: ({x}, {y})")elif event == cv2.EVENT_RBUTTONDOWN:  # 右键点击停止跟踪tracking = Falseif debug:print("Tracking stopped")

跟踪器创建

针对不同的跟踪算法，创建对应的跟踪器实例（注意 OpenCV 新版本中跟踪器位于 legacy 模块）：

def create_tracker(tracker_type):"""使用cv2.legacy模块创建跟踪器，兼容新版OpenCV"""try:if tracker_type == "mosse":return cv2.legacy.TrackerMOSSE_create()elif tracker_type == "csrt":return cv2.legacy.TrackerCSRT_create()elif tracker_type == "kcf":return cv2.legacy.TrackerKCF_create()else:print(f"Unsupported tracker: {tracker_type}, using MOSSE")return cv2.legacy.TrackerMOSSE_create()except AttributeError as e:print(f"Failed to create tracker: {e}")print("Please check OpenCV installation (must include opencv-contrib-python)")return None

OpenCV 跟踪demo源码

#!/usr/bin/env python
'''
Tracker demoFor usage download models by following links
For GOTURN:goturn.prototxt and goturn.caffemodel: https://github.com/opencv/opencv_extra/tree/c4219d5eb3105ed8e634278fad312a1a8d2c182d/testdata/tracking
For DaSiamRPN:network:     https://www.dropbox.com/s/rr1lk9355vzolqv/dasiamrpn_model.onnx?dl=0kernel_r1:   https://www.dropbox.com/s/999cqx5zrfi7w4p/dasiamrpn_kernel_r1.onnx?dl=0kernel_cls1: https://www.dropbox.com/s/qvmtszx5h339a0w/dasiamrpn_kernel_cls1.onnx?dl=0
For NanoTrack:nanotrack_backbone: https://github.com/HonglinChu/SiamTrackers/blob/master/NanoTrack/models/nanotrackv2/nanotrack_backbone_sim.onnxnanotrack_headneck: https://github.com/HonglinChu/SiamTrackers/blob/master/NanoTrack/models/nanotrackv2/nanotrack_head_sim.onnxUSAGE:tracker.py [-h] [--input INPUT_VIDEO][--tracker_algo TRACKER_ALGO (mil, goturn, dasiamrpn, nanotrack, vittrack)][--goturn GOTURN_PROTOTXT][--goturn_model GOTURN_MODEL][--dasiamrpn_net DASIAMRPN_NET][--dasiamrpn_kernel_r1 DASIAMRPN_KERNEL_R1][--dasiamrpn_kernel_cls1 DASIAMRPN_KERNEL_CLS1][--nanotrack_backbone NANOTRACK_BACKBONE][--nanotrack_headneck NANOTRACK_TARGET][--vittrack_net VITTRACK_MODEL][--vittrack_net VITTRACK_MODEL][--tracking_score_threshold TRACKING SCORE THRESHOLD FOR ONLY VITTRACK][--backend CHOOSE ONE OF COMPUTATION BACKEND][--target CHOOSE ONE OF COMPUTATION TARGET]
'''# Python 2/3 compatibility
from __future__ import print_functionimport sysimport numpy as np
import cv2 as cv
import argparsefrom video import create_capture, presetsbackends = (cv.dnn.DNN_BACKEND_DEFAULT, cv.dnn.DNN_BACKEND_HALIDE, cv.dnn.DNN_BACKEND_INFERENCE_ENGINE, cv.dnn.DNN_BACKEND_OPENCV,cv.dnn.DNN_BACKEND_VKCOM, cv.dnn.DNN_BACKEND_CUDA)
targets = (cv.dnn.DNN_TARGET_CPU, cv.dnn.DNN_TARGET_OPENCL, cv.dnn.DNN_TARGET_OPENCL_FP16, cv.dnn.DNN_TARGET_MYRIAD,cv.dnn.DNN_TARGET_VULKAN, cv.dnn.DNN_TARGET_CUDA, cv.dnn.DNN_TARGET_CUDA_FP16)class App(object):def __init__(self, args):self.args = argsself.trackerAlgorithm = args.tracker_algoself.tracker = self.createTracker()def createTracker(self):if self.trackerAlgorithm == 'mil':tracker = cv.TrackerMIL_create()elif self.trackerAlgorithm == 'goturn':params = cv.TrackerGOTURN_Params()params.modelTxt = self.args.goturnparams.modelBin = self.args.goturn_modeltracker = cv.TrackerGOTURN_create(params)elif self.trackerAlgorithm == 'dasiamrpn':params = cv.TrackerDaSiamRPN_Params()params.model = self.args.dasiamrpn_netparams.kernel_cls1 = self.args.dasiamrpn_kernel_cls1params.kernel_r1 = self.args.dasiamrpn_kernel_r1params.backend = args.backendparams.target = args.targettracker = cv.TrackerDaSiamRPN_create(params)elif self.trackerAlgorithm == 'nanotrack':params = cv.TrackerNano_Params()params.backbone = args.nanotrack_backboneparams.neckhead = args.nanotrack_headneckparams.backend = args.backendparams.target = args.targettracker = cv.TrackerNano_create(params)elif self.trackerAlgorithm == 'vittrack':params = cv.TrackerVit_Params()params.net = args.vittrack_netparams.tracking_score_threshold = args.tracking_score_thresholdparams.backend = args.backendparams.target = args.targettracker = cv.TrackerVit_create(params)else:sys.exit("Tracker {} is not recognized. Please use one of three available: mil, goturn, dasiamrpn, nanotrack.".format(self.trackerAlgorithm))return trackerdef initializeTracker(self, image):while True:print('==> Select object ROI for tracker ...')bbox = cv.selectROI('tracking', image)print('ROI: {}'.format(bbox))if bbox[2] <= 0 or bbox[3] <= 0:sys.exit("ROI selection cancelled. Exiting...")try:self.tracker.init(image, bbox)except Exception as e:print('Unable to initialize tracker with requested bounding box. Is there any object?')print(e)print('Try again ...')continuereturndef run(self):videoPath = self.args.inputprint('Using video: {}'.format(videoPath))camera = create_capture(cv.samples.findFileOrKeep(videoPath), presets['cube'])if not camera.isOpened():sys.exit("Can't open video stream: {}".format(videoPath))ok, image = camera.read()if not ok:sys.exit("Can't read first frame")assert image is not Nonecv.namedWindow('tracking')self.initializeTracker(image)print("==> Tracking is started. Press 'SPACE' to re-initialize tracker or 'ESC' for exit...")while camera.isOpened():ok, image = camera.read()if not ok:print("Can't read frame")breakok, newbox = self.tracker.update(image)#print(ok, newbox)if ok:cv.rectangle(image, newbox, (200,0,0))cv.imshow("tracking", image)k = cv.waitKey(1)if k == 32:  # SPACEself.initializeTracker(image)if k == 27:  # ESCbreakprint('Done')if __name__ == '__main__':print(__doc__)parser = argparse.ArgumentParser(description="Run tracker")parser.add_argument("--input", type=str, default="vtest.avi", help="Path to video source")parser.add_argument("--tracker_algo", type=str, default="nanotrack", help="One of available tracking algorithms: mil, goturn, dasiamrpn, nanotrack, vittrack")parser.add_argument("--goturn", type=str, default="goturn.prototxt", help="Path to GOTURN architecture")parser.add_argument("--goturn_model", type=str, default="goturn.caffemodel", help="Path to GOTERN model")parser.add_argument("--dasiamrpn_net", type=str, default="dasiamrpn_model.onnx", help="Path to onnx model of DaSiamRPN net")parser.add_argument("--dasiamrpn_kernel_r1", type=str, default="dasiamrpn_kernel_r1.onnx", help="Path to onnx model of DaSiamRPN kernel_r1")parser.add_argument("--dasiamrpn_kernel_cls1", type=str, default="dasiamrpn_kernel_cls1.onnx", help="Path to onnx model of DaSiamRPN kernel_cls1")parser.add_argument("--nanotrack_backbone", type=str, default="nanotrack_backbone_sim.onnx", help="Path to onnx model of NanoTrack backBone")parser.add_argument("--nanotrack_headneck", type=str, default="nanotrack_head_sim.onnx", help="Path to onnx model of NanoTrack headNeck")parser.add_argument("--vittrack_net", type=str, default="vitTracker.onnx", help="Path to onnx model of  vittrack")parser.add_argument('--tracking_score_threshold', type=float,  help="Tracking score threshold. If a bbox of score >= 0.3, it is considered as found ")parser.add_argument('--backend', choices=backends, default=cv.dnn.DNN_BACKEND_DEFAULT, type=int,help="Choose one of computation backends: ""%d: automatically (by default), ""%d: Halide language (http://halide-lang.org/), ""%d: Intel's Deep Learning Inference Engine (https://software.intel.com/openvino-toolkit), ""%d: OpenCV implementation, ""%d: VKCOM, ""%d: CUDA"% backends)parser.add_argument("--target", choices=targets, default=cv.dnn.DNN_TARGET_CPU, type=int,help="Choose one of target computation devices: "'%d: CPU target (by default), ''%d: OpenCL, ''%d: OpenCL fp16 (half-float precision), ''%d: VPU, ''%d: VULKAN, ''%d: CUDA, ''%d: CUDA fp16 (half-float preprocess)'% targets)args = parser.parse_args()App(args).run()cv.destroyAllWindows()

目标检测与处理

使用 YOLO 模型进行目标检测，并将结果转换为便于处理的格式：

def detect_objects(frame, model, confidence):results = model(frame, conf=confidence)boxes = []  # 存储边界框 (x, y, w, h)class_names = []  # 存储目标类别for result in results:for box in result.boxes:x1, y1, x2, y2 = box.xyxy[0].tolist()boxes.append((int(x1), int(y1), int(x2 - x1), int(y2 - y1)))class_names.append(model.names[int(box.cls[0])])return boxes, class_names

判断鼠标点击是否在某个目标框内：

def point_in_box(x, y, box):bx, by, bw, bh = boxreturn bx <= x <= bx + bw and by <= y <= by + bh

主程序逻辑

主函数协调整个系统的工作流程：

def main():global selected_box, tracking, target_class, mouse_clicked, debugargs = parse_args()debug = args.debug# 验证跟踪器是否可用test_tracker = create_tracker(args.tracker)if test_tracker is None:returndel test_tracker# 加载YOLO模型try:model = YOLO(args.model)except Exception as e:print(f"Failed to load YOLO model: {e}")return# 打开视频源cap = cv2.VideoCapture(args.video)if not cap.isOpened():print(f"Cannot open video source: {args.video}")return# 创建窗口和鼠标回调window_name = "Tracker"cv2.namedWindow(window_name)cv2.setMouseCallback(window_name, mouse_callback)tracker = Nonefont = cv2.FONT_HERSHEY_SIMPLEXwhile True:ret, frame = cap.read()if not ret:print("End of video")break# 检测目标boxes, class_names = detect_objects(frame, model, args.confidence)# 处理点击选择目标if mouse_clicked and not tracking:for i, (box, cls) in enumerate(zip(boxes, class_names)):if point_in_box(click_x, click_y, box):selected_box = boxtarget_class = clstracker = create_tracker(args.tracker)if tracker:tracker.init(frame, selected_box)tracking = Trueprint(f"Tracking {target_class} with {args.tracker}")breakmouse_clicked = False# 跟踪逻辑if tracking and tracker:ok, bbox = tracker.update(frame)if ok:x, y, w, h = [int(v) for v in bbox]cv2.rectangle(frame, (x, y), (x + w, y + h), (0, 255, 0), 2)cv2.putText(frame, f"Track: {target_class}", (x, y - 10),font, 0.5, (0, 255, 0), 2)else:cv2.putText(frame, "Lost", (10, 30), font, 0.7, (0, 0, 255), 2)tracking = False# 未跟踪时显示所有目标if not tracking:for box, cls in zip(boxes, class_names):x, y, w, h = boxcv2.rectangle(frame, (x, y), (x + w, y + h), (255, 0, 0), 2)cv2.putText(frame, cls, (x, y - 10), font, 0.5, (255, 0, 0), 2)# 显示提示信息hint = "Right click to stop" if tracking else "Left click to select"cv2.putText(frame, hint, (10, 30), font, 0.7, (0, 255, 255), 2)cv2.putText(frame, f"Tracker: {args.tracker}", (10, frame.shape[0] - 30),font, 0.5, (255, 255, 0), 2)cv2.putText(frame, f"FPS: {int(cap.get(cv2.CAP_PROP_FPS))}",(10, frame.shape[0] - 10), font, 0.5, (255, 255, 0), 2)# 显示窗口cv2.imshow(window_name, frame)# 退出条件if cv2.waitKey(1) & 0xFF == ord('q'):breakcap.release()cv2.destroyAllWindows()

使用方法

环境配置

首先安装必要的依赖库：

pip install opencv-python opencv-contrib-python ultralytics numpy

运行参数

程序支持以下命令行参数：

--video：视频源路径，默认为 “xxxxx.mp4”，使用 “0” 可调用摄像头
--confidence：目标检测置信度阈值，默认 0.5
--model：YOLO 模型路径，默认 “yolo12n.pt”（会自动下载）
--tracker：跟踪算法选择，可选 “mosse”、“csrt”、“kcf”，默认 “mosse”
--debug：启用调试模式，打印额外信息

操作指南

运行程序后，系统默认处于目标检测模式，所有检测到的目标用蓝色框标记
左键点击某个目标框，系统将开始用绿色框跟踪该目标
右键点击可停止跟踪，回到目标检测模式
按 “q” 键退出程序

算法特性对比

三种跟踪算法各有特点：

MOSSE：速度最快，适合实时性要求高的场景，但精度相对较低
CSRT：精度最高，但速度较慢，适合对精度要求高的场景
KCF：介于前两者之间，平衡了速度和精度

根据实际应用场景选择合适的跟踪算法可以获得更好的效果。

总结与扩展

本文介绍的目标跟踪系统结合了 YOLO 的强检测能力和传统跟踪算法的高效性，通过简单的交互实现了灵活的目标跟踪功能。该系统可进一步扩展，例如：

增加多目标跟踪功能
结合 ReID（重识别）技术解决目标遮挡问题
加入目标行为分析模块
优化跟踪失败后的自动重新检测机制

通过这个系统，不仅可以快速实现实用的目标跟踪应用，也有助于理解目标检测与跟踪相结合的技术路线，为更复杂的计算机视觉系统开发奠定基础。