【重要说明】
该系统以opencvsharp作图像处理,onnxruntime做推理引擎,使用CPU进行推理,适合有显卡或者没有显卡windows x64系统均可,不支持macOS和Linux系统,不支持x86的windows操作系统。由于采用CPU推理,要比GPU慢。为了适合大部分操作系统我们暂时只写了CPU推理源码,GPU推理源码后期根据需要可能会调整,目前只考虑CPU推理,主要是为了照顾现在大部分使用该源码是学生,很多人并没有显卡的电脑情况。
【算法介绍】
基于YOLOv8的人员溺水检测告警监控系统是一种高效、智能的安全监控解决方案。该系统利用YOLOv8目标检测算法的先进性能,实现了对水域中人员溺水行为的精准识别与快速响应。
YOLOv8继承了YOLO系列算法的优点,如单次预测、速度快、精度高,并进一步优化了模型性能和泛化能力。该系统通过实时视频分析,能够在复杂环境中快速锁定水中人员的位置,并准确判断其是否处于溺水状态。
一旦检测到溺水行为,系统会立即触发告警,通知相关人员及时介入救援,从而有效减少溺水事故的发生。同时,该系统还支持视频数据采集自动化、数据传输和智能分析,为水域安全管理提供了全面的技术支持。
此外,基于YOLOv8的人员溺水检测告警监控系统还具备高度的灵活性和可扩展性,可根据实际需求进行定制化开发和部署,适用于各类水域场景的安全监控需求。
【效果展示】
【测试环境】
windows10 x64系统
VS2019
netframework4.7.2
opencvsharp4.8.0
onnxruntime1.16.3
【模型可以检测出类别】
Drowning
Person out of water
Swimming
【训练数据集】
https://blog.csdn.net/FL1623863129/article/details/140031636
注意模型训练数据集可能和上述数据集存在差异,源于对数据集优化整理可能会导致数据集类别或者图片数增加或者减少,大体数据集保持一致
【训练信息】
参数 | 值 |
训练集图片数 | 23898 |
验证集图片数 | 1173 |
训练map | 84.6% |
训练精度(Precision) | 88.2% |
训练召回率(Recall) | 77.2% |
验证集测试精度信息
类别 |
MAP50(单位:%) |
all |
83 |
Drowning |
87 |
Swimming |
83 |
【部分实现源码】
using OpenCvSharp;
using System;
using System.Collections.Generic;
using System.ComponentModel;
using System.Data;
using System.Diagnostics;
using System.Drawing;
using System.IO;
using System.Linq;
using System.Text;
using System.Threading;
using System.Threading.Tasks;
using System.Windows.Forms;
namespace FIRC
{
public partial class Form1 : Form
{
public bool videoStart = false;//视频停止标志
string weightsPath = Application.StartupPath + "\\weights";//模型目录
string labelTxt= Application.StartupPath + "\\weights\\class_names.txt";//类别文件
Yolov8Manager detetor = new Yolov8Manager();//推理引擎
public Form1()
{
InitializeComponent();
CheckForIllegalCrossThreadCalls = false;//线程更新控件不报错
}
private void LoadWeightsFromDir()
{
var di = new DirectoryInfo(weightsPath);
foreach(var fi in di.GetFiles("*.onnx"))
{
comboBox1.Items.Add(fi.Name);
}
if(comboBox1.Items.Count>0)
{
comboBox1.SelectedIndex = 0;
}
else
{
tssl_show.Text = "未找到模型,请关闭程序,放入模型到weights文件夹!";
tsb_pic.Enabled = false;
tsb_video.Enabled = false;
tsb_camera.Enabled = false;
}
}
private void Form1_Load(object sender, EventArgs e)
{
LoadWeightsFromDir();//从目录加载模型
}
public string GetResultString(Result result)
{
Dictionary<string, int> resultDict = new Dictionary<string, int>();
for (int i = 0; i < result.length; i++)
{
if(resultDict.ContainsKey( result.classes[i]) )
{
resultDict[result.classes[i]]++;
}
else
{
resultDict[result.classes[i]]=1;
}
}
var resultStr = "";
foreach(var item in resultDict)
{
resultStr += string.Format("{0}:{1}\n",item.Key,item.Value);
}
return resultStr;
}
private void tsb_pic_Click(object sender, EventArgs e)
{
OpenFileDialog ofd = new OpenFileDialog();
ofd.Filter = "*.*|*.bmp;*.jpg;*.jpeg;*.tiff;*.tiff;*.png";
if (ofd.ShowDialog() != DialogResult.OK) return;
tssl_show.Text = "正在检测中...";
Task.Run(() => {
var sw = new Stopwatch();
sw.Start();
Mat image = Cv2.ImRead(ofd.FileName);
detetor.Confidence =Convert.ToSingle(numericUpDown1.Value);
detetor.IOU = Convert.ToSingle(numericUpDown2.Value);
var results=detetor.Inference(image);
var resultImage = detetor.DrawImage(OpenCvSharp.Extensions.BitmapConverter.ToBitmap(image), results);
sw.Stop();
pb_show.Image = resultImage;
tb_res.Text = GetResultString(results);
tssl_show.Text = "检测已完成!总计耗时"+sw.Elapsed.TotalSeconds+"秒";
});
}
public void VideoProcess(string videoPath)
{
Task.Run(() => {
detetor.Confidence = Convert.ToSingle(numericUpDown1.Value);
detetor.IOU = Convert.ToSingle(numericUpDown2.Value);
VideoCapture capture = new VideoCapture(videoPath);
if (!capture.IsOpened())
{
tssl_show.Text="视频打开失败!";
return;
}
Mat frame = new Mat();
var sw = new Stopwatch();
int fps = 0;
while (videoStart)
{
capture.Read(frame);
if (frame.Empty())
{
Console.WriteLine("data is empty!");
break;
}
sw.Start();
var results = detetor.Inference(frame);
var resultImg = detetor.DrawImage(frame,results);
sw.Stop();
fps = Convert.ToInt32(1 / sw.Elapsed.TotalSeconds);
sw.Reset();
Cv2.PutText(resultImg, "FPS=" + fps, new OpenCvSharp.Point(30, 30), HersheyFonts.HersheyComplex, 1.0, new Scalar(255, 0, 0), 3);
//显示结果
pb_show.Image = OpenCvSharp.Extensions.BitmapConverter.ToBitmap(resultImg);
tb_res.Text = GetResultString(results);
Thread.Sleep(5);
}
capture.Release();
pb_show.Image = null;
tssl_show.Text = "视频已停止!";
tsb_video.Text = "选择视频";
});
}
public void CameraProcess(int cameraIndex=0)
{
Task.Run(() => {
detetor.Confidence = Convert.ToSingle(numericUpDown1.Value);
detetor.IOU = Convert.ToSingle(numericUpDown2.Value);
VideoCapture capture = new VideoCapture(cameraIndex);
if (!capture.IsOpened())
{
tssl_show.Text = "摄像头打开失败!";
return;
}
Mat frame = new Mat();
var sw = new Stopwatch();
int fps = 0;
while (videoStart)
{
capture.Read(frame);
if (frame.Empty())
{
Console.WriteLine("data is empty!");
break;
}
sw.Start();
var results = detetor.Inference(frame);
var resultImg = detetor.DrawImage(frame, results);
sw.Stop();
fps = Convert.ToInt32(1 / sw.Elapsed.TotalSeconds);
sw.Reset();
Cv2.PutText(resultImg, "FPS=" + fps, new OpenCvSharp.Point(30, 30), HersheyFonts.HersheyComplex, 1.0, new Scalar(255, 0, 0), 3);
//显示结果
pb_show.Image = OpenCvSharp.Extensions.BitmapConverter.ToBitmap(resultImg);
tb_res.Text = GetResultString(results);
Thread.Sleep(5);
}
capture.Release();
pb_show.Image = null;
tssl_show.Text = "摄像头已停止!";
tsb_camera.Text = "打开摄像头";
});
}
private void tsb_video_Click(object sender, EventArgs e)
{
if(tsb_video.Text=="选择视频")
{
OpenFileDialog ofd = new OpenFileDialog();
ofd.Filter = "视频文件(*.*)|*.mp4;*.avi";
if (ofd.ShowDialog() != DialogResult.OK) return;
videoStart = true;
VideoProcess(ofd.FileName);
tsb_video.Text = "停止";
tssl_show.Text = "视频正在检测中...";
}
else
{
videoStart = false;
}
}
private void tsb_camera_Click(object sender, EventArgs e)
{
if (tsb_camera.Text == "打开摄像头")
{
videoStart = true;
CameraProcess(0);
tsb_camera.Text = "停止";
tssl_show.Text = "摄像头正在检测中...";
}
else
{
videoStart = false;
}
}
private void tsb_exit_Click(object sender, EventArgs e)
{
videoStart = false;
this.Close();
}
private void trackBar1_Scroll(object sender, EventArgs e)
{
numericUpDown1.Value = Convert.ToDecimal(trackBar1.Value / 100.0f);
}
private void trackBar2_Scroll(object sender, EventArgs e)
{
numericUpDown2.Value = Convert.ToDecimal(trackBar2.Value / 100.0f);
}
private void numericUpDown1_ValueChanged(object sender, EventArgs e)
{
trackBar1.Value = (int)(Convert.ToSingle(numericUpDown1.Value) * 100);
}
private void numericUpDown2_ValueChanged(object sender, EventArgs e)
{
trackBar2.Value = (int)(Convert.ToSingle(numericUpDown2.Value) * 100);
}
private void comboBox1_SelectedIndexChanged(object sender, EventArgs e)
{
tssl_show.Text="加载模型:"+comboBox1.Text;
detetor.LoadWeights(weightsPath+"\\"+comboBox1.Text,labelTxt);
tssl_show.Text = "模型加载已完成!";
}
}
}
【使用步骤】
使用步骤:
(1)首先根据官方框架https://github.com/ultralytics/ultralytics安装教程安装好yolov8环境,并根据官方export命令将自己pt模型转成onnx模型
(2)使用vs2019打开sln项目,选择x64 release并且修改一些必要的参数,比如输入shape等,点击运行即可查看最后效果
特别注意如果运行报错了,请参考我的博文进行重新引用我源码的DLL:[C#]opencvsharp报错System.Memory,Version=4.0.1.2,Culture=neutral,PublicKeyToken=cc7b13fcd2ddd51“版本高于所引_未能加载文件或程序集“system.memory, version=4.0.1.2, culture-CSDN博客
【提供文件】
C#源码
yolov8n.onnx模型(不提供pytorch模型)
训练的map,P,R曲线图(在weights\results.png)
测试图片(在test_img文件夹下面)
【源码下载地址】
关注下方名片并回复【firc8】即可获取下载方式