QT部署YOLO11实时检测
实验视频
qt_yolo11测试视频配合博客
这篇文章主要讲述的是在QT中编辑检测界面以及输出实时目标检测的画框界面，是基于YOLOV11进行开发的，在完成这篇文章之前请首先参考这篇文章正点原子瑞芯微RK3568部署YOLOV11模型，因为这篇文章主要是在之前那篇文章的基础上进行试验的。绝大多数文件都是基于官方代码移植的。
步骤1：我的思路是TakePhoto 按钮拍摄图片然后检测将图片存储到当前路径下，而VideoFlow主要就是实时监测目标然后在Frame中实时显示,stop按钮就是退出当前所有功能。代码主要集中也是两块mainwindow.cpp、mainwindow.h。也就实现按键拍照，和调用摄像头驱动读取视频流的功能。
mainwindow.h
cpp
#ifndef MAINWINDOW_H
#define MAINWINDOW_H
#include <QMainWindow>
class QPushButton;
class QLabel;
class QVBoxLayout;
class QCamera;
class QVideoWidget;
class QCameraImageCapture; // take photo class
class MainWindow : public QMainWindow
{
Q_OBJECT
public:
MainWindow(QWidget *parent = nullptr);
~MainWindow();
private slots:
void onTakePhotoClicked(); // 预留，暂不实现
void onVideoFlowClicked(); // 启动摄像头实时显示
void onStopClicked(); // 停止摄像头并清理
private:
void setupUI();
void startCamera(); // 开始摄像头预览
void stopCamera(); // 停止摄像头预览
QWidget *centralWidget;
QLabel *frameLabel; // 占位标签（无视频时显示）
QVideoWidget *videoWidget; // 实际视频显示控件
QPushButton *takePhotoButton;
QPushButton *videoFlowButton;
QPushButton *stopButton;
QVBoxLayout *mainLayout;
QCamera *camera;
QCameraImageCapture *imageCapture; // 新增：用于拍照
};
#endif // MAINWINDOW_Hmainwindow.cpp
c
#include "mainwindow.h"
#include <QPushButton>
#include <QLabel>
#include <QVBoxLayout>
#include <QHBoxLayout>
#include <QCamera>
#include <QCameraInfo>
#include <QVideoWidget>
#include <QCameraImageCapture> //
#include <QDebug>
#include <QDateTime> // 用于生成文件名
#include <QFile> // 文件操作
#include <QEventLoop> // 用于同步等待
#include <QFileInfo>
MainWindow::MainWindow(QWidget *parent)
: QMainWindow(parent)
, videoWidget(nullptr)
, camera(nullptr)
, imageCapture(nullptr)
{
setupUI();
}
MainWindow::~MainWindow()
{
stopCamera(); // 析构时确保释放摄像头
}
void MainWindow::setupUI()
{
centralWidget = new QWidget(this);
setCentralWidget(centralWidget);
// 显示区域：初始是一个 QLabel，用于提示
frameLabel = new QLabel("点击 VideoFlow 启动摄像头", this);
frameLabel->setAlignment(Qt::AlignCenter);
frameLabel->setStyleSheet("border: 1px solid gray; background-color: black; color: white;");
frameLabel->setFixedSize(640, 480);
// 三个按钮
takePhotoButton = new QPushButton("TakePhoto", this);
videoFlowButton = new QPushButton("VideoFlow", this);
stopButton = new QPushButton("Stop", this);
QHBoxLayout *buttonLayout = new QHBoxLayout;
buttonLayout->addWidget(takePhotoButton);
buttonLayout->addWidget(videoFlowButton);
buttonLayout->addWidget(stopButton);
mainLayout = new QVBoxLayout(centralWidget);
mainLayout->addWidget(frameLabel);
mainLayout->addLayout(buttonLayout);
// 连接信号槽
connect(takePhotoButton, &QPushButton::clicked, this, &MainWindow::onTakePhotoClicked);
connect(videoFlowButton, &QPushButton::clicked, this, &MainWindow::onVideoFlowClicked);
connect(stopButton, &QPushButton::clicked, this, &MainWindow::onStopClicked);
}
void MainWindow::startCamera()
{
if (camera) {
qDebug() << "Camera already started";
return;
}
// 获取可用摄像头
const QList<QCameraInfo> cameras = QCameraInfo::availableCameras();
if (cameras.isEmpty()) {
qDebug() << "No camera found!";
frameLabel->setText("未找到摄像头");
return;
}
// 使用第一个摄像头
camera = new QCamera(cameras.first(), this);
// 创建视频显示控件
videoWidget = new QVideoWidget(this);
videoWidget->setFixedSize(640, 480);
// 用 videoWidget 替换布局中的 frameLabel
mainLayout->replaceWidget(frameLabel, videoWidget);
frameLabel->hide(); // 隐藏占位标签，但保留以便后面恢复
// 设置摄像头取景器为 videoWidget
camera->setViewfinder(videoWidget);
// 可选：设置分辨率（根据摄像头支持情况）
QCameraViewfinderSettings settings;
settings.setResolution(640, 480);
if (camera->supportedViewfinderSettings().contains(settings)) {
camera->setViewfinderSettings(settings);
}
// *** 关键：创建 QCameraImageCapture 对象 ***
imageCapture = new QCameraImageCapture(camera, this);
imageCapture->setCaptureDestination(QCameraImageCapture::CaptureToFile);
// 启动摄像头
camera->start();
qDebug() << "Camera started";
}
void MainWindow::stopCamera()
{
if (camera) {
camera->stop();
delete imageCapture; // 删除拍照对象
imageCapture = nullptr;
delete camera;
camera = nullptr;
}
if (videoWidget) {
mainLayout->removeWidget(videoWidget);
delete videoWidget;
videoWidget = nullptr;
}
// 恢复显示 frameLabel
frameLabel->show();
frameLabel->setText("摄像头已停止");
qDebug() << "Camera stopped";
}
void MainWindow::onTakePhotoClicked()
{
if (!camera || camera->state() != QCamera::ActiveState || !imageCapture) {
qDebug() << "Camera not active, cannot take photo";
return;
}
QEventLoop loop;
QImage capturedImage;
connect(imageCapture, &QCameraImageCapture::imageCaptured, &loop,
[&](int id, const QImage &preview) {
Q_UNUSED(id);
capturedImage = preview;
loop.quit();
});
imageCapture->capture();
loop.exec();
// 生成带时间戳的文件名
QString fileName = QString("photo_%1.jpg")
.arg(QDateTime::currentDateTime().toString("yyyyMMdd_hhmmss_zzz"));
// 获取当前工作目录的绝对路径
QString absolutePath = QFileInfo(fileName).absoluteFilePath();
if (capturedImage.save(fileName)) {
qDebug() << "Photo saved:" << absolutePath;
} else {
qDebug() << "Failed to save photo!";
}
}
void MainWindow::onVideoFlowClicked()
{
startCamera(); // 启动实时显示
}
void MainWindow::onStopClicked()
{
stopCamera(); // 停止摄像头并清理显示
}测试拉取图片拍照效果如下。
c
alientek@alientek-virtual-machine:~/qt_project/qt_yolo11$ adb pull /userdata/aidemo/rknn_yolo11_demo/photo_19700317_120709_094.jpg
/userdata/aidemo/rknn_yolo11_demo/phot...pped. 0.6 MB/s (51265 bytes in 0.081s)
你可以看到我还有一个yolo11文件这是我将/home/alientek/software/rknn_model_zoo/examples/yolo11
移植拷贝过来的。
步骤二：一个按钮YOLOFlowButton，当按下VideoFlowButton后再按下此按键开启yolo检测，构建一个新类YOLO类全部放到detector.cpp中。首先构思计划的是，类似于官方的main.cc应用程序，直接调用模型，将调用的一系列接口全部放到detector.cpp中，然后用图像处理的库推送图片进行检测即可。
官方的yolo11的CPP中，首先观察到main.cc中调用的头文件比较多，那么需要将头文件以及.c文件导入到qt的pro中。
还得查找这些头文件存放的地址。地址如下在rknn_model_zoo -> utils 。
后续写完代码后运行出现这个问题，主要是因为
QVideoFrame::imageFormatFromPixelFormat() 无法识别摄像头实际输出的格式。日志中显示使用了 mplane plugin，这是 V4L2 的 VIDIOC_G_FMT 获取的格式，常见的有 V4L2_PIX_FMT_NV12、V4L2_PIX_FMT_YUYV 或 V4L2_PIX_FMT_MJPEG。Qt 的 QVideoFrame::PixelFormat 枚举可能不包含这些，导致返回 Invalid。
修改 processFrame 函数，手动处理摄像头常见的格式。由于 RK3568 平台通常使用 V4L2 驱动，帧数据多为 NV12 或 YUYV。我们将添加对这些格式的支持。
你的摄像头支持 NV12、YU12、UYVY 等多种格式。Qt 无法直接将这些格式转为 QImage，所以你需要手动实现转换。最简单高效的方法是使用 libyuv 库。
后来我想那是不是opencv库也可以实现这个工作呢？正点原子的RK3568的opencv为4.5.5版本，SDK中的也是这个版本，我只要pro文件中导入这个库，头文件地址即可。
路径地址，一定要查找是不是这个文件地址正点原子的应该都是这个地址。
从文件路径中可以看出我使用了官方的yolo11文件等。
文件有多处地方是需要注意使用自己的库文件地址以及rknn模型地址的指定。这是编译完成后的样子。
qt_yolo11.pro文件参考
c
#-------------------------------------------------
#
# Project created by QtCreator 2026-04-06T16:20:49
#
#-------------------------------------------------
QT += core gui multimedia multimediawidgets widgets
greaterThan(QT_MAJOR_VERSION, 4): QT += widgets
TARGET = qt_yolo11
TEMPLATE = app
# The following define makes your compiler emit warnings if you use
# any feature of Qt which has been marked as deprecated (the exact warnings
# depend on your compiler). Please consult the documentation of the
# deprecated API in order to know how to port your code away from it.
DEFINES += QT_DEPRECATED_WARNINGS
DEFINES += RKNPU2
# You can also make your code fail to compile if you use deprecated APIs.
# In order to do so, uncomment the following line.
# You can also select to disable deprecated APIs only up to a certain version of Qt.
#DEFINES += QT_DISABLE_DEPRECATED_BEFORE=0x060000 # disables all the APIs deprecated before Qt 6.0.0
CONFIG += c++11
# 定义OpenCV路径
TOOLCHAIN_PATH = /opt/atk-dlrk356x-toolchain
SYSROOT_PATH = $$TOOLCHAIN_PATH/aarch64-buildroot-linux-gnu/sysroot
# OpenCV头文件路径
INCLUDEPATH += /opt/atk-dlrk356x-toolchain/aarch64-buildroot-linux-gnu/sysroot/usr/include/opencv4
LIBS += -L/opt/atk-dlrk356x-toolchain/aarch64-buildroot-linux-gnu/sysroot/usr/lib
LIBS += -lopencv_core
LIBS += -lopencv_imgproc
LIBS += -lopencv_imgcodecs
INCLUDEPATH += $$PWD
INCLUDEPATH += $$PWD/yolo11/cpp
INCLUDEPATH += /home/alientek/software/rknn_model_zoo/utils
INCLUDEPATH += /home/alientek/software/rknn_model_zoo/examples/yolo11/cpp
INCLUDEPATH += /home/alientek/software/rknn_model_zoo/3rdparty/rknpu2/include
INCLUDEPATH += /home/alientek/software/rknn_model_zoo/3rdparty/librga/include
INCLUDEPATH += /home/alientek/software/rknn_model_zoo/3rdparty/stb_image
# OpenCV库路径
LIBS += -L$$SYSROOT_PATH/usr/lib
LIBS += -lopencv_core
LIBS += -lopencv_imgproc
LIBS += -lopencv_imgcodecs
LIBS += -lopencv_videoio
LIBS += -lopencv_highgui
LIBS += -lopencv_video
SOURCES += \
/home/alientek/software/rknn_model_zoo/examples/yolo11/cpp/postprocess.cc \
/home/alientek/software/rknn_model_zoo/examples/yolo11/cpp/rknpu2/yolo11.cc \
main.cpp \
mainwindow.cpp \
detector.cpp \
/home/alientek/software/rknn_model_zoo/utils/file_utils.c \
/home/alientek/software/rknn_model_zoo/utils/image_utils.c \
/home/alientek/software/rknn_model_zoo/utils/image_drawing.c
HEADERS += mainwindow.h \
detector.h \
yolo11/cpp/postprocess.h \
yolo11/cpp/yolo11.h \
rknn_api.h \
common.h \
/home/alientek/software/rknn_model_zoo/3rdparty/librga/include/im2d.h
/home/alientek/software/rknn_model_zoo/3rdparty/stb_image/stb_image.h
FORMS += \
mainwindow.ui
LIBS += -L/home/alientek/software/rknn_model_zoo/3rdparty/rknpu2/runtime/Linux/aarch64
LIBS += -lrknnrt
LIBS += -lpthread -ldl -lm
LIBS += -L/usr/lib/aarch64-linux-gnu -lturbojpeg
LIBS += -L/home/alientek/software/rknn_model_zoo/install/rk356x_linux_aarch64/rknn_yolo11_demo/lib -lrga
# Default rules for deployment.
qnx: target.path = /tmp/$${TARGET}/bin
else: unix:!android: target.path = /opt/$${TARGET}/bin
!isEmpty(target.path): INSTALLS += target上传之后打开文件。
c
root@ATK-DLRK3568:/userdata/aidemo/rknn_yolo11_demo# ./qt_yolo11
c
alientek@alientek-virtual-machine:~/qt_project/build-qt_yolo11-rk3568-Debug$ adb push qt_yolo11 /userdata/aidemo/rknn_yolo11_demo
qt_yolo11: 1 file pushed, 0 skipped. 900.8 MB/s (244600 bytes in 0.000s)实验结果：
代码程序：网盘分享qt_yolov11 链接
https://pan.baidu.com/s/11ClnOe37rOUHA5pUxI2H_g?pwd=3qtc 提取码: 3qtc