Github Copilot 实战: 从零开始用AI写一个OCR工具 (3)
源码https://github.com/densen2014/Blazor100/tree/master/AI/MiOcr
添加一个屏幕截图功能,显示截图起始点,结束点,截图区域,按键ESC取消截图
这里AI就比较中规中矩,很快就能得到我要的功能了.下面只简单贴一下代码
ScreenCaptureWindow.xaml
<Window x:
xmlns:x="http://schemas.microsoft.com/winfx/2006/xaml"
xmlns="http://schemas.microsoft.com/winfx/2006/xaml/presentation"
Window AllowsTransparency="True" Background="#01000000"
Topmost="True" ShowInTaskbar="False" WindowState="Maximized">
<Canvas x:Name="CaptureCanvas">
<TextBlock x:Name="StartCoordText"
Foreground="Yellow"
Background="#80000000"
FontSize="14"
Visibility="Collapsed"/>
<TextBlock x:Name="CurrentCoordText"
Foreground="Yellow"
Background="#80000000"
FontSize="14"
Visibility="Collapsed"/>
<TextBlock x:Name="SizeText"
Foreground="Yellow"
Background="#80000000"
FontSize="14"
Visibility="Collapsed"/>
</Canvas>
</Window>ScreenCaptureWindow.xaml.cs
using System.Windows;
using System.Windows.Controls;
using System.Windows.Input;
using System.Windows.Interop;
using System.Windows.Media;
using System.Windows.Media.Imaging;
using System.Windows.Shapes;
namespace MiOcr;
public partial class ScreenCaptureWindow : Window
{
public Rect SelectedRect { get; private set; }
public BitmapSource? CapturedImage { get; private set; }
private System.Windows.Point? _start;
private Rectangle? _rectShape;
public ScreenCaptureWindow()
{
InitializeComponent();
MouseLeftButtonDown += OnMouseDown;
MouseMove += OnMouseMove;
MouseLeftButtonUp += OnMouseUp;
Cursor = Cursors.Cross;
PreviewKeyDown += ScreenCaptureWindow_PreviewKeyDown;
Focusable = true;
Loaded += (s, e) => Keyboard.Focus(this);
}
private void ScreenCaptureWindow_PreviewKeyDown(object sender, KeyEventArgs e)
{
if (e.Key == Key.Escape)
{
CapturedImage = null;
DialogResult = false;
Close();
}
}
private void PositionTextBlocks(double x, double y, double w, double h)
{
double margin = 8;
double canvasWidth = CaptureCanvas.ActualWidth;
double canvasHeight = CaptureCanvas.ActualHeight;
// 先测量文本大小
StartCoordText.Measure(new Size(double.PositiveInfinity, double.PositiveInfinity));
SizeText.Measure(new Size(double.PositiveInfinity, double.PositiveInfinity));
double startW = StartCoordText.DesiredSize.Width;
double startH = StartCoordText.DesiredSize.Height;
double sizeW = SizeText.DesiredSize.Width;
double sizeH = SizeText.DesiredSize.Height;
// 1. 左上优先
double startX = x - startW - margin;
double startY = y - startH - margin;
if (startX >= 0 && startY >= 0)
{
Canvas.SetLeft(StartCoordText, startX);
Canvas.SetTop(StartCoordText, startY);
Canvas.SetLeft(SizeText, startX);
Canvas.SetTop(SizeText, startY + startH + 4);
return;
}
// 2. 右上
startX = x + w + margin;
startY = y - startH - margin;
if (startX + startW <= canvasWidth && startY >= 0)
{
Canvas.SetLeft(StartCoordText, startX);
Canvas.SetTop(StartCoordText, startY);
Canvas.SetLeft(SizeText, startX);
Canvas.SetTop(SizeText, startY + startH + 4);
return;
}
// 3. 左下
startX = x - startW - margin;
startY = y + h + margin;
if (startX >= 0 && startY + startH + sizeH + 4 <= canvasHeight)
{
Canvas.SetLeft(StartCoordText, startX);
Canvas.SetTop(StartCoordText, startY);
Canvas.SetLeft(SizeText, startX);
Canvas.SetTop(SizeText, startY + startH + 4);
return;
}
// 4. 右下
startX = x + w + margin;
startY = y + h + margin;
if (startX + startW <= canvasWidth && startY + startH + sizeH + 4 <= canvasHeight)
{
Canvas.SetLeft(StartCoordText, startX);
Canvas.SetTop(StartCoordText, startY);
Canvas.SetLeft(SizeText, startX);
Canvas.SetTop(SizeText, startY + startH + 4);
return;
}
// 5. 屏幕内兜底
Canvas.SetLeft(StartCoordText, Math.Max(margin, Math.Min(canvasWidth - startW - margin, x)));
Canvas.SetTop(StartCoordText, Math.Max(margin, Math.Min(canvasHeight - startH - margin, y)));
Canvas.SetLeft(SizeText, Math.Max(margin, Math.Min(canvasWidth - sizeW - margin, x)));
Canvas.SetTop(SizeText, Math.Max(margin, Math.Min(canvasHeight - sizeH - margin, y + startH + 4)));
}
private void OnMouseDown(object sender, MouseButtonEventArgs e)
{
_start = e.GetPosition(this);
_rectShape = new Rectangle
{
Stroke = Brushes.Red,
StrokeThickness = 2,
Fill = new SolidColorBrush(Color.FromArgb(40, 0, 0, 255))
};
CaptureCanvas.Children.Add(_rectShape);
Canvas.SetLeft(_rectShape, _start.Value.X);
Canvas.SetTop(_rectShape, _start.Value.Y);
StartCoordText.Text = $"起点: ({(int)_start.Value.X}, {(int)_start.Value.Y})";
StartCoordText.Visibility = Visibility.Visible;
CurrentCoordText.Text = $"当前: ({(int)_start.Value.X}, {(int)_start.Value.Y})";
CurrentCoordText.Visibility = Visibility.Visible;
SizeText.Text = $"大小: 0 x 0";
SizeText.Visibility = Visibility.Visible;
// 初始位置
PositionTextBlocks(_start.Value.X, _start.Value.Y, 0, 0);
}
private void OnMouseMove(object sender, MouseEventArgs e)
{
if (_start.HasValue && _rectShape != null)
{
var pos = e.GetPosition(this);
double x = Math.Min(_start.Value.X, pos.X);
double y = Math.Min(_start.Value.Y, pos.Y);
double w = Math.Abs(_start.Value.X - pos.X);
double h = Math.Abs(_start.Value.Y - pos.Y);
Canvas.SetLeft(_rectShape, x);
Canvas.SetTop(_rectShape, y);
_rectShape.Width = w;
_rectShape.Height = h;
// 更新当前点坐标
CurrentCoordText.Text = $"当前: ({(int)pos.X}, {(int)pos.Y})";
Canvas.SetLeft(CurrentCoordText, pos.X + 2);
Canvas.SetTop(CurrentCoordText, pos.Y + 2);
// 更新区域大小
SizeText.Text = $"大小: {(int)w} x {(int)h}";
// 动态调整文本位置
PositionTextBlocks(x, y, w, h);
}
}
private void OnMouseUp(object sender, MouseButtonEventArgs e)
{
if (_start.HasValue && _rectShape != null)
{
var end = e.GetPosition(this);
double x = Math.Min(_start.Value.X, end.X);
double y = Math.Min(_start.Value.Y, end.Y);
double w = Math.Abs(_start.Value.X - end.X);
double h = Math.Abs(_start.Value.Y - end.Y);
SelectedRect = new Rect(x, y, w, h);
// 隐藏坐标
StartCoordText.Visibility = Visibility.Collapsed;
CurrentCoordText.Visibility = Visibility.Collapsed;
// 隐藏区域大小
SizeText.Visibility = Visibility.Collapsed;
// 截图
CapturedImage = CaptureScreenArea(SelectedRect);
DialogResult = true;
Close();
}
}
private BitmapSource CaptureScreenArea(Rect rect)
{
double dpiScale = NativeMethods.GetDpiScale(this);
int x = (int)(rect.X * dpiScale);
int y = (int)(rect.Y * dpiScale);
int w = (int)(rect.Width * dpiScale);
int h = (int)(rect.Height * dpiScale);
IntPtr hdcSrc = NativeMethods.GetDC(IntPtr.Zero);
IntPtr hdcDest = NativeMethods.CreateCompatibleDC(hdcSrc);
IntPtr hBitmap = NativeMethods.CreateCompatibleBitmap(hdcSrc, w, h);
IntPtr hOld = NativeMethods.SelectObject(hdcDest, hBitmap);
NativeMethods.BitBlt(hdcDest, 0, 0, w, h, hdcSrc, x, y, 0x00CC0020); // SRCCOPY
NativeMethods.SelectObject(hdcDest, hOld);
NativeMethods.DeleteDC(hdcDest);
NativeMethods.ReleaseDC(IntPtr.Zero, hdcSrc);
try
{
var source = Imaging.CreateBitmapSourceFromHBitmap(
hBitmap, IntPtr.Zero, Int32Rect.Empty, BitmapSizeOptions.FromEmptyOptions());
source.Freeze();
return source;
}
finally
{
NativeMethods.DeleteObject(hBitmap);
}
}
}步骤二：确保 PaddleOCRService.cs 的 StartOCR 支持 onStatusChanged
如前所述，PaddleOCRService.cs 需如下实现：
using System.Windows;
namespace MiOcr;
public static class NativeMethods
{
public static extern bool DeleteObject(IntPtr hObject);
public static extern IntPtr GetDC(IntPtr hWnd);
public static extern int ReleaseDC(IntPtr hWnd, IntPtr hDC);
public static extern IntPtr CreateCompatibleDC(IntPtr hdc);
public static extern IntPtr CreateCompatibleBitmap(IntPtr hdc, int nWidth, int nHeight);
public static extern IntPtr SelectObject(IntPtr hdc, IntPtr hgdiobj);
public static extern bool BitBlt(IntPtr hdcDest, int nXDest, int nYDest, int nWidth, int nHeight,
IntPtr hdcSrc, int nXSrc, int nYSrc, int dwRop);
public static extern bool DeleteDC(IntPtr hdc);
/// <summary>
/// 获取指定窗口的 DPI 缩放比例。
/// </summary>
/// <param name="window"></param>
/// <returns></returns>
public static double GetDpiScale(Window window)
{
var source = PresentationSource.FromVisual(window);
if (source?.CompositionTarget != null)
{
return source.CompositionTarget.TransformToDevice.M11; // X 方向缩放
}
return 1.0;
}
}这样，用户在模型首次下载或初始化时会看到“正在初始化OCR模型，请稍候...”，其余时间显示“正在识别...”，体验更友好。
回顾
项目目标
本项目旨在利用 Github Copilot 辅助开发，从零实现一个基于 AI 的 OCR（光学字符识别）工具。项目采用 .NET 9 和 WPF 技术栈，集成了 PaddleOCR 作为核心识别引擎，实现了图片文字识别、区域选择、结果高亮与复制等实用功能。
主要技术与依赖
• 开发语言与平台：C#，.NET 9，WPF
• OCR引擎：Sdcb.OpenVINO.PaddleOCR
• 图像处理：OpenCvSharp
• 界面交互：WPF，支持拖拽、粘贴、截图等多种图片输入方式
• AI辅助开发：Github Copilot 提供代码建议与自动补全
核心功能
[*]图片输入
• 支持文件选择、拖拽、粘贴、屏幕截图等多种方式加载图片。
[*]OCR识别
• 调用 PaddleOCR 进行文字识别，支持中文、英文等多语种。
• 识别结果实时显示，支持区域高亮和文字复制。
[*]用户体验优化
• 首次模型下载时，异步回调 UI，友好提示“正在初始化OCR模型，请稍候...”，避免用户等待时无响应。
• 识别过程有进度提示，提升交互体验。
[*]结果交互
• 支持鼠标框选图片区域，提取并复制选中区域的文字。
• 右键点击可复制单个识别文本。
关键实现思路
• AI驱动开发：通过 Copilot 自动生成代码骨架、方法实现和注释，大幅提升开发效率。
• 异步与回调：模型下载和识别过程均为异步，UI 通过回调及时反馈进度和状态。
• 图像与坐标映射：实现了图片与控件坐标的精准映射，保证高亮和选区准确。
总结
本项目充分发挥了 Github Copilot 在 AI 辅助开发中的优势，实现了一个功能完善、交互友好的 OCR 工具。开发过程中 Copilot 提供了大量代码建议，极大提升了开发效率和代码质量。项目结构清晰，易于扩展，适合 AI+开发实战学习与参考。
来源：程序园用户自行投稿发布，如果侵权，请联系站长删除
免责声明：如果侵犯了您的权益，请联系站长，我们会及时删除侵权内容，谢谢合作！
页:
[1]