C# から 顔認識によく用いられる OpenCV 2.4 の CascadeClassifier をつかうための C++/CLI のコード。System.Drawing の Bitmap が利用可能です。C++/CLI のプロジェクトを作ったら、まず System::Drawing の参照設定を行い、その後 nuget で OpenCV を追加してください。逆だとうまく参照設定できないようです (Visual Studio 2012, 2013)。このコードで作成した DLL のほかに、opencv_core24x.dll, opencv_highgui24x.dll, opencv_imgproc24x.dll, opencv_objdetect24x.dll だけあればうごきます。

CascadeClassifier.h

#pragma once

#include <opencv2/opencv.hpp>

using namespace System;
using namespace System::Drawing;
using namespace System::Collections::Generic;

namespace OpenCv {

	public ref class CascadeClassifier
	{
		public:
			property bool Loaded{
				bool get(){ return loaded; }
			}

		private:
			cv::CascadeClassifier* cascade;
			bool loaded;

		public:
			CascadeClassifier(System::String^ cascadeFileName){
				std::string filename = convertToStdString(cascadeFileName);
				cascade = new cv::CascadeClassifier();

				loaded = cascade->load(filename);
			}

		public:
			List<Rectangle>^ DetectMultiScale(Bitmap^ bitmap, double scaleFactor, int minNeighbors, Size minSize, Size maxSize){
				auto iplImage = getIplImage(bitmap);
				cv::Mat image = cv::cvarrToMat(iplImage);

				cv::Mat gray;
				cv::cvtColor(image, gray, CV_BGR2GRAY);
				cvReleaseImage(&iplImage);

				std::vector<cv::Rect> objs;

				cascade->detectMultiScale(gray, objs,
					scaleFactor, minNeighbors,
					CV_HAAR_SCALE_IMAGE,
					cv::Size(minSize.Width, minSize.Height),
					cv::Size(maxSize.Width, maxSize.Height));

				std::vector<cv::Rect>::const_iterator r = objs.begin();

				List<Rectangle>^ rectangles = gcnew List<Rectangle>();
				for (; r != objs.end(); ++r) {
					Rectangle rect = Rectangle(r->x, r->y, r->width, r->height);
					rectangles->Add(rect);
				}

				return rectangles;
			}

		private:
			IplImage* getIplImage(Bitmap^ bitmap){
				Drawing::Imaging::BitmapData^ data;

				IplImage *image = cvCreateImage(cvSize(bitmap->Width, bitmap->Height), IPL_DEPTH_8U, 3);

				data = bitmap->LockBits(
					Drawing::Rectangle(0, 0, bitmap->Width, bitmap->Height),
					Drawing::Imaging::ImageLockMode::ReadOnly,
					Drawing::Imaging::PixelFormat::Format24bppRgb
					);

				memcpy(image->imageData, data->Scan0.ToPointer(), image->imageSize);

				bitmap->UnlockBits(data);
				return image;
			}

		private:
			std::string convertToStdString(String^ str)
			{
				std::string stdStr;
				if (str != nullptr &&  str->Length > 0){
					array<Byte>^ data = System::Text::Encoding::Convert(
						System::Text::Encoding::Unicode,
						System::Text::Encoding::Default,
						System::Text::Encoding::Unicode->GetBytes(str));
					pin_ptr<Byte> pin = &data[0];
					stdStr.assign(reinterpret_cast<char*>(pin), data->Length);
				}
				return stdStr;
			}
	};
}

Form1.cs

using OpenCv;
using System;
using System.Collections.Generic;
using System.Drawing;
using System.Windows.Forms;

namespace ObjectRecognitionTest
{
    public partial class Form1 : Form
    {
        public Form1()
        {
            InitializeComponent();
        }

        private void button1_Click(object sender, EventArgs e)
        {
            CascadeClassifier detector = new CascadeClassifier(@"cascade.xml");

            if (detector.Loaded)
            {
                Bitmap bmp = new Bitmap("test.png");
                List<Rectangle> rects = detector.DetectMultiScale(bmp, 1.1, 3, new Size(20,20), new Size(100,100));
                using (Graphics g = Graphics.FromImage(bmp))
                {
                    foreach (Rectangle r in rects)
                    {
                        g.DrawRectangle(Pens.Red, r);
                    }
                }
                this.pictureBox1.Image = bmp;
            }
        }
    }
}