EmguCv3中使用决策树

决策树训练与图像分类

最新推荐文章于 2025-09-12 09:59:14 发布

cccccc1212

最新推荐文章于 2025-09-12 09:59:14 发布

阅读量1.8k

点赞数

CC 4.0 BY-SA版权

文章标签： emgucv 机器学习

本文链接：https://blog.youkuaiyun.com/qq_22033759/article/details/48815757

EmguCV 专栏收录该内容

27 篇文章

订阅专栏

本文介绍了一个使用决策树进行训练的过程，并将其应用于图像分类任务中。通过具体的代码实现，展示了如何构造训练数据集，训练决策树模型，并利用该模型预测不同图像像素点的类别，最终形成可视化分类结果。

由于我训练的范围太小，图片也要放大到像素级才能看得清。

        public Form1()
        {
            InitializeComponent();
            DTrees tree = new DTrees();
            float[,] fdata = new float[5, 2] { { 1, 1 }, { 2, 2 }, { 1, 0 }, { 0, 1 }, { 0, 0 } };
            Image<Gray, float> data = new Image<Gray, float>(5, 2);
            float[] fresponses = new float[5] { 1, 1, 0, 0, 0 };
            Image<Gray, float> responses = new Image<Gray, float>(5, 1);
            for (int i = 0; i < 5; i++)
            {
                for (int j = 0; j < 2; j++)
                    data[j, i] = new Gray(fdata[i, j]);
            }
            for (int i = 0; i < 5; i++)
            {
                responses[0, i] = new Gray(fresponses[i]);
            }
            tree.Use1SERule = true;
            tree.UseSurrogates = false;
            tree.TruncatePrunedTree = true;
            tree.MaxDepth = 8;
            tree.MinSampleCount = 1;
            tree.RegressionAccuracy = 0;
            tree.MaxCategories = 15;
            tree.CVFolds = 0;

            try
            {
                tree.Train(data, DataLayoutType.ColSample, responses);
            }
            catch (Exception ex)
            {
                MessageBox.Show(ex.Message);
            }
            Image<Bgr, byte> aaa = new Image<Bgr, byte>(4, 4);
            for (int i = 0; i < aaa.Height; i++)
                for (int j = 0; j < aaa.Width; j++)
                {
                    //int k = i * 512 + j;
                    Image<Gray, float> bbb = new Image<Gray, float>(5, 1);
                    bbb[0, 0] = new Gray(i);
                    bbb[0, 1] = new Gray(j);
                    bbb[0, 2] = bbb[0, 3] = bbb[0, 4] = new Gray(0);
                    //Mat res = new Mat();
                    float tt = tree.Predict(bbb);
                    if (tt == 1)
                        aaa[i, j] = new Bgr(0, 0, 255);
                    else if (tt == 0)
                        aaa[i, j] = new Bgr(0, 255, 0);
                }
            //CvInvoke.Circle(aaa, new Point(100, 250), 5, new MCvScalar(0, 0, 0), -1);
            //CvInvoke.Circle(aaa, new Point(75, 11), 5, new MCvScalar(0, 0, 0), -1);
            //CvInvoke.Circle(aaa, new Point(500, 400), 5, new MCvScalar(255, 255, 255), -1);
            //CvInvoke.Circle(aaa, new Point(350, 20), 5, new MCvScalar(255, 255, 255), -1);
            //CvInvoke.Circle(aaa, new Point(190, 100), 5, new MCvScalar(255, 255, 255), -1);
            aaa[1, 1] = new Bgr(0, 0, 0);
            aaa[2, 2] = new Bgr(0, 0, 0);
            aaa[0, 0] = new Bgr(255, 255, 255);
            aaa[1, 0] = new Bgr(255, 255, 255);
            aaa[0, 1] = new Bgr(255, 255, 255);
            imageBox1.Image = aaa;
        }