opencv实现KNN手写数字的识别

　　人工智能是当下很热门的话题，手写识别是一个典型的应用。为了进一步了解这个领域，我阅读了大量的论文，并借助opencv完成了对28x28的数字图片（预处理后的二值图像）的识别任务。

　　预处理一张图片：

　　首先采用opencv读取图片的构造函数读取灰度的图片，再采用大津法求出图片的二值化的阈值，并且将图片二值化。

 int otsu(const IplImage* src_image) {
     double sum = 0.0;
     double w0 = 0.0;
     double w1 = 0.0;
     double u0_temp = 0.0;
     double u1_temp = 0.0;
     double u0 = 0.0;
     double u1 = 0.0;
     double delta_temp = 0.0;
     double delta_max = 0.0;

     int pixel_count[] = {  };
     float pixel_pro[] = {  };
     int threshold = ;
     uchar* data = (uchar*)src_image->imageData;
     for (int i = ; i < src_image->height; i++) {
         for (int j = ; j < src_image->width; j++) {
             pixel_count[(int)data[i * src_image->width + j]]++;
             sum += (int)data[i * src_image->width + j];
         }
     }
     for (int i = ; i < ; i++) {
         pixel_pro[i] = (float)pixel_count[i] / (src_image->height * src_image->width);
     }
     for (int i = ; i < ; i++) {
         w0 = w1 = u0_temp = u1_temp = u0 = u1 = delta_temp = ;
         for (int j = ; j < ; j++) {
             if (j <= i) {
                 w0 += pixel_pro[j];
                 u0_temp += j * pixel_pro[j];
             }
             else {
                 w1 += pixel_pro[j];
                 u1_temp += j * pixel_pro[j];
             }
         }
         u0 = u0_temp / w0;
         u1 = u1_temp / w1;
         delta_temp = (float)(w0 *w1* pow((u0 - u1), ));
         if (delta_temp > delta_max) {
             delta_max = delta_temp;
             threshold = i;
         }
     }
     return threshold;
 }

大津法

 void imageBinarization(IplImage* src_image) {
     IplImage* binImg = cvCreateImage(cvGetSize(src_image), src_image->depth, src_image->nChannels);
     CvScalar s;
     int ave = ;
     int binThreshold = otsu(src_image);

     for (int i = ; i < src_image->height; i++) {
         for (int j = ; j < src_image->width; j++) {
             s = cvGet2D(src_image, i, j);
             ave = (s.val[] + s.val[] + s.val[]) / ;
             if (ave < binThreshold) {
                 s.val[] = s.val[] = s.val[] = 0xff;
                 cvSet2D(src_image, i, j, s);
             }
             else {
                 s.val[] = s.val[] = s.val[] = 0x00;
                 cvSet2D(src_image, i, j, s);
             }
         }
     }
     cvCopy(src_image, binImg);
     cvSaveImage(bined, binImg);
     //cvShowImage("binarization", binImg);
     //waitKey(0);
 }

二值化

　　由于是只进行简单的识别模拟，因此没有做像素断点的处理。获取minst提供的数据集，提取每个图片的hog特征，参数如下：

 HOGDescriptor *hog = new HOGDescriptor(
             cvSize(ImgWidht, ImgHeight), cvSize(, ), cvSize(, ), cvSize(, ), );

　　（9个方向换成18个可能会取得更准确的结果，这取决于对图片本身的复杂程度的分析

　　之后即可训练knn分类器，进行分类了。

 void knnTrain() {
 #ifdef SAVETRAINED
     //knn training;
     samples.clear();
     dat_mat = Mat::zeros( * nImgNum, , CV_32FC1);
     res_mat = Mat::zeros( * nImgNum, , CV_32FC1);
     for (int i = ; i != ; i++) {
         getFile(dirNames[i], i);
     }
     preTrain();
     cout << "------ Training finished. -----" << endl << endl;
     knn.train(dat_mat, res_mat, Mat(), false, );

 #ifdef SAVEASXML
     knn.save("./trained/knnTrained.xml");
 #endif

 #else
     knn.load("./trained/knnTrained.xml");
 #endif

     //knn test
     cout << endl << "--- KNN test mode : ---" << endl;
     int tCnt = ;
     int tAc = ;
     selfknnTest(tCnt, tAc);

     cout << endl << endl << "Total number of test samples : " << tCnt << endl;

     cout << "Accuracy : " << float(float(tAc) / float(tCnt)) *  << "%" << endl;
 }

train

　训练结果如下，准确率还是很令人满意的。