c语言0xC0000005
// SupportVectormachineClassification.cpp : 定义控制台应用程序的入口点。
//
#include "stdafx.h"
struct PTS{
int x;// x coordinate
int y;// y coordinate
int l;// label
};
void SVMClassification()
{
// image files (directories should be marked by '')
const TCHAR * trianing_sample_pts =_T("C:UsersAdministratorDesktopBYSJyangbenjieguojieguo1_30jieguo1subset_1.csv");
const TCHAR * image_data =_T("C:UsersAdministratorDesktopBYSJBYLWSYSJITUXIANG");// image data should be BSQ format!
const TCHAR * classification_result =_T("C:UsersAdministratorDesktopBYSJ uxiangfenleijieguo esult_1_30");
int w = 1500;// image width
int h = 1000;// image height
int dim = 9;// number of image feature bands
// read training samples
_AutoFilePtr train_pts_fp(trianing_sample_pts,_T("r"));
if(0 == train_pts_fp.get()){
std::cerr<<"open training sample file error!"<<std::endl;
return;
}
int num_class;// number of classes
int num_pts;// number of training samples points
_ftscanf(train_pts_fp.get(),_T("%d,%d "),&num_class,&num_pts);
_AutoPtr<PTS> points(num_pts);
if(0 == points.get()){
std::cerr<<"memory application error for training points!"<<std::endl;
return;
}
for(int i=0;i<num_pts;++i){
_ftscanf(train_pts_fp.get(),_T("%d,%d,%d "),&points[i].l,&points[i].x,&points[i].y);
}
_AutoPtr<float> train_label(num_pts);
if(0 == train_label.get()){
std::cerr<<"memory application for training data label information"<<std::endl;
return;
}
for(int i=0;i<num_pts;++i){
train_label[i] = static_cast<float>(points[i].l);
}
// read sample feature data
_AutoFilePtr data_fp(image_data,_T("rb"));
if(0 == data_fp.get()){
std::cerr<<"file open error for image data!"<<std::endl;
return;
}
int all_size = w*h*dim;
_AutoPtr<float> image_data_all(all_size);
if(0 == image_data_all.get()){
std::cerr<<"memory application error for image data!"<<std::endl;
return;
}
fread(image_data_all.get(),sizeof(float),all_size,data_fp.get());
int data_size = num_pts*dim;
_AutoPtr<float> image_data_train_pts(data_size);
if(0 == image_data_train_pts.get()){
std::cerr<<"memory application error for training points feature data!"<<std::endl;
return;
}
int num_pixels = w*h;
for(int i=0;i<num_pts;++i){
int image_offset = points[i].y*w + points[i].x;
for(int j=0;j<dim;++j){
//rewind(data_fp.get());
//fseek(data_fp.get(),image_offset+num_pixels*j,SEEK_SET);
//float value;
//fread(&value,sizeof(float),1,data_fp.get());
float value = image_data_all[image_offset+j*num_pixels];
image_data_train_pts[i*dim+j] = value;
}
}
// construct data for random forest training
Mat train_label_mat(num_pts,1,CV_32FC1,train_label.get());
Mat train_data_mat(num_pts,dim,CV_32FC1,image_data_train_pts.get());
// train random forest
CvSVM svm_alg;
CvSVMParams params;
params.svm_type = CvSVM::C_Svc;
params.kernel_type = CvSVM::RBF;// radial basis function
params.term_crit = cvTermCriteria(CV_TERMCRIT_ITER,100,1e-6);//termination: end when maximal iterative number is reached
int cv_fold = 10;// very important! 10 as a default setting, but smaller number is needed when the number of training samples are too low
svm_alg.train_auto(train_data_mat,train_label_mat,Mat(),Mat(),params,cv_fold);// this method allows grid search
Mat var_type = Mat(dim + 1, 1, CV_8U );
var_type.setTo(Scalar(CV_VAR_NUMERICAL) ); // all inputs are numerical
var_type.at<uchar>(dim, 0) = CV_VAR_CATEGORICAL;// this is a classification problem
svm_alg.clear();
svm_alg.train_auto(train_data_mat,train_label_mat,Mat(),Mat(),params,cv_fold);
// prediction (classification)
_AutoPtr<float> vec_data(dim);// to store the feature vector for each pixel
if(0 == vec_data.get()){
std::cerr<<"memory application error for pixel feature vector data!"<<std::endl;
return;
}
_AutoPtr<short int> result(num_pixels);
if(0 == result.get()){
std::cerr<<"memory application error for result!"<<std::endl;
return;
}
for(int i=0;i<h;++i){
for(int j=0;j<w;++j){
int image_offset = i*w+j;// current pixel position
for(int k=0;k<dim;++k){
//rewind(data_fp.get());
//fseek(data_fp.get(),image_offset+num_pixels*k,SEEK_SET);
//float value;
//fread(&value,sizeof(float),1,data_fp.get());
float value = image_data_all[image_offset + k*num_pixels];
vec_data[k] = value;
}
Mat feature_vec(1,dim,CV_32FC1,vec_data.get());
short int result_label = static_cast<short int>(svm_alg.predict(feature_vec));
result[image_offset] = result_label;
}
}
// output classification result
_AutoFilePtr result_fp(classification_result,_T("wb+"));
if(0 == result_fp.get()){
std::cerr<<"open file error for result!"<<std::endl;
return;
}
fwrite(result.get(),num_pixels,sizeof(short int),result_fp.get());
}
int _tmain(int argc, _TCHAR* argv[])
{
SVMClassification();
return 0;
}
怎么改啊 各位大神
首先，你的程序不是c语言的而是C++的（它们不是同一语言啊）
其次，你确定你的两个原始文件确实存在且图像一定是BSQ格式？
个人一向的建议，初学者不要尝试自己看不懂的程序，这样没有意义
（网上程序大多不能直接编译运行的）
上一篇：用c语言完成下列问题！！！