p86743915
/
pp6hrvtlm
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

no commit message

Browse Source
devA
yuxue 5 years ago
parent 474df80320
commit aae605be80
1 changed files with 72 additions and 199 deletions
Show Stats Download Patch File Download Diff File

271
src/main/java/com/yuxue/train/ANNTrain.java
Unescape View File

@ -6,11 +6,8 @@ import java.util.Vector;
import org.opencv.core.Core;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
import org.opencv.core.Point;
import org.opencv.core.Size;
import org.opencv.core.TermCriteria;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import org.opencv.ml.ANN_MLP;
import org.opencv.ml.Ml;
import org.opencv.ml.TrainData;
@ -49,96 +46,6 @@ public class ANNTrain {
// 训练模型文件保存位置
private static final String MODEL_PATH = DEFAULT_PATH + "ann.xml";
/**
* 进行膨胀操作
* @param inMat
* @return
*/
public Mat dilate(Mat inMat) {
Mat result = inMat.clone();
Mat element = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(2, 2));
Imgproc.dilate(inMat, result, element);
return result;
}
/**
* 进行腐蚀操作
* @param inMat
* @return
*/
public Mat erode(Mat inMat) {
Mat result = inMat.clone();
Mat element = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(2, 2));
Imgproc.erode(inMat, result, element);
return result;
}
/**
* 随机数平移
* @param inMat
* @return
*/
public Mat randTranslate(Mat inMat) {
Random rand = new Random();
Mat result = inMat.clone();
int ran_x = rand.nextInt(10000) % 5 - 2; // 控制在-2~3个像素范围内
int ran_y = rand.nextInt(10000) % 5 - 2;
return translateImg(result, ran_x, ran_y);
}
/**
* 随机数旋转
* @param inMat
* @return
*/
public Mat randRotate(Mat inMat) {
Random rand = new Random();
Mat result = inMat.clone();
float angle = (float) (rand.nextInt(10000) % 15 - 7); // 旋转角度控制在-7~8°范围内
return rotateImg(result, angle);
}
/**
* 平移
* @param img
* @param offsetx
* @param offsety
* @return
*/
public Mat translateImg(Mat img, int offsetx, int offsety){
Mat dst = new Mat();
//定义平移矩阵
Mat trans_mat = Mat.zeros(2, 3, CvType.CV_32FC1);
trans_mat.put(0, 0, 1);
trans_mat.put(0, 2, offsetx);
trans_mat.put(1, 1, 1);
trans_mat.put(1, 2, offsety);
Imgproc.warpAffine(img, dst, trans_mat, img.size()); // 仿射变换
return dst;
}
/**
* 旋转角度
* @param source
* @param angle
* @return
*/
public Mat rotateImg(Mat source, float angle){
Point src_center = new Point(source.cols() / 2.0F, source.rows() / 2.0F);
Mat rot_mat = Imgproc.getRotationMatrix2D(src_center, angle, 1);
Mat dst = new Mat();
// 仿射变换 可以考虑使用投影变换; 这里使用放射变换进行旋转，对于实际效果来说感觉意义不大，反而会干扰结果预测
Imgproc.warpAffine(source, dst, rot_mat, source.size());
return dst;
}
public void train(int _predictsize, int _neurons) {
Mat samples = new Mat(); // 使用push_back，行数列数不能赋初始值
Vector<Integer> trainingLabels = new Vector<Integer>();
@ -149,68 +56,46 @@ public class ANNTrain {
Vector<String> files = new Vector<String>();
FileUtil.getFiles(str, files); // 文件名不能包含中文
int count = 200; // 控制从训练样本中，抽取指定数量的样本
// int count = 200; // 控制从训练样本中，抽取指定数量的样本
int count = files.size(); // 控制从训练样本中，抽取指定数量的样本
for (int j = 0; j < count; j++) {
Mat img = Imgcodecs.imread(files.get(rand.nextInt(files.size() - 1)), 0);
String filename = "";
if(j < files.size()) {
filename = files.get(j);
} else {
filename = files.get(rand.nextInt(files.size() - 1)); // 样本不足，随机重复提取已有的样本
}
Mat img = Imgcodecs.imread(filename, 0);
Mat f = PlateUtil.features(img, _predictsize);
samples.push_back(f);
trainingLabels.add(i); // 每一幅字符图片所对应的字符类别索引下标
// 增加随机平移样本
samples.push_back(PlateUtil.features(randTranslate(img), _predictsize));
samples.push_back(PlateUtil.features(PlateUtil.randTranslate(img), _predictsize));
trainingLabels.add(i);
// 增加随机旋转样本
samples.push_back(PlateUtil.features(randRotate(img), _predictsize));
samples.push_back(PlateUtil.features(PlateUtil.randRotate(img), _predictsize));
trainingLabels.add(i);
// 增加膨胀样本
/*samples.push_back(PlateUtil.features(dilate(img), _predictsize));
trainingLabels.add(i);*/
samples.push_back(PlateUtil.features(PlateUtil.dilate(img), _predictsize));
trainingLabels.add(i);
// 增加腐蚀样本
/*samples.push_back(PlateUtil.features(erode(img), _predictsize));
/*samples.push_back(PlateUtil.features(PlateUtil.erode(img), _predictsize));
trainingLabels.add(i); */
}
}
// 加载汉字字符
for (int i = 0; i < Constant.strChinese.length; i++) {
String str = DEFAULT_PATH + "learn/" + Constant.strChinese[i];
Vector<String> files = new Vector<String>();
FileUtil.getFiles(str, files);
int count = 200; // 控制从训练样本中，抽取指定数量的样本
for (int j = 0; j < count; j++) {
Mat img = Imgcodecs.imread(files.get(rand.nextInt(files.size() - 1)), 0);
Mat f = PlateUtil.features(img, _predictsize);
samples.push_back(f);
trainingLabels.add(i + Constant.numCharacter);
// 增加随机平移样本
samples.push_back(PlateUtil.features(randTranslate(img), _predictsize));
trainingLabels.add(i + Constant.numCharacter);
// 增加随机旋转样本
samples.push_back(PlateUtil.features(randRotate(img), _predictsize));
trainingLabels.add(i + Constant.numCharacter);
// 增加膨胀样本
/*samples.push_back(PlateUtil.features(dilate(img), _predictsize));
trainingLabels.add(i + Constant.numCharacter);*/
// 增加腐蚀样本
samples.push_back(PlateUtil.features(erode(img), _predictsize));
trainingLabels.add(i + Constant.numCharacter);
}
}
samples.convertTo(samples, CvType.CV_32F);
//440 vhist.length + hhist.length + lowData.cols() * lowData.rows();
// CV_32FC1 CV_32SC1 CV_32F
Mat classes = Mat.zeros(trainingLabels.size(), Constant.numAll, CvType.CV_32F);
Mat classes = Mat.zeros(trainingLabels.size(), Constant.strCharacters.length, CvType.CV_32F);
float[] labels = new float[trainingLabels.size()];
for (int i = 0; i < labels.length; ++i) {
@ -244,78 +129,66 @@ public class ANNTrain {
public void predict() {
ann.clear();
ann = ANN_MLP.load(MODEL_PATH);
Vector<String> files = new Vector<String>();
FileUtil.getFiles(DEFAULT_PATH + "test/", files); // 获取测试文件
String plate = "";
for (String string : files) {
Mat img = Imgcodecs.imread(string, 0);
Mat f = PlateUtil.features(img, Constant.predictSize);
int index = 0;
double maxVal = -2;
Mat output = new Mat(1, Constant.numAll, CvType.CV_32F);
ann.predict(f, output); // 预测结果
for (int j = 0; j < Constant.numAll; j++) {
double val = output.get(0, j)[0];
if (val > maxVal) {
maxVal = val;
index = j;
}
}
// 随机平移
/*f = PlateUtil.features(randTranslate(img), Constant.predictSize);
ann.predict(f, output); // 预测结果
for (int j = 0; j < Constant.numAll; j++) {
double val = output.get(0, j)[0];
if (val > maxVal) {
maxVal = val;
index = j;
}
}*/
// 随机旋转
/*f = PlateUtil.features(randRotate(img), Constant.predictSize);
ann.predict(f, output); // 预测结果
for (int j = 0; j < Constant.numAll; j++) {
double val = output.get(0, j)[0];
if (val > maxVal) {
maxVal = val;
index = j;
}
}*/
// 膨胀
/*f = PlateUtil.features(dilate(img), Constant.predictSize);
ann.predict(f, output); // 预测结果
for (int j = 0; j < Constant.numAll; j++) {
double val = output.get(0, j)[0];
if (val > maxVal) {
maxVal = val;
index = j;
int total = 0;
int correct = 0;
// 遍历测试样本下的所有文件，计算预测准确率
for (int i = 0; i < Constant.strCharacters.length; i++) {
char c = Constant.strCharacters[i];
String path = DEFAULT_PATH + "learn/" + c;
Vector<String> files = new Vector<String>();
FileUtil.getFiles(path, files);
for (String filePath : files) {
Mat img = Imgcodecs.imread(filePath, 0);
Mat f = PlateUtil.features(img, Constant.predictSize);
int index = 0;
double maxVal = -2;
Mat output = new Mat(1, Constant.strCharacters.length, CvType.CV_32F);
ann.predict(f, output); // 预测结果
for (int j = 0; j < Constant.strCharacters.length; j++) {
double val = output.get(0, j)[0];
if (val > maxVal) {
maxVal = val;
index = j;
}
}
}*/
// 腐蚀 -- 识别中文字符效果会好一点，识别数字及字母效果会更差
/*f = PlateUtil.features(erode(img), Constant.predictSize);
ann.predict(f, output); // 预测结果
for (int j = 0; j < Constant.numAll; j++) {
double val = output.get(0, j)[0];
if (val > maxVal) {
maxVal = val;
index = j;
// 膨胀
f = PlateUtil.features(PlateUtil.dilate(img), Constant.predictSize);
ann.predict(f, output); // 预测结果
for (int j = 0; j < Constant.strCharacters.length; j++) {
double val = output.get(0, j)[0];
if (val > maxVal) {
maxVal = val;
index = j;
}
}
}*/
if (index < Constant.numCharacter) {
plate += String.valueOf(Constant.strCharacters[index]);
} else {
String s = Constant.strChinese[index - Constant.numCharacter];
plate += Constant.KEY_CHINESE_MAP.get(s);
String result = String.valueOf(Constant.strCharacters[index]);
if(result.equals(String.valueOf(c))) {
correct++;
} else {
System.err.print(filePath);
System.err.println("\t预测结果：" + result);
}
total++;
}
}
System.err.println("===>" + plate);
System.out.print("total:" + total);
System.out.print("\tcorrect:" + correct);
System.out.print("\terror:" + (total - correct));
System.out.println("\t计算准确率为：" + correct / (total * 1.0));
//牛逼，我操 total:13178 correct:13139 error:39 计算准确率为：0.9970405220822584
return;
}

Write Preview
Loading…
Cancel
Save