Java OpenCV FileStorage 和 Mat.push_back
Posted
技术标签:
【中文标题】Java OpenCV FileStorage 和 Mat.push_back【英文标题】：Java OpenCV FileStorage and Mat.push_back 【发布时间】：2016-07-07 10:20:10 【问题描述】：我正在尝试用 Java 为 KNN 分类器实现 this 项目，即 GenData.cpp（用 C++ 编写）。 我已经到达这些代码行并卡住了：
matClassificationInts.push_back(intChar);
cv::FileStorage fsClassifications("classifications.xml", cv::FileStorage::WRITE);
fsClassifications << "classifications" << matClassificationInts;
fsClassifications.release();
在 c++ 中，我们可以将整数传递给 push_back()，但在 Java 中我收到错误：“int cannot be convert to Mat”。 所以，第一个问题是：如何将 int 传递给 someMat.push_back()？ 第二个：如何在 Java 中实现 FileStorage 或将 Mat 写入 *.xml 格式（并从 *.xml 读取 Mat）？
到目前为止，我的代码：
import java.io.IOException;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Scanner;
import org.opencv.core.Core;
import static org.opencv.core.CvType.CV_32FC1;
import org.opencv.core.Mat;
import org.opencv.core.MatOfInt4;
import org.opencv.core.MatOfPoint;
import org.opencv.core.Rect;
import org.opencv.core.Scalar;
import org.opencv.core.Size;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import static org.opencv.imgproc.Imgproc.ADAPTIVE_THRESH_GAUSSIAN_C;
import static org.opencv.imgproc.Imgproc.CHAIN_APPROX_SIMPLE;
import static org.opencv.imgproc.Imgproc.RETR_EXTERNAL;
import static org.opencv.imgproc.Imgproc.THRESH_BINARY_INV;
public class genData
private static final int
MIN_CONTOUR_AREA = 100,
RESIZED_IMAGE_WIDTH = 20,
RESIZED_IMAGE_HEIGHT = 30;
public static void main(String[] args) throws IOException
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
Scanner keyboard = new Scanner(System.in);
boolean exit = false;
Mat imgTrainingNumbers;
Mat imgGrayscale = new Mat();
Mat imgBlurred = new Mat();
Mat imgThresh = new Mat();
Mat imgThreshCopy = new Mat();
ArrayList<MatOfPoint> ptContours = new ArrayList<MatOfPoint>();
MatOfInt4 v4iHierarchy;
Mat matClassificationInts = new Mat();
Mat matTrainingImagesAsFlattenedFloats = new Mat();
int[] intValidChars = '0', '1', '2',
'A', 'B', 'C'; //Here I did not make List<Integer>, because I can't pass char to Integer.
Arrays.sort(intValidChars); //for binary search
imgTrainingNumbers = Imgcodecs.imread("test.png"); //here Text on white paper.
if (imgTrainingNumbers.empty())
System.out.println("err");
return;
Imgproc.cvtColor(imgTrainingNumbers, imgGrayscale, Imgproc.COLOR_BGR2GRAY);
Imgproc.GaussianBlur(imgGrayscale, imgBlurred, new Size(5, 5), 0);
Imgproc.adaptiveThreshold(imgBlurred, imgThresh, 255, ADAPTIVE_THRESH_GAUSSIAN_C, THRESH_BINARY_INV, 11, 2);
/*
//imshow class implementation (found via google, works properly, but this block is commented for now)
Imshow im = new Imshow("imgThresh");
im.showImage(imgThresh);
imgThreshCopy = imgThresh.clone();
*/
Imgproc.findContours(imgThreshCopy, ptContours, new Mat(), RETR_EXTERNAL, CHAIN_APPROX_SIMPLE);
for (int i = 0; i < ptContours.size(); i++)
if (Imgproc.contourArea(ptContours.get(i)) > MIN_CONTOUR_AREA)
Rect boundingRect = Imgproc.boundingRect(ptContours.get(i));
Imgproc.rectangle(imgTrainingNumbers, boundingRect.tl(), boundingRect.br(), new Scalar(0, 0, 255), 2);
Mat matROI = imgThresh.submat(boundingRect.y, boundingRect.y + boundingRect.height, boundingRect.x, boundingRect.x + boundingRect.width);
Mat matROIResized = new Mat();
Imgproc.resize(matROI, matROIResized, new Size(RESIZED_IMAGE_WIDTH, RESIZED_IMAGE_HEIGHT));
/*
im.showImage(matROI);
im.showImage(matROIResized);
im.showImage(imgTrainingNumbers);
*/
String input = keyboard.nextLine();
int intChar = (int)input.charAt(0);
if (Arrays.binarySearch(intValidChars, intChar) >=0)
/*
matClassificationInts.push_back(intChar);
//Here I'm getting an error.
*/
Mat matImageFloat = new Mat();
matROIResized.convertTo(matImageFloat, CV_32FC1);
Mat matImageFlattenedFloat = matImageFloat.reshape(1, 1);
matTrainingImagesAsFlattenedFloats.push_back(matImageFlattenedFloat);
//Here should go FileStorage stuff.
提前致谢。 附：使用 OpenCV_310 + Java（不是 JavaCV）
【问题讨论】：
【参考方案1】：好吧，我已经自己解决了。 很脏，我猜，但就是这样。如果您知道如何对我的代码进行良好的改进，我会很高兴阅读您的 cmets。 1）我的第一个问题是关于将 int 放入 Mat （用于进一步制作 *.xml）。我避免了这种方法，并决定将 int（实际上是 Integer）放入 List。
Scanner keyboard = new Scanner(System.in);
String input = keyboard.nextLine();
int intChar = (int)input.charAt(0);
List<Integer> matClassificationInts = new ArrayList<Integer>();
if (Arrays.binarySearch(intValidChars, intChar) >=0)
matClassificationInts.add(new Integer(intChar));
......
String dataImages = "";
for (Integer i : matClassificationInts)
dataImages += i + " ";
我可以制作字符串（例如“49 48”字符“1 0”整数）以将其保存在 *.xml 中（参见下一段）。 2) 第二个问题是关于从 Mat 中提取数据并将其存储在 *.xml 中。好吧，通过 C++，我可以通过 FileStorage 来实现：
cv::FileStorage fsClassifications("classifications.xml", cv::FileStorage::WRITE);
fsClassifications << "classifications" << matClassificationInts;
fsClassifications.release();
但是 Java OpenCV 没有这样的功能，所以我循环遍历二维数组（Mat.rows() 和 Mat.cols()）并通过 get() 方法提取所需的数据（Mat.get(row, col) -给出双精度数组，数组长度 = 1)：
String dataClassifications = "";
for (int i = 0; i < matTrainingImagesAsFlattenedFloats.rows(); i++)
for (int j = 0; j < matTrainingImagesAsFlattenedFloats.cols(); j++)
double[] temp = matTrainingImagesAsFlattenedFloats.get(i, j);
dataClassifications += temp[0] + " ";
dataClassifications += "\n";
现在，关于将数据保存到 *.xml： 我刚刚使用了 javafx.xml 和 org.wc3.dom 库。 制作了两个用于返回 DOM 节点的函数：
private static Node getMatXML(Document doc, String option_id, String type_id, String rows, String cols, String dt, String data)
Element elem = doc.createElement(option_id);
elem.setAttribute("type_id", type_id);
elem.appendChild(getMatXMLElement(doc,"rows", rows));
elem.appendChild(getMatXMLElement(doc, "cols", cols));
elem.appendChild(getMatXMLElement(doc, "dt", dt));
elem.appendChild(getMatXMLElement(doc, "data", data));
return elem;
private static Node getMatXMLElement(Document doc, String name, String value)
Element node = doc.createElement(name);
node.appendChild(doc.createTextNode(value));
return node;
并使用这些函数来创建 *.xml: 分类.xml：
DocumentBuilderFactory icFactory_images = DocumentBuilderFactory.newInstance();
DocumentBuilder icBuilder_images;
try
icBuilder_images = icFactory_images.newDocumentBuilder();
Document doc = icBuilder_images.newDocument();
Element mainRootElement = doc.createElement("opencv_storage");
doc.appendChild(mainRootElement);
mainRootElement.appendChild(getMatXML(doc, "classifications", "opencv-matrix", rowsImages, colsImages, "i", dataImages));
Transformer transformer = TransformerFactory.newInstance().newTransformer();
transformer.setOutputProperty(OutputKeys.INDENT, "yes");
DOMSource source = new DOMSource(doc);
String filename = "classifications.xml";
File file = new File(filename);
StreamResult console = new StreamResult(file); //(System.out)
transformer.transform(source, console);
catch (Exception e)
e.printStackTrace();
Images.xml：
DocumentBuilderFactory icFactory_classifications = DocumentBuilderFactory.newInstance();
DocumentBuilder icBuilder_classifications;
try
icBuilder_classifications = icFactory_classifications.newDocumentBuilder();
Document doc = icBuilder_classifications.newDocument();
Element mainRootElement = doc.createElement("opencv_storage");
doc.appendChild(mainRootElement);
mainRootElement.appendChild(getMatXML(doc, "images", "opencv-matrix", rowsClassifications, colsClassifications, "f", dataClassifications));
Transformer transformer = TransformerFactory.newInstance().newTransformer();
transformer.setOutputProperty(OutputKeys.INDENT, "yes");
DOMSource source = new DOMSource(doc);
String filename = "images.xml";
File file = new File(filename);
StreamResult console = new StreamResult(file); //(System.out)
transformer.transform(source, console);
catch (Exception e)
e.printStackTrace();
例如，生成的分类文件是：
<?xml version="1.0" encoding="UTF-8" standalone="no"?>
<opencv_storage>
<classifications type_id="opencv-matrix">
<rows>2</rows>
<cols>1</cols>
<dt>i</dt>
<data>49 48 </data>
</classifications>
</opencv_storage>
我为这张照片做了测试： 通过 GenData.cpp（参见相关链接 - 第 1 行）和我的 Java 代码（完整代码见下文）。两个程序都给了我相同的结果：对于 Java OpenCV Imshow 实现，您可以查看 this link（不是我的）。
import java.io.File;
import java.io.FileNotFoundException;
import java.io.FileOutputStream;
import java.io.FileReader;
import java.io.IOException;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.List;
import java.util.Map;
import java.util.Scanner;
import org.opencv.core.Core;
import static org.opencv.core.CvType.CV_32FC1;
import org.opencv.core.Mat;
import org.opencv.core.MatOfInt4;
import org.opencv.core.MatOfPoint;
import org.opencv.core.Rect;
import org.opencv.core.Scalar;
import org.opencv.core.Size;
import org.opencv.imgcodecs.Imgcodecs;
import org.opencv.imgproc.Imgproc;
import static org.opencv.imgproc.Imgproc.ADAPTIVE_THRESH_GAUSSIAN_C;
import static org.opencv.imgproc.Imgproc.CHAIN_APPROX_SIMPLE;
import static org.opencv.imgproc.Imgproc.RETR_EXTERNAL;
import static org.opencv.imgproc.Imgproc.THRESH_BINARY_INV;
//XML - write.
import javax.xml.parsers.DocumentBuilder;
import javax.xml.parsers.DocumentBuilderFactory;
import javax.xml.transform.OutputKeys;
import javax.xml.transform.Transformer;
import javax.xml.transform.TransformerFactory;
import javax.xml.transform.dom.DOMSource;
import javax.xml.transform.stream.StreamResult;
import org.w3c.dom.Document;
import org.w3c.dom.Element;
import org.w3c.dom.Node;
public class genData
private static final int
MIN_CONTOUR_AREA = 100,
RESIZED_IMAGE_WIDTH = 20,
RESIZED_IMAGE_HEIGHT = 30;
public static void main(String[] args) throws IOException
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
Scanner keyboard = new Scanner(System.in);
Mat imgTrainingNumbers;
Mat imgGrayscale = new Mat();
Mat imgBlurred = new Mat();
Mat imgThresh = new Mat();
Mat imgThreshCopy = new Mat();
ArrayList<MatOfPoint> ptContours = new ArrayList<MatOfPoint>();
MatOfInt4 v4iHierarchy = new MatOfInt4();
List<Integer> matClassificationInts = new ArrayList<Integer>();
Mat matTrainingImagesAsFlattenedFloats = new Mat();
int[] intValidChars = '0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
'A', 'B', 'C', 'E', 'H',
'K', 'M', 'O', 'P', 'T',
'X', 'Y';
Arrays.sort(intValidChars);
imgTrainingNumbers = Imgcodecs.imread("01.png");
if (imgTrainingNumbers.empty())
System.out.println("Error: file is not found");
return;
Imgproc.cvtColor(imgTrainingNumbers, imgGrayscale, Imgproc.COLOR_BGR2GRAY);
Imgproc.GaussianBlur(imgGrayscale, imgBlurred, new Size(5, 5), 0);
Imgproc.adaptiveThreshold(imgBlurred, imgThresh, 255, ADAPTIVE_THRESH_GAUSSIAN_C, THRESH_BINARY_INV, 11, 2);
Imshow im = new Imshow("imgThresh");
im.showImage(imgThresh);
imgThreshCopy = imgThresh.clone();
Imgproc.findContours(imgThreshCopy, ptContours, v4iHierarchy, RETR_EXTERNAL, CHAIN_APPROX_SIMPLE);
for (int i = 0; i < ptContours.size(); i++)
if (Imgproc.contourArea(ptContours.get(i)) > MIN_CONTOUR_AREA)
Rect boundingRect = Imgproc.boundingRect(ptContours.get(i));
Imgproc.rectangle(imgTrainingNumbers, boundingRect.tl(), boundingRect.br(), new Scalar(0, 0, 255), 2);
Mat matROI = imgThresh.submat(boundingRect.y, boundingRect.y + boundingRect.height, boundingRect.x, boundingRect.x + boundingRect.width);
Mat matROIResized = new Mat();
Imgproc.resize(matROI, matROIResized, new Size(RESIZED_IMAGE_WIDTH, RESIZED_IMAGE_HEIGHT));
im.showImage(matROI);
im.showImage(matROIResized);
im.showImage(imgTrainingNumbers);
String input = keyboard.nextLine();
int intChar = (int)input.charAt(0);
if (Arrays.binarySearch(intValidChars, intChar) >=0)
matClassificationInts.add(new Integer(intChar));
Mat matImageFloat = new Mat();
matROIResized.convertTo(matImageFloat, CV_32FC1);
Mat matImageFlattenedFloat = matImageFloat.reshape(1, 1);
matTrainingImagesAsFlattenedFloats.push_back(matImageFlattenedFloat);
String dataImages = "";
for (Integer i : matClassificationInts)
dataImages += i + " ";
String dataClassifications = "";
for (int i = 0; i < matTrainingImagesAsFlattenedFloats.rows(); i++)
for (int j = 0; j < matTrainingImagesAsFlattenedFloats.cols(); j++)
double[] temp = matTrainingImagesAsFlattenedFloats.get(i, j);
dataClassifications += temp[0] + " ";
dataClassifications += "\n";
String rowsImages = String.valueOf(matClassificationInts.size());
String colsImages = "1";
String rowsClassifications = String.valueOf(matTrainingImagesAsFlattenedFloats.rows());
String colsClassifications = String.valueOf(matTrainingImagesAsFlattenedFloats.cols());
DocumentBuilderFactory icFactory_images = DocumentBuilderFactory.newInstance();
DocumentBuilder icBuilder_images;
try
icBuilder_images = icFactory_images.newDocumentBuilder();
Document doc = icBuilder_images.newDocument();
Element mainRootElement = doc.createElement("opencv_storage");
doc.appendChild(mainRootElement);
mainRootElement.appendChild(getMatXML(doc, "classifications", "opencv-matrix", rowsImages, colsImages, "i", dataImages));
Transformer transformer = TransformerFactory.newInstance().newTransformer();
transformer.setOutputProperty(OutputKeys.INDENT, "yes");
DOMSource source = new DOMSource(doc);
String filename = "classifications.xml";
File file = new File(filename);
StreamResult console = new StreamResult(file); //(System.out)
transformer.transform(source, console);
catch (Exception e)
e.printStackTrace();
DocumentBuilderFactory icFactory_classifications = DocumentBuilderFactory.newInstance();
DocumentBuilder icBuilder_classifications;
try
icBuilder_classifications = icFactory_classifications.newDocumentBuilder();
Document doc = icBuilder_classifications.newDocument();
Element mainRootElement = doc.createElement("opencv_storage");
doc.appendChild(mainRootElement);
mainRootElement.appendChild(getMatXML(doc, "images", "opencv-matrix", rowsClassifications, colsClassifications, "f", dataClassifications));
Transformer transformer = TransformerFactory.newInstance().newTransformer();
transformer.setOutputProperty(OutputKeys.INDENT, "yes");
DOMSource source = new DOMSource(doc);
String filename = "images.xml";
File file = new File(filename);
StreamResult console = new StreamResult(file); //(System.out)
transformer.transform(source, console);
catch (Exception e)
e.printStackTrace();
System.out.println("Finished.");
System.exit(0);
private static Node getMatXML(Document doc, String option_id, String type_id, String rows, String cols, String dt, String data)
Element elem = doc.createElement(option_id);
elem.setAttribute("type_id", type_id);
elem.appendChild(getMatXMLElement(doc,"rows", rows));
elem.appendChild(getMatXMLElement(doc, "cols", cols));
elem.appendChild(getMatXMLElement(doc, "dt", dt));
elem.appendChild(getMatXMLElement(doc, "data", data));
return elem;
private static Node getMatXMLElement(Document doc, String name, String value)
Element node = doc.createElement(name);
node.appendChild(doc.createTextNode(value));
return node;
【讨论】：
以上是关于Java OpenCV FileStorage 和 Mat.push_back的主要内容，如果未能解决你的问题，请参考以下文章
利用opencv中的类FileStorage生成和读取XML和YAML文件