我正在尝试使用官方OpenCV Java绑定以YAML格式加载/保存OpenCV校准数据。我知道OpenCV(至少为c
++版本)可以序列化为XML和JSON,但是我想支持较早的YAML校准文件。
校准文件如下所示:
%YAML:1.0
cameraMatrix: !!opencv-matrix
rows: 3
cols: 3
dt: d
data: [ 6.6278599887122368e+02, 0., 3.1244256016006659e+02, 0.,
6.6129276875199082e+02, 2.2747179767124251e+02, 0., 0., 1. ]
imageSize_width: 640
imageSize_height: 480
sensorSize_width: 0
sensorSize_height: 0
distCoeffs: !!opencv-matrix
rows: 5
cols: 1
dt: d
data: [ -1.8848338341464690e-01, 1.0721890419183855e+00,
-3.5244467228016116e-03, -7.0195032848241403e-04,
-2.0412827999027101e+00 ]
reprojectionError: 2.1723265945911407e-01
我正在寻找一个优雅的解决方案,因为我不太了解如何将Java类最好地映射到YAML并返回。我尝试了一些库,例如jyaml,yamlbeans(来自SourceForge的1.0和来自MavenCentral的1.13)和SnakeYAML。
我目前尝试对一些作品进行反序列化,但感觉很棘手:
CalibrationParseTest.java
import java.io.IOException;
import java.nio.charset.StandardCharsets;
import java.nio.file.Files;
import java.nio.file.Paths;
import org.opencv.core.Core;
import org.yaml.snakeyaml.Yaml;
import org.yaml.snakeyaml.constructor.Constructor;
public class CalibrationParseTest {
public static void main(String[] args) {
// load OpenCV native
System.loadLibrary(Core.NATIVE_LIBRARY_NAME);
String yamlPath = "./data/calibration.yml";
try{
String yamlString = new String(Files.readAllBytes(Paths.get(yamlPath)), StandardCharsets.UTF_8);
// remove %YAML:1.0 to avoid scan directive error
yamlString = yamlString.replaceAll("%YAML:1.0", "");
// map custom class
yamlString = yamlString.replaceAll("opencv-matrix", "MatYAML");
System.out.println("<loaded>");
System.out.println(yamlString);
System.out.println("</loaded>");
Yaml yaml = new Yaml(new Constructor(CalibrationData.class));
CalibrationData data = yaml.load(yamlString);
// currently manually parsing data from the HashMap: can this be better ?
data.populateCV();
// double check data
System.out.println("<deserialized>");
System.out.println(data);
System.out.println("</deserialized>");
}catch (IOException e) {
e.printStackTrace();
}
}
}
CalibrationData.java
import java.util.HashMap;
import org.opencv.core.Mat;
import org.opencv.core.Size;
public class CalibrationData extends HashMap{
public Mat cameraMatrix;
public Size imageSize;
public Size sensorSize;
public Mat distCoeffs;
public float reprojectionError;
public CalibrationData(){}
public void populateCV(){
cameraMatrix = ((MatYAML)get("cameraMatrix")).toMat();
imageSize = new Size((int)get("imageSize_width"),(int)get("imageSize_height"));
sensorSize = new Size((int)get("sensorSize_width"),(int)get("sensorSize_height"));
distCoeffs = ((MatYAML)get("distCoeffs")).toMat();
reprojectionError = (float)((double)get("reprojectionError"));
}
public String toString(){
if(cameraMatrix == null){
return String.format("[CalibrationData (not parsed to CV-> call populateCV()\n\tdata: %s\n]",super.toString());
}
return String.format("[CalibrationData\n" +
"\tcalibrationMatrix: %s\n" +
"\timageSize: %s\n" +
"\tsensorSize: %s\n" +
"\tdistCoeffs: %s\n" +
"\treprojectionError: %f\n]", cameraMatrix.dump(), imageSize.toString(), sensorSize.toString(), distCoeffs.dump(), reprojectionError);
}
}
MatYAML.java
import java.util.List;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
public class MatYAML{
public int rows;
public int cols;
public String dt;
public List<Double> data;
Mat toMat(){
Mat out = new Mat(rows, cols, dt.equals("d") ? CvType.CV_64F : CvType.CV_32F);
int index = 0;
for (int row = 0; row < rows; row++) {
for (int col = 0; col < cols; col++) {
out.put(row, col, data.get(index++));
}
}
return out;
}
}
这将输出预期结果:
<loaded>
cameraMatrix: !!MatYAML
rows: 3
cols: 3
dt: d
data: [ 6.6278599887122368e+02, 0., 3.1244256016006659e+02, 0.,
6.6129276875199082e+02, 2.2747179767124251e+02, 0., 0., 1. ]
imageSize_width: 640
imageSize_height: 480
sensorSize_width: 0
sensorSize_height: 0
distCoeffs: !!MatYAML
rows: 5
cols: 1
dt: d
data: [ -1.8848338341464690e-01, 1.0721890419183855e+00,
-3.5244467228016116e-03, -7.0195032848241403e-04,
-2.0412827999027101e+00 ]
reprojectionError: 2.1723265945911407e-01
</loaded>
<deserialized>
[CalibrationData
calibrationMatrix: [662.7859988712237, 0, 312.4425601600666;
0, 661.2927687519908, 227.4717976712425;
0, 0, 1]
imageSize: 640x480
sensorSize: 0x0
distCoeffs: [-0.1884833834146469; 1.072189041918385; -0.003524446722801612; -0.000701950328482414; -2.04128279990271]
reprojectionError: 0.217233
]
</deserialized>
没有这些技巧,是否有更优雅的方法在Java OpenCV类和YAML之间进行序列化/反序列化?
骇客是指:
Mat
数据?(如果可能的话 ?)Update 2 amanin的答案更干净,可以避免过分地替换“ !! opencv-matrix”,但是它不会序列化/反序列化Mat
:
OpenCVConfig{imageSize_width=640, imageSize_height=480, sensorSize_width=0, sensorSize_height=0, camerMatrix=Matrix{rows=3, cols=3, dt=d, data=[662.7859988712237, 0.0, 312.4425601600666, 0.0, 661.2927687519908, 227.4717976712425, 0.0, 0.0, 1.0]}, distCoeffs=Matrix{rows=5, cols=1, dt=d, data=[-0.1884833834146469, 1.0721890419183855, -0.0035244467228016116, -7.01950328482414E-4, -2.04128279990271]}}
---
imageSize_width: 640
imageSize_height: 480
sensorSize_width: 0
sensorSize_height: 0
reprojectionError: 0.21723265945911407
cameraMatrix:
rows: 3
cols: 3
dt: "d"
data:
- 662.7859988712237
- 0.0
- 312.4425601600666
- 0.0
- 661.2927687519908
- 227.4717976712425
- 0.0
- 0.0
- 1.0
distCoeffs:
rows: 5
cols: 1
dt: "d"
data:
- -0.1884833834146469
- 1.0721890419183855
- -0.0035244467228016116
- -7.01950328482414E-4
- -2.04128279990271
请建议将解决方案与 org.opencv.core.Mat
你看过杰克逊图书馆吗?它允许将JSON / Yaml内容映射到Java POJO。
我举了一个小例子来解决您的两个问题:
但是,对于yaml版本指令,由于它似乎无效,因此我不确定如何处理它。在我的示例中,我已预先手动将其删除。当然,可以找到更好的解决方案,但我不知道。
EDIT2 :对于矩阵对象,我做了一个笨拙的POJO,供Jackson用来内部读取brut YAML。然后,我添加了一个转换层(见
@JsonSerialize 和 @JsonDeserialize 上anotations OpenCVConfig
类)这个简单的POJO转换为专业OpenCV的矩阵。Jackson提供了多种映射技术(流技术,自定义转换器/解串器,指导注释等),因此您可以探索其功能,以找到最适合您需要的解决方案。
因此,要使该示例正常工作,您将需要两个依赖项(以maven格式给出):
<!-- (De)serialization engine -->
<dependency>
<groupId>com.fasterxml.jackson.core</groupId>
<artifactId>jackson-databind</artifactId>
<version>2.10.0</version>
</dependency>
<!-- Yaml support -->
<dependency>
<groupId>com.fasterxml.jackson.dataformat</groupId>
<artifactId>jackson-dataformat-yaml</artifactId>
<version>2.10.0</version>
</dependency>
这是Maven存储库描述页面:
Jackson-databind
:通过Maven搜索或通过mvnrepository
Jackson-dataformat-yaml
:通过Maven搜索或通过mvnrepository
注意 :我的示例包含很多样板,就像我手动(以及IDE)生成的getter / setter一样。您应该通过以下两种方式减少代码量:
使用Java 14记录(不确定当前是否可用)
package fr.amanin.stackoverflow;
import java.util.Arrays;
import java.util.stream.Stream;
import com.fasterxml.jackson.core.type.TypeReference;
import com.fasterxml.jackson.databind.JavaType;
import com.fasterxml.jackson.databind.ObjectMapper;
import com.fasterxml.jackson.databind.annotation.JsonDeserialize;
import com.fasterxml.jackson.databind.annotation.JsonSerialize;
import com.fasterxml.jackson.databind.type.TypeFactory;
import com.fasterxml.jackson.databind.util.Converter;
import com.fasterxml.jackson.dataformat.yaml.YAMLGenerator;
import com.fasterxml.jackson.dataformat.yaml.YAMLMapper;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
public class YAMLOpenCV {
/*
* Engine in charge of YAML decoding.
/
public static final ObjectMapper OPENCV_YAML_MAPPER = new YAMLMapper();
public static void main(String[] args) throws Exception {
nu.pattern.OpenCV.loadShared();
System.loadLibrary(org.opencv.core.Core.NATIVE_LIBRARY_NAME);
final String confStr =
"cameraMatrix: !!opencv-matrix\n" +
" rows: 3\n" +
" cols: 3\n" +
" dt: d\n" +
" data: [ 6.6278599887122368e+02, 0., 3.1244256016006659e+02, 0.,\n" +
" 6.6129276875199082e+02, 2.2747179767124251e+02, 0., 0., 1. ]\n" +
"imageSize_width: 640\n" +
"imageSize_height: 480\n" +
"sensorSize_width: 0\n" +
"sensorSize_height: 0\n" +
"distCoeffs: !!opencv-matrix\n" +
" rows: 5\n" +
" cols: 1\n" +
" dt: d\n" +
" data: [ -1.8848338341464690e-01, 1.0721890419183855e+00,\n" +
" -3.5244467228016116e-03, -7.0195032848241403e-04,\n" +
" -2.0412827999027101e+00 ]\n" +
"reprojectionError: 2.1723265945911407e-01";
OpenCVConfig conf = OPENCV_YAML_MAPPER.readValue(confStr, OpenCVConfig.class);
System.out.println(conf);
String serialized = OPENCV_YAML_MAPPER.writeValueAsString(conf);
System.out.println(serialized);
}
/**
* Java model mirroring YAML configuration. Jackson will fill it
* with values read from YAML configuration file, matching YAML
* fields with this class property names.
*/
public static class OpenCVConfig {
int imageSize_width;
int imageSize_height;
int sensorSize_width;
int sensorSize_height;
double reprojectionError;
/* Special case: Matrix objects are decoded in two passes:
* 1. Jackson will check below converters, and use their input
* type to decode YAML to `Matrix` object (intermediate step).
* 2. Jackson uses converter to delegate to user the mapping
* from this intermediate POJO to specialized target (`Mat` here)
*/
@JsonDeserialize(converter = ToMatConverter.class)
@JsonSerialize(converter = FromMatConverter.class)
Mat cameraMatrix;
@JsonDeserialize(converter = ToMatConverter.class)
@JsonSerialize(converter = FromMatConverter.class)
Mat distCoeffs;
public int getImageSize_width() {
return imageSize_width;
}
public OpenCVConfig setImageSize_width(int imageSize_width) {
this.imageSize_width = imageSize_width;
return this;
}
public int getImageSize_height() {
return imageSize_height;
}
public OpenCVConfig setImageSize_height(int imageSize_height) {
this.imageSize_height = imageSize_height;
return this;
}
public int getSensorSize_width() {
return sensorSize_width;
}
public OpenCVConfig setSensorSize_width(int sensorSize_width) {
this.sensorSize_width = sensorSize_width;
return this;
}
public int getSensorSize_height() {
return sensorSize_height;
}
public OpenCVConfig setSensorSize_height(int sensorSize_height) {
this.sensorSize_height = sensorSize_height;
return this;
}
public double getReprojectionError() {
return reprojectionError;
}
public OpenCVConfig setReprojectionError(double reprojectionError) {
this.reprojectionError = reprojectionError;
return this;
}
public Mat getCameraMatrix() {
return cameraMatrix;
}
public OpenCVConfig setCameraMatrix(Mat cameraMatrix) {
this.cameraMatrix = cameraMatrix;
return this;
}
public Mat getDistCoeffs() {
return distCoeffs;
}
public OpenCVConfig setDistCoeffs(Mat distCoeffs) {
this.distCoeffs = distCoeffs;
return this;
}
@Override
public String toString() {
return "OpenCVConfig{" +
"imageSize_width=" + imageSize_width +
", imageSize_height=" + imageSize_height +
", sensorSize_width=" + sensorSize_width +
", sensorSize_height=" + sensorSize_height +
", camerMatrix=" + cameraMatrix +
", distCoeffs=" + distCoeffs +
'}';
}
}
/**
* Converter used for serialization of Mat objects into YAML.
*/
private static class FromMatConverter implements Converter<Mat, Matrix> {
@Override
public Matrix convert(Mat value) {
final Matrix result = new Matrix();
result.cols = value.cols();
result.rows = value.rows();
final int type = value.type();
result.dt = Stream.of(MatrixDataType.values())
.filter(dt -> dt.mapping == type)
.findAny()
.orElseThrow(() -> new UnsupportedOperationException("No matching datatype found for "+type));
int idx = 0;
result.data = new double[result.rows * result.cols];
for (int r = 0 ; r < result.rows ; r++) {
for (int c = 0; c < result.cols; c++) {
final double[] v = value.get(r, c);
result.data[idx++] = v[0];
}
}
return result;
}
@Override
public JavaType getInputType(TypeFactory typeFactory) {
return typeFactory.constructType(new TypeReference<Mat>() {});
}
@Override
public JavaType getOutputType(TypeFactory typeFactory) {
return typeFactory.constructType(new TypeReference<Matrix>() {});
}
}
/**
* Converter used at read time, to map YAML object to OpenCV Mat.
*/
private static class ToMatConverter implements Converter<Matrix, Mat> {
@Override
public Mat convert(Matrix in) {
final Mat result = new Mat(in.rows, in.cols, in.dt.mapping);
int idx = 0;
for (int r = 0 ; r < in.rows ; r++) {
for (int c = 0; c < in.cols; c++) {
result.put(r, c, in.data[idx++]);
}
}
return result;
}
@Override
public JavaType getInputType(TypeFactory typeFactory) {
return typeFactory.constructType(new TypeReference<Matrix>() {});
}
@Override
public JavaType getOutputType(TypeFactory typeFactory) {
return typeFactory.constructType(new TypeReference<Mat>() {});
}
}
public static class Matrix {
int rows;
int cols;
MatrixDataType dt;
double[] data;
public int getRows() {
return rows;
}
public Matrix setRows(int rows) {
this.rows = rows;
return this;
}
public int getCols() {
return cols;
}
public Matrix setCols(int cols) {
this.cols = cols;
return this;
}
public MatrixDataType getDt() {
return dt;
}
public Matrix setDt(MatrixDataType dt) {
this.dt = dt;
return this;
}
public double[] getData() {
return data;
}
public Matrix setData(double[] data) {
this.data = data;
return this;
}
double at(int x, int y) {
if (x >= cols || y >= rows) throw new IllegalArgumentException("Bad coordinate");
return data[y*rows + x];
}
@Override
public String toString() {
return "Matrix{" +
"rows=" + rows +
", cols=" + cols +
", dt=" + dt +
", data=" + Arrays.toString(data) +
'}';
}
}
/*
public static class MatDeserializer extends StdDeserializer
protected MatDeserializer() {
super(Mat.class);
}
@Override
public Mat deserialize(JsonParser p, DeserializationContext ctxt) throws IOException, JsonProcessingException {
final int rows, cols;
final MatrixDataType dtype;
final double[] data;
}
}
public static class MatSerializer extends StdSerializer<Mat> {
protected MatSerializer() {
super(Mat.class);
}
@Override
public void serialize(Mat value, JsonGenerator gen, SerializerProvider provider) throws IOException {
gen.writeNumberField("rows", value.rows());
gen.writeNumberField("cols", value.cols());
gen.writeFieldName("data");
gen.writeStartArray();
gen.writeEndArray();
}
}
*/
public enum MatrixDataType {
d(CvType.CV_64F),
f(CvType.CV_32F);
public final int mapping;
MatrixDataType(int mapping) {
this.mapping = mapping;
}
}
}
希望能帮助到你,
我有一个kdtree,其节点由以下字段组成:公共静态类节点实现可序列化{ 其中DataPoint定义: 公共静态类DataPoint实现可序列化{公共可比X;公共可比Y;公共可比Z; 我想序列化树,存储在文件中并在回答范围查询时反序列化。我对这个概念od序列化的理解并不好。从我收集的任何内容中,我编写了以下函数,但不起作用。有人能帮忙吗?
问题内容: 我的直觉告诉我,必须以某种方式将其转换为字符串或byte [](在Go中甚至可能是相同的东西?),然后将其保存到磁盘。 我找到了这个包(http://golang.org/pkg/encoding/gob/),但似乎仅用于结构? 问题答案: 序列化数据有多种方法,Go为此提供了许多软件包。某些常见编码方式的软件包: 处理地图很好。以下示例显示了地图的编码/解码: 操场
主要内容:1 Java序列化和反序列化,2 Java序列化的优点,3 java.io.Serializable接口,4 Java ObjectOutputStream,5 Java ObjectInputStream,6 Java序列化的例子,7 Java反序列化的例子1 Java序列化和反序列化 Java中的序列化是一种将对象状态写入字节流的机制。它主要用于Hibernate,RMI,JPA,EJB和JMS技术。 序列化的反向操作称为反序列化,其中字节流被转换为对象。序列化和反序列化过程与平台
问题内容: 我已经开始在我的第一个android应用程序上进行工作,并且具有处理多层图像的应用程序。我能够将项目文件的平面版本导出为PNG,但我希望能够保存分层图像以供以后编辑(包括应用于某些层的任何选项,例如基于文本的层)。 无论如何,我已经确保需要写入文件的类是“可序列化的”,但是由于android.graphics.Bitmap不可序列化这一事实而遇到了一些障碍。以下代码实质上将位图作为PN
我有两个Java应用程序-和。需要通过套接字发送到类的Slave实例。 Master创建这个类的实例,序列化它并通过套接字发送到Slave。 在上,一切正常。没有异常。接收数据并尝试对其进行反序列化。引发以下异常 JAVAlang.ClassNotFoundException 在类中没有错误,因为如果我用优化函数=null创建它,那么就会毫无问题地反序列化它。我试图将实例序列化到中的文件中,然后也
如题目所述,想请教一下大佬们在Java中序列化与反序列化的意义是什么,如何理解Java的序列化和反序列化?