/// <summary>
/// job complete
/// </summary>
/// <param name="taskName"></param>
/// <param name="outputs"></param>
private void Job_OnTaskComplete(string taskName, params object[] outputs)
{
switch (taskName)
{
//load image classification result
case "COVRasterTask":
case "RFClassificationTask":
case "CnnClassificationTask":
case "DqnClassificationTask":
string fullFilename = outputs[0] as string;
ReadRaster(fullFilename);
break;
//load image
case "ReadRasterTask":
string nodeName = outputs[0] as string;
Dictionary <string, Bitmap2> dict = outputs[1] as Dictionary <string, Bitmap2>;
GRasterLayer rasterLayer = outputs[2] as GRasterLayer;
UpdateReadRasterUI(nodeName, dict, rasterLayer);
break;
// rpc rester rectify
case "RPCRasterRectifyTask":
break;
default:
break;
}
//print completed information
string msg = string.Format("time:{0},task:{1} completed", Now, taskName);
Invoke(new UpdateMapListBoxHandler(UpdateMapListBox), msg);
}
public void ClassificationByDQN()
{
double _loss = 1.0;
//
GRasterLayer featureLayer = new GRasterLayer(featureFullFilename);
GRasterLayer labelLayer = new GRasterLayer(trainFullFilename);
//create environment for agent exploring
IEnv env = new ImageClassifyEnv(featureLayer, labelLayer);
//create dqn alogrithm
DQN dqn = new DQN(env);
//in order to do this quickly, we set training epochs equals 10.
//please do not use so few training steps in actual use.
dqn.SetParameters(10, 0);
//register event to get information while training
dqn.OnLearningLossEventHandler += (double loss, double totalReward, double accuracy, double progress, string epochesTime) => { _loss = loss; };
//start dqn alogrithm learning
dqn.Learn();
//in general, loss is less than 1
Assert.IsTrue(_loss < 1.0);
//apply dqn to classify fetureLayer
//pick value
IRasterLayerCursorTool pRasterLayerCursorTool = new GRasterLayerCursorTool();
pRasterLayerCursorTool.Visit(featureLayer);
//
double[] state = pRasterLayerCursorTool.PickNormalValue(50, 50);
double[] action = dqn.ChooseAction(state).action;
int landCoverType = dqn.ActionToRawValue(NP.Argmax(action));
//do something as you need. i.e. draw landCoverType to bitmap at position ( i , j )
//the classification results are not stable because of the training epochs are too few.
Assert.IsTrue(landCoverType >= 0);
}
public Bitmap2(Bitmap bmp = null, string name = "", string dec = "", GRasterLayer gdalLayer = null, IGBand gdalBand = null)
{
_bitmap = bmp;
_name = name;
_dec = dec;
_gdalBand = gdalBand;
_gdalLayer = gdalLayer;
}
public JobCNNClassify(GRasterLayer featureRasterLayer, GRasterLayer labelRasterLayer, int epochs, int model, int width, int height, int channel)
{
_t = new Thread(() => {
ImageClassifyEnv env = new ImageClassifyEnv(featureRasterLayer, labelRasterLayer);
CNN cnn = new CNN(new int[] { channel, width, height }, env.ActionNum);
//training
Summary = "模型训练中";
for (int i = 0; i < epochs; i++)
{
int batchSize = cnn.BatchSize;
var(states, labels) = env.RandomEval(batchSize);
double[][] inputX = new double[batchSize][];
for (int j = 0; j < batchSize; j++)
{
inputX[j] = states[j];
}
double loss = cnn.Train(inputX, labels);
Process = (double)i / epochs;
}
//classify
Summary = "分类应用中";
IRasterLayerCursorTool pRasterLayerCursorTool = new GRasterLayerCursorTool();
pRasterLayerCursorTool.Visit(featureRasterLayer);
//GDI graph
Bitmap classificationBitmap = new Bitmap(featureRasterLayer.XSize, featureRasterLayer.YSize);
Graphics g = Graphics.FromImage(classificationBitmap);
//
int seed = 0;
int totalPixels = featureRasterLayer.XSize * featureRasterLayer.YSize;
//应用dqn对图像分类
for (int i = 0; i < featureRasterLayer.XSize; i++)
{
for (int j = 0; j < featureRasterLayer.YSize; j++)
{
//get normalized input raw value
double[] normal = pRasterLayerCursorTool.PickNormalValue(i, j);
//}{debug
double[] action = cnn.Predict(normal);
//convert action to raw byte value
int gray = env.RandomSeedKeys[NP.Argmax(action)];
//后台绘制，报告进度
Color c = Color.FromArgb(gray, gray, gray);
Pen p = new Pen(c);
SolidBrush brush = new SolidBrush(c);
g.FillRectangle(brush, new Rectangle(i, j, 1, 1));
//report progress
Process = (double)(seed++) / totalPixels;
}
}
//保存结果至tmp
string fullFileName = Directory.GetCurrentDirectory() + @"\tmp\" + DateTime.Now.ToFileTimeUtc() + ".png";
classificationBitmap.Save(fullFileName);
//complete
Summary = "CNN训练分类完成";
Complete = true;
OnTaskComplete?.Invoke(Name, fullFileName);
});
}
private void button3_Click(object sender, EventArgs e)
{
//计算kappa,并输出矩阵
GRasterLayer truthLayer = _rasterDic[truthKey];
GRasterLayer predLayer = _rasterDic[predKey];
var(matrix, kappa, actionsNumber, oa) = KappaIndex.Calcute(truthLayer, predLayer);
//matrix for binding and kappa for display
kappa_label.Text = string.Format("kappa:{0:P} oa:{1:P}", kappa, oa);
//绘表头
webBrowser1.DocumentText = GenericTable(matrix);
}
public void RasterBandStatisticTool()
{
GRasterLayer rasterLayer = new GRasterLayer(fullFilename);
IRasterBandStatisticTool pRasterBandStasticTool = new GRasterBandStatisticTool();
pRasterBandStasticTool.Visit(rasterLayer.BandCollection[0]);
var rawGraph = pRasterBandStasticTool.StaisticalRawGraph;
var rawTable = pRasterBandStasticTool.StatisticalRawQueryTable;
Assert.AreEqual(rawGraph.Count, 255);
Assert.AreEqual(rawTable.Length, 768000);
}
public void RasterLayerPickValueTool()
{
GRasterLayer rasterLayer = new GRasterLayer(fullFilename);
IRasterLayerCursorTool pRasterLayerCursorTool = new GRasterLayerCursorTool();
pRasterLayerCursorTool.Visit(rasterLayer);
//returen the nomalized values form each layer at given point (100,100)
double[] normalValue = pRasterLayerCursorTool.PickNormalValue(100, 100);
//the col and row should be odd number
double[] rangeNormalValue = pRasterLayerCursorTool.PickRagneNormalValue(100, 100, 3, 3);
Assert.AreEqual(string.Join(",", normalValue), "0.622047244094488,0.574803149606299,0.322834645669291,0.12992125984252,0.145669291338583,0.338582677165354,0.192913385826772,0.192913385826772,0.354330708661417,0.488188976377953,0.165354330708661,0.181102362204724,0.12992125984252,0.366141732283465,0.133858267716535,0.102362204724409,0.381889763779528,0.133858267716535");
Assert.AreEqual(string.Join(",", rangeNormalValue), "0.645669291338583,0.637795275590551,0.661417322834646,0.614173228346457,0.622047244094488,0.645669291338583,0.614173228346457,0.618110236220472,0.622047244094488,0.622047244094488,0.598425196850394,0.610236220472441,0.574803149606299,0.574803149606299,0.586614173228346,0.570866141732283,0.566929133858268,0.566929133858268,0.385826771653543,0.354330708661417,0.354330708661417,0.330708661417323,0.322834645669291,0.326771653543307,0.322834645669291,0.31496062992126,0.311023622047244,0.12992125984252,0.137795275590551,0.153543307086614,0.125984251968504,0.12992125984252,0.145669291338583,0.133858267716535,0.137795275590551,0.141732283464567,0.165354330708661,0.15748031496063,0.18503937007874,0.133858267716535,0.145669291338583,0.177165354330709,0.125984251968504,0.133858267716535,0.137795275590551,0.409448818897638,0.374015748031496,0.362204724409449,0.346456692913386,0.338582677165354,0.334645669291339,0.338582677165354,0.326771653543307,0.322834645669291,0.196850393700787,0.196850393700787,0.228346456692913,0.18503937007874,0.192913385826772,0.216535433070866,0.196850393700787,0.200787401574803,0.208661417322835,0.196850393700787,0.196850393700787,0.228346456692913,0.18503937007874,0.192913385826772,0.216535433070866,0.196850393700787,0.200787401574803,0.208661417322835,0.413385826771654,0.385826771653543,0.385826771653543,0.358267716535433,0.354330708661417,0.358267716535433,0.358267716535433,0.346456692913386,0.338582677165354,0.409448818897638,0.425196850393701,0.551181102362205,0.409448818897638,0.488188976377953,0.362204724409449,0.374015748031496,0.437007874015748,0.275590551181102,0.106299212598425,0.153543307086614,0.141732283464567,0.12992125984252,0.165354330708661,0.125984251968504,0.141732283464567,0.114173228346457,0.137795275590551,0.122047244094488,0.188976377952756,0.177165354330709,0.145669291338583,0.181102362204724,0.133858267716535,0.133858267716535,0.0984251968503937,0.118110236220472,0.12992125984252,0.133858267716535,0.153543307086614,0.125984251968504,0.12992125984252,0.145669291338583,0.133858267716535,0.137795275590551,0.141732283464567,0.437007874015748,0.401574803149606,0.397637795275591,0.374015748031496,0.366141732283465,0.37007874015748,0.366141732283465,0.358267716535433,0.354330708661417,0.153543307086614,0.145669291338583,0.169291338582677,0.125984251968504,0.133858267716535,0.161417322834646,0.118110236220472,0.125984251968504,0.125984251968504,0.094488188976378,0.102362204724409,0.118110236220472,0.0984251968503937,0.102362204724409,0.110236220472441,0.118110236220472,0.118110236220472,0.122047244094488,0.452755905511811,0.417322834645669,0.405511811023622,0.389763779527559,0.381889763779528,0.374015748031496,0.389763779527559,0.374015748031496,0.366141732283465,0.153543307086614,0.145669291338583,0.169291338582677,0.125984251968504,0.133858267716535,0.161417322834646,0.118110236220472,0.125984251968504,0.125984251968504");
}
/// <summary>
/// Update Read Raster UI
/// </summary>
private void UpdateReadRasterUI(string nodeName, Dictionary <string, Bitmap2> dict, GRasterLayer rasterLayer)
{
TreeNode node = map_treeView.Nodes[nodeName];
_imageDic[nodeName] = null;
_rasterDic[nodeName] = rasterLayer;
foreach (var key in dict.Keys)
{
TreeNode childNode = new TreeNode(key);
_imageDic[key] = dict[key];
Invoke(new UpdateTreeNodeHandler(UpdateTreeNode), node, childNode);
}
}
/// <summary>
/// 指定观察的图像，和样本所在的层位置
/// </summary>
/// <param name="featureRasterLayer"></param>
/// <param name="sampleIndex"></param>
public ImageClassifyEnv(GRasterLayer featureRasterLayer, GRasterLayer labelRasterLayer)
{
//input feature raster layer
_featureRasterLayer = featureRasterLayer;
//groundtruth raster layer
_labelRasterLayer = labelRasterLayer;
//num of categories
//标注层要求：
//1.分类按照顺序，从1开始，逐步+1
//2.背景值设置为0
//ActionNum = Convert.ToInt32(_labelRasterLayer.BandCollection[0].Max - _labelRasterLayer.BandCollection[0].Min);
ActionNum = Convert.ToInt32(_labelRasterLayer.BandCollection[0].Max - 0);
//statical graph
Prepare();
}
/// <summary>
/// 波段读取
/// </summary>
/// <param name="filePath"></param>
/// <param name="parentNode"></param>
private void ReadBand(string rasterFilename, TreeNode parentNode)
{
string name = parentNode.Text;
GRasterLayer _layer = new GRasterLayer(rasterFilename);
for (int i = 0; i < _layer.BandCollection.Count; i++)
{
IGBand band = _layer.BandCollection[i];
band.BandName = name + "_波段_" + i;
Bitmap2 bmp2 = new Bitmap2(bmp: band.GetBitmap(), name: band.BandName, gdalBand: band, gdalLayer: _layer);
//获取band对应的bitmap格式图像，载入treedNode中
_imageDic.Add(band.BandName, bmp2);
TreeNode childrenNode = new TreeNode(band.BandName);
Invoke(new UpdateTreeNodeHandler(UpdateTreeNode), parentNode, childrenNode);
}
}
/// <summary>
/// DQN classify task
/// </summary>
/// <param name="featureRasterLayer"></param>
/// <param name="labelRasterLayer"></param>
/// <param name="epochs"></param>
public JobDQNClassify(GRasterLayer featureRasterLayer, GRasterLayer labelRasterLayer, int epochs = 3000)
{
_t = new Thread(() =>
{
ImageClassifyEnv env = new ImageClassifyEnv(featureRasterLayer, labelRasterLayer);
_dqn = new DQN(env);
_dqn.SetParameters(epochs: epochs, gamma: _gamma);
_dqn.OnLearningLossEventHandler += _dqn_OnLearningLossEventHandler;
//training
Summary = "模型训练中";
_dqn.Learn();
//classification
Summary = "分类应用中";
IRasterLayerCursorTool pRasterLayerCursorTool = new GRasterLayerCursorTool();
pRasterLayerCursorTool.Visit(featureRasterLayer);
Bitmap classificationBitmap = new Bitmap(featureRasterLayer.XSize, featureRasterLayer.YSize);
Graphics g = Graphics.FromImage(classificationBitmap);
int seed = 0;
int totalPixels = featureRasterLayer.XSize * featureRasterLayer.YSize;
for (int i = 0; i < featureRasterLayer.XSize; i++)
{
for (int j = 0; j < featureRasterLayer.YSize; j++)
{
//get normalized input raw value
double[] normal = pRasterLayerCursorTool.PickNormalValue(i, j);
var(action, q) = _dqn.ChooseAction(normal);
//convert action to raw byte value
int gray = _dqn.ActionToRawValue(NP.Argmax(action));
Color c = Color.FromArgb(gray, gray, gray);
Pen p = new Pen(c);
SolidBrush brush = new SolidBrush(c);
g.FillRectangle(brush, new Rectangle(i, j, 1, 1));
//report progress
Process = (double)(seed++) / totalPixels;
}
}
//save result
string fullFileName = Directory.GetCurrentDirectory() + @"\tmp\" + DateTime.Now.ToFileTimeUtc() + ".png";
classificationBitmap.Save(fullFileName);
//complete
Summary = "DQN训练分类完成";
Complete = true;
OnTaskComplete?.Invoke(Name, fullFileName);
});
}
public IJobRPCRectify(double[] a, double[] b, double[] c, double[] d, Dictionary <string, double> paramaters, List <string> rawBinRasterFullFilenames)
{
_t = new Thread(() =>
{
using (IRasterRPCTool pRasterRPCTool = new GRasterRPCTool(a, b, c, d, paramaters))
{
for (int i = 0; i < rawBinRasterFullFilenames.Count; i++)
{
string rasterFilename = rawBinRasterFullFilenames[i];
GRasterLayer rasterLayer = new GRasterLayer(rasterFilename);
Summary = string.Format("total:{1}/{2}, RPC rectify for {0} is in progress.... ", rasterLayer.Name, i + 1, rawBinRasterFullFilenames.Count);
pRasterRPCTool.Visit(rasterLayer);
pRasterRPCTool.DoRPCRectify();
Process = i / (double)rawBinRasterFullFilenames.Count;
}
OnTaskComplete?.Invoke(Name);
}
});
}
public void RasterBandPickValueTool()
{
GRasterLayer rasterLayer = new GRasterLayer(fullFilename);
IRasterBandCursorTool pRasterBandCursorTool = new GRasterBandCursorTool();
pRasterBandCursorTool.Visit(rasterLayer.BandCollection[0]);
double normalValue = pRasterBandCursorTool.PickNormalValue(100, 100);
//the col and row should be odd number
double[] rangeNormalValue = pRasterBandCursorTool.PickRangeNormalValue(100, 100, 3, 3);
double rawValue = pRasterBandCursorTool.PickRawValue(100, 100);
//the col and row should be odd number
double[] rangeRawValue = pRasterBandCursorTool.PickRangeRawValue(100, 100, 3, 3);
//
Assert.AreEqual(normalValue, 0.62204724409448819);
Assert.AreEqual(string.Join(",", rangeNormalValue), "0.645669291338583,0.637795275590551,0.661417322834646,0.614173228346457,0.622047244094488,0.645669291338583,0.614173228346457,0.618110236220472,0.622047244094488");
Assert.AreEqual(rawValue, 159);
Assert.AreEqual(string.Join(",", rangeRawValue), "165,163,169,157,159,165,157,158,159");
}
public void ClassificationByCNN()
{
//loss
double _loss = 1.0;
//training epochs
int epochs = 100;
//
GRasterLayer featureLayer = new GRasterLayer(featureFullFilename);
GRasterLayer labelLayer = new GRasterLayer(trainFullFilename);
//create environment for agent exploring
IEnv env = new ImageClassifyEnv(featureLayer, labelLayer);
//assume 18dim equals 3x6 (image)
CNN cnn = new CNN(new int[] { 1, 3, 6 }, env.ActionNum);
//training
for (int i = 0; i < epochs; i++)
{
int batchSize = cnn.BatchSize;
var(states, labels) = env.RandomEval(batchSize);
double[][] inputX = new double[batchSize][];
for (int j = 0; j < batchSize; j++)
{
inputX[j] = states[j];
}
_loss = cnn.Train(inputX, labels);
}
//in general, loss is less than 5
Assert.IsTrue(_loss < 5.0);
//apply cnn to classify featureLayer
IRasterLayerCursorTool pRasterLayerCursorTool = new GRasterLayerCursorTool();
pRasterLayerCursorTool.Visit(featureLayer);
//get normalized input raw value
double[] normal = pRasterLayerCursorTool.PickNormalValue(50, 50);
double[] action = cnn.Predict(normal);
int landCoverType = env.RandomSeedKeys[NP.Argmax(action)];
//do something as you need. i.e. draw landCoverType to bitmap at position ( i , j )
//the classification results are not stable because of the training epochs are too few.
Assert.IsTrue(landCoverType >= 0);
}
/// <summary>
///
/// </summary>
/// <param name="fullFilename"></param>
public JobReadRaster(string fullFilename)
{
_t = new Thread(() =>
{
Dictionary <string, Bitmap2> dict = new Dictionary <string, Bitmap2>();
string name = Path.GetFileNameWithoutExtension(fullFilename);
GRasterLayer rasterLayer = new GRasterLayer(fullFilename);
//reading
Summary = "数据读取中";
for (int i = 0; i < rasterLayer.BandCount; i++)
{
GRasterBand band = rasterLayer.BandCollection[i];
band.BandName = name + "_band_" + i;
Bitmap2 bmp2 = new Bitmap2(bmp: band.GrayscaleImage, name: band.BandName, gdalBand: band, gdalLayer: rasterLayer);
dict[band.BandName] = bmp2;
Process = (double)(i + 1) / rasterLayer.BandCount;
}
//read complete
Summary = "读取完毕";
Complete = true;
OnTaskComplete?.Invoke(Name, name, dict, rasterLayer);
});
}
/// <summary>
/// 指定观察的图像，和样本所在的层位置
/// </summary>
/// <param name="featureRasterLayer"></param>
/// <param name="sampleIndex"></param>
public ExtractRoadEnv(GRasterLayer featureRasterLayer, GRasterLayer labelRasterLayer)
{
//input
_featureRasterLayer = featureRasterLayer;
//output
_labelRasterLayer = labelRasterLayer;
// 探索方式，沿顺时针
// -----------------------------------------------------
// * 0 | 1 | 2
// * -----------------------
// * 7 | p | 3
// * -----------------------
// * 6 | 5 | 4
//represent eight direction actions
ActionNum = 8;
//
_channel = _featureRasterLayer.BandCount;
//limit of x
_limit_x = _labelRasterLayer.XSize;
//limit of y
_limit_y = _labelRasterLayer.YSize;
//read labellayer
Prepare();
}
public void Visit(GRasterLayer pLayer)
{
_pLayer = pLayer;
}
/// <summary>
///
/// </summary>
/// <param name="treeCount"></param>
/// <param name="fullFilename"></param>
/// <param name="rasterLayer"></param>
public JobRFClassify(int treeCount, string fullFilename, GRasterLayer rasterLayer)
{
_t = new Thread(() =>
{
RF rf = new RF(treeCount);
//training
Summary = "随机森林训练中";
using (StreamReader sr = new StreamReader(fullFilename))
{
List <List <double> > inputList = new List <List <double> >();
List <int> outputList = new List <int>();
string text = sr.ReadLine().Replace("\t", ",");
do
{
string[] rawdatas = text.Split(',');
outputList.Add(Convert.ToInt32(rawdatas.Last()));
List <double> inputItem = new List <double>();
for (int i = 0; i < rawdatas.Length - 1; i++)
{
inputItem.Add(Convert.ToDouble(rawdatas[i]));
}
inputList.Add(inputItem);
text = sr.ReadLine();
} while (text != null);
double[][] inputs = new double[inputList.Count][];
int[] outputs = outputList.ToArray();
for (int i = 0; i < inputList.Count; i++)
{
inputs[i] = inputList[i].ToArray();
}
rf.Train(inputs, outputs);
}
//image classify
Summary = "分类应用中";
IRasterLayerCursorTool pRasterLayerCursorTool = new GRasterLayerCursorTool();
pRasterLayerCursorTool.Visit(rasterLayer);
//GDI graph
Bitmap classificationBitmap = new Bitmap(rasterLayer.XSize, rasterLayer.YSize);
Graphics g = Graphics.FromImage(classificationBitmap);
//
int seed = 0;
int totalPixels = rasterLayer.XSize * rasterLayer.YSize;
Process = 0.0;
//应用dqn对图像分类
for (int i = 0; i < rasterLayer.XSize; i++)
{
for (int j = 0; j < rasterLayer.YSize; j++)
{
//get normalized input raw value
double[] raw = pRasterLayerCursorTool.PickRawValue(i, j);
double[][] inputs = new double[1][];
inputs[0] = raw;
//}{debug
int[] ouputs = rf.Predict(inputs);
int gray = ouputs[0];
//convert action to raw byte value
Color c = Color.FromArgb(gray, gray, gray);
Pen p = new Pen(c);
SolidBrush brush = new SolidBrush(c);
g.FillRectangle(brush, new Rectangle(i, j, 1, 1));
//report progress
Process = (double)seed++ / totalPixels;
}
}
//保存结果至tmp
string fullFileName = Directory.GetCurrentDirectory() + @"\tmp\" + DateTime.Now.ToFileTimeUtc() + ".png";
classificationBitmap.Save(fullFileName);
//rf complete
Summary = "RF训练分类完成";
Complete = true;
OnTaskComplete?.Invoke(Name, fullFileName);
});
}
/// <summary>
/// 底图区域功能按钮
/// </summary>
/// <param name="sender"></param>
/// <param name="e"></param>
private void Map_function_Click(object sender, EventArgs e)
{
ToolStripItem item = sender as ToolStripItem;
switch (item.Name)
{
case "Export_Bitmap_ToolStripMenuItem":
Bitmap saveBmp = map_pictureBox.Image as Bitmap;
SaveBitmap(saveBmp);
//导出图片
break;
case "DL_CLASS_toolStripButton":
if (map_treeView.SelectedNode == null)
{
UpdateStatusLabel("请选择一副图像或者一个图层后，进行分类操作", STATUE_ENUM.ERROR);
return;
}
else
{
DLClassifyForm dlclassify = new DLClassifyForm();
if (dlclassify.ShowDialog() == DialogResult.OK)
{
//1.选择处理那副图像
string imageName = map_treeView.SelectedNode.Text;
Bitmap2 imageBitmap2 = _imageDic[imageName];
GRasterLayer rasterLayer = imageBitmap2.GdalLayer;
ThreadStart clsfy_ts = delegate { RunClassify(rasterLayer, dlclassify.UseSLIC, dlclassify.PBName, dlclassify.CenterName, dlclassify.LabelName); };
Thread clsfy_t = new Thread(clsfy_ts);
clsfy_t.IsBackground = true;
clsfy_t.Start();
}
}
break;
case "open_toolstripmenuitem": //添加图像
ReadImage();
break;
case "open_contextMenuStrip":
ReadImage();
break;
//超像素分割
case "SLIC_toolStripButton":
case "SLIC_toolStripMenu":
Bitmap bmp = map_pictureBox.Image as Bitmap;
if (bmp != null)
{
ThreadStart slic_ts = delegate { RunSLIC(bmp); };
Thread slic_t = new Thread(slic_ts);
slic_t.IsBackground = true;
slic_t.Start();
}
else
{
UpdateStatusLabel("未选中待计算图像，地图区域无图片", STATUE_ENUM.ERROR);
}
break;
//超像素中心应用
case "SLIC_Center_toolStripButton":
case "SLIC_Center_toolStripMenu":
OpenFileDialog opg = new OpenFileDialog
{
Filter = "JSON文件|*.json"
};
if (opg.ShowDialog() == DialogResult.OK)
{
//1.读取center中心
using (StreamReader sr = new StreamReader(opg.FileName))
{
List <byte> colors = new List <byte>();
Center[] centers = SuperPixelSegment.ReadCenter(sr.ReadToEnd());
//2.设置使用图层
CenterApplyForm centerApplyForm = new CenterApplyForm();
if (centerApplyForm.ShowDialog() == DialogResult.OK)
{
ThreadStart s = delegate { RunCenter(centerApplyForm.FileNameCollection, centers); };
Thread t = new Thread(s)
{
IsBackground = true
};
t.Start();
}
}
}
break;
default:
break;
}
}
public static (int[, ] matrix, double kappa, int actionsNumber, double oa) Calcute(GRasterLayer truthLayer, GRasterLayer predLayer)
{
//statical label band graph
IRasterBandStatisticTool pBandStaticTool = new GRasterBandStatisticTool();
pBandStaticTool.Visit(truthLayer.BandCollection[0]);
Dictionary <int, List <Point> > memory = pBandStaticTool.StaisticalRawGraph;
//key index
List <int> Keys = memory.Keys.ToList();
int actionsNumber = Keys.Count;
int[,] matrix = new int[actionsNumber, actionsNumber];
IRasterBandCursorTool pBandCursorTool = new GRasterBandCursorTool();
pBandCursorTool.Visit(predLayer.BandCollection[0]);
//
pBandStaticTool.Visit(predLayer.BandCollection[0]);
var m = pBandStaticTool.StaisticalRawGraph;
//
for (int i = 0; i < actionsNumber; i++)
{
int key = Keys[i];
List <Point> points = memory[key];
//计算realKey类分类结果,存入混淆矩阵
points.ForEach(p =>
{
int rawType = (int)pBandCursorTool.PickRawValue(p.X, p.Y);
int indexType = Keys.IndexOf(rawType);
if (indexType != -1)
{
matrix[i, indexType]++;
}
});
}
// Create a new multi-class Confusion Matrix
var cm = new GeneralConfusionMatrix(matrix);
//
int totalNum = cm.NumberOfSamples;
//p0
double p0 = 0;
for (int i = 0; i < actionsNumber; i++)
{
p0 += Convert.ToDouble(matrix[i, i]);
}
//pc
double pc = 0;
for (int i = 0; i < actionsNumber; i++)
{
pc += Convert.ToDouble(cm.ColumnTotals[i]) * Convert.ToDouble(cm.RowTotals[i]);
}
pc = pc / totalNum;
//
double kappa = (p0 - pc) / (totalNum - pc);
double oa = p0 / totalNum;
//
return(matrix, kappa, actionsNumber, oa);
}
public GoutputRaster(GRasterLayer rasterLayer)
{
_rasterLayer = rasterLayer;
}
public void ReadRasterLayer()
{
GRasterLayer rasterLayer = new GRasterLayer(fullFilename);
Assert.AreEqual(rasterLayer.BandCollection.Count, 18);
}
/// <summary>
/// 应用深度学习模型进行分类
/// </summary>
/// <param name="rasterLayer"></param>
private void RunClassify(GRasterLayer rasterLayer, bool useSLIC, string pbName, string centerName, string labelName)
{
//判断图层结构，选用不同的tensor输入
ShapeEnum shapeEuum;
if (rasterLayer.BandCount == 130)
{
shapeEuum = ShapeEnum.THIRTEEN_TEN;
}
else if (rasterLayer.BandCount == 100)
{
shapeEuum = ShapeEnum.TEN_TEN;
}
else if (rasterLayer.BandCount == 64)
{
shapeEuum = ShapeEnum.EIGHT_EIGHT;
}
else
{
shapeEuum = ShapeEnum.TEN_TEN;
}
//构建结果图层，用于动态绘制
Bitmap bmp = new Bitmap(rasterLayer.XSize, rasterLayer.YSize);
string nodeName = rasterLayer.Name + "结果图层";
TreeNode childrenNode = new TreeNode(nodeName);
_imageDic.Add(nodeName, new Bitmap2(name: nodeName, bmp: bmp));
Invoke(new UpdateTreeNodeHandler(UpdateTreeNode), null, childrenNode);
//获取波段
TensorflowBootstrap model = new TensorflowBootstrap(pbName);
//判断是否基于超像素
if (!useSLIC)
{
for (int i = 0; i < rasterLayer.XSize; i++)
{
for (int j = 0; j < rasterLayer.YSize; j++)
{
float[] input = rasterLayer.GetPixelFloat(i, j).ToArray();
long classified = model.Classify(input, shapeEuum);
Invoke(new PaintPointHandler(PaintPoint), bmp, i, j, Convert.ToByte(classified * 15));
Invoke(new UpdateStatusLabelHandler(UpdateStatusLabel), "应用分类中，总进度：" + i + "列" + j + "行", STATUE_ENUM.WARNING);
}
}
}
else
{
//基于slic的超像素绘制方法
using (StreamReader sr_center = new StreamReader(centerName))
using (StreamReader sr_label = new StreamReader(labelName))
{
Center[] centers = SuperPixelSegment.ReadCenter(sr_center.ReadToEnd());
Bitplane labels = SuperPixelSegment.ReadLabel(sr_label.ReadToEnd());
int[] mask = new int[] { 1, 1, 1, 1, 1, 1, 1, 1, 1 };
for (int i = 0; i < centers.Length; i++)
{
Center center = centers[i];
float[] input = rasterLayer.GetPixelFloatWidthConv((int)center.X, (int)center.Y, mask).ToArray();
long classified = model.Classify(input, shapeEuum);
center.L = classified * 15;
center.A = classified * 15;
center.B = classified * 15;
Invoke(new UpdateStatusLabelHandler(UpdateStatusLabel), "已处理第" + i + "/" + centers.Length + "个中心", STATUE_ENUM.WARNING);
}
//遍历图片进行绘制
for (int i = 0; i < rasterLayer.XSize; i++)
{
for (int j = 0; j < rasterLayer.YSize; j++)
{
Invoke(new PaintPointHandler(PaintPoint), bmp, i, j, Convert.ToByte(centers[(int)Math.Floor(labels.GetPixel(i, j))].L));
}
}
}
}
}
public void Dispose()
{
_pLayer = null;
}
