From bdb39c22d561e4a26b0c0c6d9f9e4ff5021865d4 Mon Sep 17 00:00:00 2001 From: lidapeng Date: Wed, 22 Jan 2020 16:23:17 +0800 Subject: [PATCH] =?UTF-8?q?=E6=B7=BB=E5=8A=A0=E7=9F=A9=E9=98=B5PHSH?= =?UTF-8?q?=E5=85=83=E7=B4=A0=EF=BC=8C=E7=9F=A9=E9=98=B5=E7=AC=9B=E5=8D=A1?= =?UTF-8?q?=E5=B0=94=E7=A7=AF?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- .../org/wlld/MatrixTools/MatrixOperation.java | 86 +++++++++++++++++++ .../wlld/imageRecognition/Convolution.java | 15 ++-- .../org/wlld/imageRecognition/Operation.java | 8 +- .../wlld/imageRecognition/TempleConfig.java | 44 ++++++++++ .../wlld/imageRecognition/border/Border.java | 67 +++++++++++++++ 5 files changed, 211 insertions(+), 9 deletions(-) create mode 100644 src/main/java/org/wlld/imageRecognition/border/Border.java diff --git a/src/main/java/org/wlld/MatrixTools/MatrixOperation.java b/src/main/java/org/wlld/MatrixTools/MatrixOperation.java index c953912..9d19ea5 100644 --- a/src/main/java/org/wlld/MatrixTools/MatrixOperation.java +++ b/src/main/java/org/wlld/MatrixTools/MatrixOperation.java @@ -1,5 +1,6 @@ package org.wlld.MatrixTools; +import com.sun.org.apache.bcel.internal.generic.FADD; import org.wlld.tools.ArithUtil; public class MatrixOperation { @@ -23,6 +24,91 @@ public class MatrixOperation { } } + public static Matrix pushVector(Matrix myMatrix, Matrix matrix, boolean addRow) throws Exception { + if (matrix.getX() == 1 || matrix.getY() == 1) { + Matrix addMatrix; + if (addRow) {//增加一行 + if (matrix.getY() != myMatrix.getY()) { + throw new Exception("this matrix column is not equals"); + } + addMatrix = new Matrix(myMatrix.getX() + 1, myMatrix.getY()); + } else {//增加一列 + if (matrix.getX() != myMatrix.getX()) { + throw new Exception("this matrix row is not equals"); + } + addMatrix = new Matrix(myMatrix.getX(), myMatrix.getY() + 1); + } + for (int i = 0; i < addMatrix.getX(); i++) { + for (int j = 0; j < addMatrix.getY(); j++) { + if (addRow) { + if (i == addMatrix.getX() - 1) {//最后一行 + addMatrix.setNub(i, j, matrix.getNumber(0, j)); + } else { + addMatrix.setNub(i, j, myMatrix.getNumber(i, j)); + } + } else { + if (j == addMatrix.getY() - 1) {//最后一列 + addMatrix.setNub(i, j, matrix.getNumber(i, 0)); + } else { + addMatrix.setNub(i, j, myMatrix.getNumber(i, j)); + } + } + } + } + return addMatrix; + } else { + throw new Exception("this matrix is not a vector"); + } + } + + public static Matrix push(Matrix matrix, double nub) throws Exception {//向一个向量里PUSH一个值 + if (matrix.getX() == 1 || matrix.getY() == 1) { + Matrix myMatrix; + int nubs; + boolean isRow = true; + if (matrix.getX() == 1) {//行向量 + nubs = matrix.getY() + 1; + myMatrix = new Matrix(1, nubs); + } else {//列向量 + isRow = false; + nubs = matrix.getX() + 1; + myMatrix = new Matrix(nubs, 1); + } + for (int i = 0; i < nubs; i++) { + if (i == nubs - 1) { + if (isRow) { + myMatrix.setNub(0, i, nub); + } else { + myMatrix.setNub(i, 0, nub); + } + } else { + if (isRow) {//行向量 + myMatrix.setNub(0, i, matrix.getNumber(0, i)); + } else {//列向量 + myMatrix.setNub(i, 0, matrix.getNumber(i, 0)); + } + } + } + return myMatrix; + } else { + throw new Exception("this matrix is not a vector"); + } + } + + public static Matrix matrixToRow(Matrix matrix) throws Exception {//将一个矩阵转成行向量 + int x = matrix.getX(); + int y = matrix.getY(); + Matrix myMatrix = new Matrix(1, x * y); + int t = 0; + for (int i = 0; i < x; i++) { + for (int j = 0; j < y; j++) { + myMatrix.setNub(0, t, matrix.getNumber(i, j)); + t++; + } + } + return myMatrix; + } + public static double innerProduct(Matrix matrix1, Matrix matrix2) throws Exception {//两个向量内积 Matrix matrix = transPosition(matrix1); Matrix matrix3 = mulMatrix(matrix, matrix2); diff --git a/src/main/java/org/wlld/imageRecognition/Convolution.java b/src/main/java/org/wlld/imageRecognition/Convolution.java index 9ae6131..be217a4 100644 --- a/src/main/java/org/wlld/imageRecognition/Convolution.java +++ b/src/main/java/org/wlld/imageRecognition/Convolution.java @@ -10,16 +10,16 @@ import org.wlld.config.Kernel; * @date 9:23 上午 2020/1/2 */ public class Convolution { - public Matrix getFeatures(Matrix matrix, int maxNub) throws Exception { + public Matrix getFeatures(Matrix matrix, int maxNub, boolean isBorder) throws Exception { do { - matrix = convolution(matrix, Kernel.ALL_Two); + matrix = convolution(matrix, Kernel.ALL_Two, isBorder); } - while (matrix.getX() > maxNub && matrix.getY() > maxNub); + while (matrix.getX() > maxNub && matrix.getY() > maxNub && !isBorder); //已经不可以再缩小了，最后做一层卷积，然后提取最大值 return matrix; } - private Matrix convolution(Matrix matrix, Matrix kernel) throws Exception { + private Matrix convolution(Matrix matrix, Matrix kernel, boolean isBorder) throws Exception { int x = matrix.getX() - 2;//求导后矩阵的行数 int y = matrix.getY() - 2;//求导后矩阵的列数 Matrix myMatrix = new Matrix(x, y);//最终合成矩阵 @@ -31,8 +31,13 @@ public class Convolution { } } } - return late(myMatrix); + if (isBorder) { + return myMatrix; + } else { + return late(myMatrix); + } } + public Matrix late(Matrix matrix) throws Exception {//迟化处理 int xn = matrix.getX(); int yn = matrix.getY(); diff --git a/src/main/java/org/wlld/imageRecognition/Operation.java b/src/main/java/org/wlld/imageRecognition/Operation.java index f33b951..0aa9d41 100644 --- a/src/main/java/org/wlld/imageRecognition/Operation.java +++ b/src/main/java/org/wlld/imageRecognition/Operation.java @@ -19,7 +19,7 @@ public class Operation {//进行计算 this.templeConfig = templeConfig; } - public List convolution(Matrix matrix) throws Exception { + public List convolution(Matrix matrix, boolean isBorder) throws Exception { //进行卷积 int maxNub = 0; if (templeConfig.getRow() >= templeConfig.getColumn()) { @@ -27,14 +27,14 @@ public class Operation {//进行计算 } else { maxNub = templeConfig.getColumn(); } - Matrix matrix1 = convolution.getFeatures(matrix, maxNub); + Matrix matrix1 = convolution.getFeatures(matrix, maxNub, isBorder); return sub(matrix1); } //模板学习 public void study(Matrix matrix, Map tagging) throws Exception { if (templeConfig.getStudyPattern() == StudyPattern.Speed_Pattern) { - List list = convolution(matrix); + List list = convolution(matrix, templeConfig.isHavePosition()); intoNerve(1, list, templeConfig.getSensoryNerves(), true, tagging); } else { throw new Exception("pattern is wrong"); @@ -58,7 +58,7 @@ public class Operation {//进行计算 //图像视觉 speed 模式 public void look(Matrix matrix, long eventId) throws Exception { if (templeConfig.getStudyPattern() == StudyPattern.Speed_Pattern) { - List list = convolution(matrix); + List list = convolution(matrix, templeConfig.isHavePosition()); intoNerve(eventId, list, templeConfig.getSensoryNerves(), false, null); } else if (templeConfig.getStudyPattern() == StudyPattern.Accuracy_Pattern) { see(matrix, eventId); diff --git a/src/main/java/org/wlld/imageRecognition/TempleConfig.java b/src/main/java/org/wlld/imageRecognition/TempleConfig.java index 7778387..857d0e9 100644 --- a/src/main/java/org/wlld/imageRecognition/TempleConfig.java +++ b/src/main/java/org/wlld/imageRecognition/TempleConfig.java @@ -25,6 +25,50 @@ public class TempleConfig { private int classificationNub = 1;//分类的数量 private int studyPattern;//学习模式 private OutBack outBack; + private boolean isHavePosition = false;//是否需要锁定物体位置 + private int region = 10;//将一个图片均匀分成几个区域 默认是10 + private int regionRow = 8;//目标检测物体大概会在图像中占的行份 + private int regionColumn = 8;//目标检测物体大概会在图像中占的列份 + + public int getRegionRow() { + return regionRow; + } + + public void setRegionRow(int regionRow) { + this.regionRow = regionRow; + } + + public int getRegionColumn() { + return regionColumn; + } + + public void setRegionColumn(int regionColumn) { + this.regionColumn = regionColumn; + } + + public NerveManager getNerveManager() { + return nerveManager; + } + + public void setNerveManager(NerveManager nerveManager) { + this.nerveManager = nerveManager; + } + + public int getRegion() { + return region; + } + + public void setRegion(int region) { + this.region = region; + } + + public boolean isHavePosition() { + return isHavePosition; + } + + public void setHavePosition(boolean havePosition) { + isHavePosition = havePosition; + } public NerveManager getConvolutionNerveManager() { return convolutionNerveManager; diff --git a/src/main/java/org/wlld/imageRecognition/border/Border.java b/src/main/java/org/wlld/imageRecognition/border/Border.java new file mode 100644 index 0000000..2027a82 --- /dev/null +++ b/src/main/java/org/wlld/imageRecognition/border/Border.java @@ -0,0 +1,67 @@ +package org.wlld.imageRecognition.border; + +import org.wlld.MatrixTools.Matrix; +import org.wlld.imageRecognition.TempleConfig; +import org.wlld.test.Ma; +import org.wlld.tools.ArithUtil; + +/** + * @author lidapeng + * @description 图像边框 + * @date 10:37 上午 2020/1/22 + */ +public class Border { + private int minX; + private int minY; + private int maxX; + private int maxY; + private int width;//宽度 + private int height;//高度 + private double modelX;//标准左上角X坐标 + private double modelY;//标准左上角Y坐标 + private double modelHeight;//标准高度 + private double modelWidth;//标准宽度 + + Border(TempleConfig templeConfig, int imageWidth, int imageHeight) { + int region = templeConfig.getRegion();//将一个区域分成几份 + double row = ArithUtil.div(imageHeight, region);//每一行的像素 + double column = ArithUtil.div(imageWidth, region);//每一列的像素 + int divRow = templeConfig.getRegionRow();//检测物体行占据的份数 + int divColumn = templeConfig.getRegionColumn();//检测物体列占据的份数 + modelHeight = ArithUtil.mul(divRow, row); + modelWidth = ArithUtil.mul(divColumn, column); + double regionRow = (region - divRow) / 2;//左上角X份数 + double regionColumn = (region - divColumn) / 2;//左上角Y份数 + modelX = ArithUtil.mul(regionRow, row); + modelY = ArithUtil.mul(regionColumn, column); + } + + //输入像素值查找边框四角坐标 + public void setPosition(int x, int y) { + if (minX > x || minX == 0) { + minX = x; + } + if (minY > y || minY == 0) { + minY = y; + } + if (x > maxX) { + maxX = x; + } + if (y > maxY) { + maxY = y; + } + } + + public void end(Matrix matrix) {//长宽 + height = maxX - minX; + width = maxY - minY; + //多元线性回归的四个输出值 + double tx = ArithUtil.div(ArithUtil.sub(minX, modelX), modelHeight); + double ty = ArithUtil.div(ArithUtil.sub(minY, modelY), modelWidth); + double tw = Math.log(ArithUtil.div(width, modelWidth)); + double th = Math.log(ArithUtil.div(height, modelHeight)); + //进行参数汇集 + + } + +}