diff --git a/src/main/java/org/wlld/imageRecognition/Convolution.java b/src/main/java/org/wlld/imageRecognition/Convolution.java index 58a2e2a..87efbfe 100644 --- a/src/main/java/org/wlld/imageRecognition/Convolution.java +++ b/src/main/java/org/wlld/imageRecognition/Convolution.java @@ -132,6 +132,49 @@ public class Convolution extends Frequency { } + private double[] compareDis(double[] rgbTest, List rgbNorms) { + double[] feature = null; + double minDis = -1; + for (int i = 0; i < 3; i++) { + double[] rgb = rgbNorms.get(i).getRgb(); + double sigma = 0; + for (int j = 0; j < 3; j++) { + sigma = sigma + Math.pow(rgbTest[j] - rgb[j], 2); + } + if (sigma < minDis || minDis == -1) { + minDis = sigma; + feature = rgb; + } + } + return feature; + } + + private void dispersed(Matrix matrixR, Matrix matrixG, Matrix matrixB, List rgbNorms) throws Exception {//图像离散化 + ThreeChannelMatrix threeChannelMatrix = new ThreeChannelMatrix(); + int x = matrixR.getX(); + int y = matrixR.getY(); + Matrix matrixRD = new Matrix(x, y); + Matrix matrixGD = new Matrix(x, y); + Matrix matrixBD = new Matrix(x, y); + threeChannelMatrix.setMatrixR(matrixRD); + threeChannelMatrix.setMatrixG(matrixGD); + threeChannelMatrix.setMatrixB(matrixBD); + for (int i = 0; i < x; i++) { + for (int j = 0; j < y; j++) { + double r = matrixR.getNumber(i, j); + double g = matrixG.getNumber(i, j); + double b = matrixB.getNumber(i, j); + double[] rgb = new double[]{r, g, b}; + double[] rgbNow = compareDis(rgb, rgbNorms); + matrixRD.setNub(i, j, rgbNow[0]); + matrixGD.setNub(i, j, rgbNow[1]); + matrixBD.setNub(i, j, rgbNow[2]); + } + } + //输入结束进行卷积 + //System.out.println(matrixBD.getString()); + } + public List getCenterColor(ThreeChannelMatrix threeChannelMatrix, int poolSize, int sqNub) throws Exception { Matrix matrixR = threeChannelMatrix.getMatrixR(); Matrix matrixG = threeChannelMatrix.getMatrixG(); @@ -159,6 +202,8 @@ public class Convolution extends Frequency { features.add(rgb[j]); } } + //测试卷积 + //dispersed(matrixR, matrixG, matrixB, rgbNorms); //System.out.println("feature==" + feature); return features; } @@ -252,6 +297,33 @@ public class Convolution extends Frequency { return frameBodies; } + private void convolutionByThree(ThreeChannelMatrix threeChannelMatrix, Matrix kernel) throws Exception { + Matrix matrixR = threeChannelMatrix.getMatrixR(); + Matrix matrixG = threeChannelMatrix.getMatrixG(); + Matrix matrixB = threeChannelMatrix.getMatrixB(); + int x = matrixR.getX() - 2;//求导后矩阵的行数 + int y = matrixR.getY() - 2;//求导后矩阵的列数 + Matrix myMatrixR = new Matrix(x, y); + Matrix myMatrixG = new Matrix(x, y); + Matrix myMatrixB = new Matrix(x, y); + for (int i = 0; i < x; i++) {//遍历行 + for (int j = 0; j < y; j++) {//遍历每行的列 + double rm = MatrixOperation.convolution(matrixR, kernel, i, j, false); + double gm = MatrixOperation.convolution(matrixG, kernel, i, j, false); + double bm = MatrixOperation.convolution(matrixB, kernel, i, j, false); + myMatrixR.setNub(i, j, rm); + myMatrixG.setNub(i, j, gm); + myMatrixB.setNub(i, j, bm); + } + } + myMatrixR = late(myMatrixR, 2); + myMatrixG = late(myMatrixG, 2); + myMatrixB = late(myMatrixB, 2); + threeChannelMatrix.setMatrixR(myMatrixR); + threeChannelMatrix.setMatrixG(myMatrixG); + threeChannelMatrix.setMatrixB(myMatrixB); + } + private Matrix convolution(Matrix matrix, Matrix kernel, boolean isFirst , Border border, boolean isOnce) throws Exception { int x = matrix.getX() - 2;//求导后矩阵的行数 diff --git a/src/main/java/org/wlld/imageRecognition/MeanClustering.java b/src/main/java/org/wlld/imageRecognition/MeanClustering.java index e58afd9..4bc622d 100644 --- a/src/main/java/org/wlld/imageRecognition/MeanClustering.java +++ b/src/main/java/org/wlld/imageRecognition/MeanClustering.java @@ -90,15 +90,6 @@ public class MeanClustering { break; } } -// boolean isNext; -// do { -// averageMatrix(); -// isNext = isNext(); -// if (isNext) { -// clear(); -// } -// } -// while (isNext); } else { throw new Exception("matrixList number less than 2"); diff --git a/src/main/java/org/wlld/imageRecognition/Operation.java b/src/main/java/org/wlld/imageRecognition/Operation.java index fc87c5d..13c9495 100644 --- a/src/main/java/org/wlld/imageRecognition/Operation.java +++ b/src/main/java/org/wlld/imageRecognition/Operation.java @@ -105,7 +105,7 @@ public class Operation {//进行计算 } } } else { - throw new Exception("Parameter exception"); + throw new Exception("Parameter exception region size==" + regionBodies.size()); } } diff --git a/src/main/java/org/wlld/imageRecognition/RGBNorm.java b/src/main/java/org/wlld/imageRecognition/RGBNorm.java index 19b5c04..ce4817c 100644 --- a/src/main/java/org/wlld/imageRecognition/RGBNorm.java +++ b/src/main/java/org/wlld/imageRecognition/RGBNorm.java @@ -2,8 +2,6 @@ package org.wlld.imageRecognition; import org.wlld.tools.ArithUtil; -import java.util.Arrays; - public class RGBNorm { private double[] rgbAll = new double[3]; private double norm; diff --git a/src/main/java/org/wlld/imageRecognition/border/Knn.java b/src/main/java/org/wlld/imageRecognition/border/Knn.java index 70fdcfd..a8702b0 100644 --- a/src/main/java/org/wlld/imageRecognition/border/Knn.java +++ b/src/main/java/org/wlld/imageRecognition/border/Knn.java @@ -68,105 +68,6 @@ public class Knn {//KNN分类器 } } - public int getType2(Matrix vector) throws Exception { - Map map = new HashMap<>(); - double[] sub = new double[vector.getY()];//记录最小差 - for (int t = 0; t < sub.length; t++) { - sub[t] = -1; - } - int[] types = new int[vector.getY()];//记录每个位置的最小类别 - for (Map.Entry> entry : featureMap.entrySet()) { - int type = entry.getKey(); - List matrices = entry.getValue(); - double minDist = -1; - Matrix minMatrix = null; - for (Matrix matrix : matrices) { - double dist = MatrixOperation.getEDist(matrix, vector); - if (minDist > dist || minDist == -1) { - minDist = dist; - minMatrix = matrix; - } - } - map.put(type, minMatrix); - } - //从所有最小的里面选取三名最小的 - int[] ty = getTypes(vector); - System.out.println(Arrays.toString(ty)); - //从所有最小的里面选取 - for (Map.Entry entry : map.entrySet()) { - int type = entry.getKey(); - if (isLive(ty, type)) { - Matrix matrix = entry.getValue(); - setMin(matrix, types, sub, vector, type); - } - } - System.out.println("最小位置：" + Arrays.toString(types)); - Map typeMap = new HashMap<>(); - for (int i = 0; i < types.length; i++) { - int type = types[i]; - if (typeMap.containsKey(type)) { - typeMap.put(type, typeMap.get(type) + 1); - } else { - typeMap.put(type, 1); - } - } - int type = 0; - int max = 0; - for (Map.Entry entry : typeMap.entrySet()) { - int key = entry.getKey(); - int nub = entry.getValue(); - if (nub > max) { - max = nub; - type = key; - } - } - return type; - } - - private boolean isLive(int[] types, int type) { - boolean isLive = false; - for (int ty : types) { - if (ty == type) { - isLive = true; - break; - } - } - return isLive; - } - - private void setMin(Matrix matrix, int[] types, double[] sub, Matrix vector, int type) throws Exception { - int y = vector.getY(); - for (int j = 0; j < y; j++) { - double nub = vector.getNumber(0, j); - double myNub = matrix.getNumber(0, j); - double subs = sub[j]; - double cha = Math.abs(nub - myNub); - if (subs > cha || subs == -1) { - sub[j] = cha; - types[j] = type; - } - } - } - - public int[] getTypes(Matrix vector) throws Exception { - double[] dists = new double[nub]; - // System.out.println("测试：" + vector.getString()); - int[] types = new int[nub]; - for (int i = 0; i < nub; i++) { - dists[i] = -1; - } - for (Map.Entry> entry : featureMap.entrySet()) { - int type = entry.getKey(); - List matrices = entry.getValue(); - for (Matrix matrix : matrices) { - //double dist = MatrixOperation.errorNub(vector, matrix, 0); - double dist = MatrixOperation.getEDist(matrix, vector); - compare(dists, types, dist, type); - } - } - return types; - } - public int getType(Matrix vector) throws Exception {//识别分类 int ty = 0; double[] dists = new double[nub]; diff --git a/src/main/java/org/wlld/tools/Frequency.java b/src/main/java/org/wlld/tools/Frequency.java index 6a41a20..d7bb2cd 100644 --- a/src/main/java/org/wlld/tools/Frequency.java +++ b/src/main/java/org/wlld/tools/Frequency.java @@ -59,7 +59,7 @@ public abstract class Frequency {//统计频数 return var; } - public double varianceByAve(double[] m, double ave) {//计算方差 计算方差，依赖平均值 + public double varianceByAve(double[] m, double ave) {// 计算方差，依赖平均值 double allNub = 0; for (int i = 0; i < m.length; i++) { allNub = allNub + Math.pow(m[i] - ave, 2); diff --git a/src/test/java/coverTest/FoodTest.java b/src/test/java/coverTest/FoodTest.java index 71b0457..d9cd62e 100644 --- a/src/test/java/coverTest/FoodTest.java +++ b/src/test/java/coverTest/FoodTest.java @@ -51,8 +51,8 @@ public class FoodTest { specifications.setMaxHeight(750); specificationsList.add(specifications); Operation operation = new Operation(templeConfig); - for (int i = 1; i <= 28; i++) { - ThreeChannelMatrix threeChannelMatrix1 = picture.getThreeMatrix("D:\\share\\cai\\g/g" + i + ".jpg"); + for (int i = 1; i <= 1; i++) { + ThreeChannelMatrix threeChannelMatrix1 = picture.getThreeMatrix("D:\\share\\cai\\g/g2.jpg"); List regionBody = operation.colorLook(threeChannelMatrix1, specificationsList); for (int j = 0; j < regionBody.size(); j++) { RegionBody regionBody1 = regionBody.get(j); @@ -80,7 +80,7 @@ public class FoodTest { //聚类 templeConfig.setFeatureNub(3);//聚类特征数量 //菜品识别实体类 - food.setShrink(60);//缩紧像素 + food.setShrink(50);//缩紧像素 food.setTimes(2);//聚类数据增强 templeConfig.setClassifier(Classifier.KNN); templeConfig.init(StudyPattern.Cover_Pattern, true, 400, 400, 3); @@ -124,12 +124,7 @@ public class FoodTest { operation.colorStudy(threeChannelMatrix10, 10, specificationsList); System.out.println("=======================================" + i); } - ModelParameter modelParameter = templeConfig.getModel(); - String model = JSON.toJSONString(modelParameter); - System.out.println(model); - ModelParameter modelParameter1 = JSONObject.parseObject(model, ModelParameter.class); - //templeConfig.finishStudy(); - test2(getTemple(modelParameter1)); + test2(templeConfig); } public static void study() throws Exception {