图像识别——（java）opencv(答题卡识别过程优化)

参考https://blog.csdn.net/qq_39246466/article/details/123740015

package com.acts.opencv.base;import java.util.ArrayList;
import java.util.Date;
import java.util.List;
import java.util.TreeMap;import javax.servlet.http.HttpServletResponse;import org.apache.commons.lang3.StringUtils;
import org.opencv.core.Core;
import org.opencv.core.Core.MinMaxLocResult;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
import org.opencv.core.MatOfFloat;
import org.opencv.core.MatOfInt;
import org.opencv.core.Point;
import org.opencv.core.Range;
import org.opencv.core.Rect;
import org.opencv.core.Scalar;
import org.opencv.core.Size;
import org.opencv.highgui.Highgui;
import org.opencv.imgproc.Imgproc;
import org.opencv.photo.Photo;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.RequestMapping;import com.acts.opencv.common.utils.Constants;
import com.acts.opencv.common.utils.OpenCVUtil;
import com.acts.opencv.common.web.BaseController;@Controller
@RequestMapping(value = "cardPlus")
public class CardPlusController extends BaseController {private static final Logger logger = LoggerFactory.getLogger(CardPlusController.class);/*** 答题卡识别优化* 创建者 Songer* 创建时间  2018年3月23日*/@RequestMapping(value = "answerSheet")public void answerSheet(HttpServletResponse response, String imagefile, Integer binary_thresh,String blue_red_thresh) {System.loadLibrary(Core.NATIVE_LIBRARY_NAME);logger.info("\n 完整答题卡识别");String sourcePath = Constants.PATH + imagefile;logger.info("url==============" + sourcePath);Mat sourceMat = Highgui.imread(sourcePath, Highgui.CV_LOAD_IMAGE_COLOR);long t1 = new Date().getTime();String destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk0.png";Highgui.imwrite(destPath, sourceMat);logger.info("原答题卡图片======" + destPath);// 初始图片灰度图Mat sourceMat1 = Highgui.imread(sourcePath, Highgui.CV_LOAD_IMAGE_GRAYSCALE);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk1.png";Highgui.imwrite(destPath, sourceMat1);logger.info("生成灰度图======" + destPath);// 先膨胀 后腐蚀算法，开运算消除细小杂点Mat element = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(2 * 1 + 1, 2 * 1 + 1));Imgproc.morphologyEx(sourceMat1, sourceMat1, Imgproc.MORPH_OPEN, element);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk2.png";Highgui.imwrite(destPath, sourceMat1);logger.info("生成膨胀腐蚀后的图======" + destPath);// 切割右侧和底部标记位图片Mat rightMark = new Mat(sourceMat1, new Rect(sourceMat1.cols() - 100, 0, 100, sourceMat1.rows()));destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk3.png";Highgui.imwrite(destPath, rightMark);logger.info("截取右侧定位点图======" + destPath);// 平滑处理消除噪点毛刺等等Imgproc.GaussianBlur(rightMark, rightMark, new Size(3, 3), 0);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk4.png";Highgui.imwrite(destPath, rightMark);logger.info("平滑处理后的右侧定位点图======" + destPath);// 根据右侧定位获取水平投影，并获取纵向坐标Mat matright = horizontalProjection(rightMark);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk5.png";Highgui.imwrite(destPath, matright);logger.info("右侧水平投影图======" + destPath);// 获取y坐标点，返回的是横向条状图集合List<Rect> listy = getBlockRect(matright, 1, 0);Mat footMark = new Mat(sourceMat1, new Rect(0, sourceMat1.rows() - 150, sourceMat1.cols(), 50));destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk6.png";Highgui.imwrite(destPath, footMark);logger.info("截取底部定位点图======" + destPath);Imgproc.GaussianBlur(footMark, footMark, new Size(3, 3), 0);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk7.png";Highgui.imwrite(destPath, footMark);logger.info("平滑处理后的底部定位点图======" + destPath);// 根据底部定位获取垂直投影，并获取横向坐标Mat matbootom = verticalProjection(footMark);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk8.png";Highgui.imwrite(destPath, matbootom);logger.info("底部垂直投影图======" + destPath);// 获取x坐标点，返回的是竖向的柱状图集合List<Rect> listx = getBlockRect(matbootom, 0, 0);// 高阶处理：增加HSV颜色查找，查找红色像素点Mat matRed = findColorbyHSV(sourceMat, 156, 180);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk9.png";Highgui.imwrite(destPath, matRed);logger.info("HSV找出红色像素点======" + destPath);Mat dstNoRed = new Mat(sourceMat1.rows(), sourceMat1.cols(), sourceMat1.type());dstNoRed = OpenCVUtil.dilation(sourceMat1);// Imgproc.threshold(sourceMat1, dstNoRed, 190, 255, Imgproc.THRESH_BINARY);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk10.png";Highgui.imwrite(destPath, dstNoRed);logger.info("原灰度图的图片======" + destPath);Photo.inpaint(dstNoRed, matRed, dstNoRed, 1, Photo.INPAINT_NS);// findBlackColorbyHSV(sourceMat);// for (int i = 0;i<dstNoRed.rows();i++) {// for (int j = 0; j < dstNoRed.cols(); j++) {// if(matRed.get(i, j)[0]==255){//代表识别出的红色区域// dstNoRed.put(i,j,255);// }// }// }destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk11.png";Highgui.imwrite(destPath, dstNoRed);logger.info("去除红颜色后的图片======" + destPath);Mat grayHistogram1 = getGrayHistogram(dstNoRed);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk12.png";Highgui.imwrite(destPath, grayHistogram1);logger.info("灰度直方图图片1======" + destPath);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk13.png";Mat answerMat = dstNoRed.submat(new Rect(41, 895, 278, 133));Mat grayHistogram2 = getGrayHistogram(answerMat);Highgui.imwrite(destPath, grayHistogram2);logger.info("灰度直方图图片2======" + destPath);destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk14.png";Imgproc.threshold(dstNoRed, dstNoRed, binary_thresh, 255, Imgproc.THRESH_BINARY_INV);Highgui.imwrite(destPath, dstNoRed);logger.info("去除红色基础上进行二值化======" + destPath);String redvalue = StringUtils.split(blue_red_thresh, ",")[0];String bluevalue = StringUtils.split(blue_red_thresh, ",")[1];System.out.println(bluevalue + "          " + redvalue);TreeMap<Integer, String> resultMap = new TreeMap<Integer, String>();StringBuffer resultValue = new StringBuffer();for (int no = 0; no < listx.size(); no++) {Rect rectx = listx.get(no);for (int an = 0; an < listy.size(); an++) {Rect recty = listy.get(an);Mat selectdst = new Mat(dstNoRed, new Range(recty.y, recty.y + recty.height), new Range(rectx.x,rectx.x+ rectx.width));// 本来是在每个区域内进行二值化，后来挪至了14步，整体进行二值化，因此注释掉此处2行// Mat selectdst = new Mat(select.rows(), select.cols(), select.type());// Imgproc.threshold(select, selectdst, 170, 255, Imgproc.THRESH_BINARY);// System.out.println("rectx.x, recty.y=="+rectx.x+","+recty.y+"rectx.width,recty.height=="+rectx.width+","+recty.height);double p100 = Core.countNonZero(selectdst) * 100 / (selectdst.size().area());String que_answer = getQA(no, an);Integer que = Integer.valueOf(que_answer.split("_")[0]);String answer = que_answer.split("_")[1];// System.out.println(Core.countNonZero(selectdst) + "/" + selectdst.size().area());System.out.println(que_answer + ":            " + p100);if (p100 >= Integer.valueOf(bluevalue)) {// 蓝色Core.rectangle(sourceMat, new Point(rectx.x, recty.y), new Point(rectx.x + rectx.width, recty.y+ recty.height), new Scalar(255, 0, 0), 2);// logger.info(que_answer + ":填涂");if (StringUtils.isNotEmpty(resultMap.get(que))) {resultMap.put(que, resultMap.get(que) + "," + answer);} else {resultMap.put(que, answer);}} else if (p100 > Integer.valueOf(redvalue) && p100 < Integer.valueOf(bluevalue)) {// 红色Core.rectangle(sourceMat, new Point(rectx.x, recty.y), new Point(rectx.x + rectx.width, recty.y+ recty.height), new Scalar(0, 0, 255), 2);// logger.info(que_answer + ":临界");if (StringUtils.isNotEmpty(resultMap.get(que))) {resultMap.put(que, resultMap.get(que) + ",(" + answer + ")");} else {resultMap.put(que, "(" + answer + ")");}} else {// 绿色Core.rectangle(sourceMat, new Point(rectx.x, recty.y), new Point(rectx.x + rectx.width, recty.y+ recty.height), new Scalar(0, 255, 0), 1);// logger.info(que_answer + ":未涂");}}}// for (Object result : resultMap.keySet()) {for (int i = 1; i <= 100; i++) {// logger.info("key=" + result + " value=" + resultMap.get(result));resultValue.append("  " + i + "=" + (StringUtils.isEmpty(resultMap.get(i)) ? "未填写" : resultMap.get(i)));if (i % 5 == 0) {resultValue.append("<br>");}}destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk15.png";Highgui.imwrite(destPath, sourceMat);logger.info("框选填图区域，绿色为选项，蓝色为填图，红色为临界======" + destPath);long t2 = new Date().getTime();System.out.println(t2 - t1);// logger.info("输出最终结果：" + resultValue.toString());renderString(response, resultValue.toString());}/*** 绘制灰度直方图用于调整识别区域阈值判断* @Author 王嵩* @param 输入Mat对象img* @return Mat* @Date 2018年3月28日* 更新日志* 2018年3月28日 王嵩  首次创建**/public Mat getGrayHistogram(Mat img) {List<Mat> images = new ArrayList<Mat>();images.add(img);MatOfInt channels = new MatOfInt(0); // 图像通道数，0表示只有一个通道MatOfInt histSize = new MatOfInt(256); // CV_8U类型的图片范围是0~255，共有256个灰度级Mat histogramOfGray = new Mat(); // 输出直方图结果，共有256行，行数的相当于对应灰度值，每一行的值相当于该灰度值所占比例MatOfFloat histRange = new MatOfFloat(0, 255);Imgproc.calcHist(images, channels, new Mat(), histogramOfGray, histSize, histRange, false); // 计算直方图MinMaxLocResult minmaxLoc = Core.minMaxLoc(histogramOfGray);// 按行归一化// Core.normalize(histogramOfGray, histogramOfGray, 0, histogramOfGray.rows(), Core.NORM_MINMAX, -1, new Mat());// 创建画布int histImgRows = 600;int histImgCols = 1300;System.out.println("---------" + histSize.get(0, 0)[0]);int colStep = (int) Math.floor(histImgCols / histSize.get(0, 0)[0]);// 舍去小数，不能四舍五入，有可能列宽不够Mat histImg = new Mat(histImgRows, histImgCols, CvType.CV_8UC3, new Scalar(255, 255, 255)); // 重新建一张图片，绘制直方图int max = (int) minmaxLoc.maxVal;System.out.println("--------" + max);double bin_u = (double) (histImgRows - 20) / max; // max: 最高条的像素个数，则 bin_u 为单个像素的高度，因为画直方图的时候上移了20像素，要减去int kedu = 0;for (int i = 1; kedu <= minmaxLoc.maxVal; i++) {kedu = i * max / 10;// 在图像中显示文本字符串Core.putText(histImg, kedu + "", new Point(0, histImgRows - kedu * bin_u), 1, 1, new Scalar(0, 0, 0));}for (int i = 0; i < histSize.get(0, 0)[0]; i++) { // 画出每一个灰度级分量的比例，注意OpenCV将Mat最左上角的点作为坐标原点// System.out.println(i + ":=====" + histogramOfGray.get(i, 0)[0]);Core.rectangle(histImg, new Point(colStep * i, histImgRows - 20), new Point(colStep * (i + 1), histImgRows- bin_u * Math.round(histogramOfGray.get(i, 0)[0]) - 20),new Scalar(0, 0, 0), 1, 8, 0);kedu = i * 10;// 每隔10画一下刻度Core.rectangle(histImg, new Point(colStep * kedu, histImgRows - 20), new Point(colStep * (kedu + 1),histImgRows - 20), new Scalar(255, 0, 0), 2, 8, 0);Core.putText(histImg, kedu + "", new Point(colStep * kedu, histImgRows - 5), 1, 1, new Scalar(255, 0, 0)); // 附上x轴刻度}return histImg;}// 获取题号及选项填涂情况public String getQA(int no,int an){//返回1A、1B、1C...2A类似这样的返回值int first = no + 1 + an / 4 * 20;String second = "";if (an % 4 == 0) {second = "A";} else if (an % 4 == 1) {second = "B";} else if (an % 4 == 2) {second = "C";} else if (an % 4 == 3) {second = "D";}return first + "_" + second;}public static void main(String[] args) {System.out.println(5 / 3);}/*** 红色色系0-20，160-180* 蓝色色系100-120* 绿色色系60-80* 黄色色系23-38* 识别出的颜色会标记为白色，其他的为黑色* @param min* @param max*/public static Mat findColorbyHSV(Mat source, int min, int max) {Mat hsv_image = new Mat();Imgproc.GaussianBlur(source, source, new Size(3, 3), 0, 0);Imgproc.cvtColor(source, hsv_image, Imgproc.COLOR_BGR2HSV);// String imagenameb = "D:\\test\\testImge\\ttbefore.jpg";// Highgui.imwrite(imagenameb, hsv_image);Mat thresholded = new Mat();Core.inRange(hsv_image, new Scalar(min, 90, 90), new Scalar(max, 255, 255), thresholded);return thresholded;}/*** 查找黑色* @param source* @param min* @param max* @return*/public static Mat findBlackColorbyHSV(Mat source) {Mat hsv_image = new Mat();Imgproc.GaussianBlur(source, source, new Size(3, 3), 0, 0);Imgproc.cvtColor(source, hsv_image, Imgproc.COLOR_BGR2HSV);String imagenameb = "D:\\test\\testImge\\ttbefore.jpg";Highgui.imwrite(imagenameb, hsv_image);Mat thresholded = new Mat();Core.inRange(hsv_image, new Scalar(0, 0, 0), new Scalar(180, 255, 46), thresholded);String ttblack = "D:\\test\\testImge\\ttblack.jpg";Highgui.imwrite(ttblack, thresholded);return thresholded;}/*** 水平投影* @param source 传入灰度图片Mat* @return*/public static Mat horizontalProjection(Mat source) {Mat dst = new Mat(source.rows(), source.cols(), source.type());// 先进行反转二值化Imgproc.threshold(source, dst, 150, 255, Imgproc.THRESH_BINARY_INV);// 水平积分投影// 每一行的白色像素的个数int[] rowswidth = new int[dst.rows()];for (int i = 0; i < dst.rows(); i++) {for (int j = 0; j < dst.cols(); j++) {if (dst.get(i, j)[0] == 255) {rowswidth[i]++;}}}// 定义一个白色跟原图一样大小的画布Mat matResult = new Mat(dst.rows(), dst.cols(), CvType.CV_8UC1, new Scalar(255, 255, 255));// 将每一行按照行像素值大小填充像素宽度for (int i = 0; i < matResult.rows(); i++) {for (int j = 0; j < rowswidth[i]; j++) {matResult.put(i, j, 0);}}return matResult;}/*** 垂直投影* @param source 传入灰度图片Mat* @return*/public static Mat verticalProjection(Mat source) {// 先进行反转二值化Mat dst = new Mat(source.rows(), source.cols(), source.type());Imgproc.threshold(source, dst, 150, 255, Imgproc.THRESH_BINARY_INV);// 垂直积分投影// 每一列的白色像素的个数int[] colswidth = new int[dst.cols()];for (int j = 0; j < dst.cols(); j++) {for (int i = 0; i < dst.rows(); i++) {if (dst.get(i, j)[0] == 255) {colswidth[j]++;}}}Mat matResult = new Mat(dst.rows(), dst.cols(), CvType.CV_8UC1, new Scalar(255, 255, 255));// 将每一列按照列像素值大小填充像素宽度for (int j = 0; j < matResult.cols(); j++) {for (int i = 0; i < colswidth[j]; i++) {matResult.put(matResult.rows() - 1 - i, j, 0);}}return matResult;}/*** 图片切块* @param srcMat 传入水平或垂直投影的图片对象Mat* @param proType 传入投影Mat对象的 投影方式0：垂直投影图片,竖向切割；1：水平投影图片，横向切割* @param rowY 由于传来的是可能是原始图片的部分切片，要计算切块的实际坐标位置需要给出切片时所在的坐标，所以需要传递横向切片的y坐标或者纵向切片的横坐标* 如当proType==0时，传入的是切片的垂直投影，那么切成块后能得出x坐标及块宽高度，但是实际y坐标需要加上原切片的y坐标值，所以rowXY为切片的y坐标点，* 同理当proType==1时，rowXY应该为x坐标* @return*/public static List<Rect> getBlockRect(Mat srcImg, Integer proType, int rowXY) {Imgproc.threshold(srcImg, srcImg, 150, 255, Imgproc.THRESH_BINARY_INV);// 注意 countNonZero 方法是获取非0像素（白色像素）数量，所以一般要对图像进行二值化反转List<Rect> rectList = new ArrayList<Rect>();int size = proType == 0 ? srcImg.cols() : srcImg.rows();int[] pixNum = new int[size];if (proType == 0) {for (int i = 0; i < srcImg.cols(); i++) {Mat col = srcImg.col(i);pixNum[i] = Core.countNonZero(col) > 1 ? Core.countNonZero(col) : 0;}} else {// 水平投影只关注行for (int i = 0; i < srcImg.rows(); i++) {Mat row = srcImg.row(i);pixNum[i] = Core.countNonZero(row) > 1 ? Core.countNonZero(row) : 0;}}int startIndex = 0;// 记录进入字符区的索引int endIndex = 0;// 记录进入空白区域的索引boolean inBlock = false;// 是否遍历到了字符区内for (int i = 0; i < size; i++) {if (!inBlock && pixNum[i] != 0) {// 进入字符区，上升跳变沿inBlock = true;startIndex = i;} else if (pixNum[i] == 0 && inBlock) {// 进入空白区，下降跳变沿存储endIndex = i;inBlock = false;Rect rect = null;if (proType == 0) {rect = new Rect(startIndex, rowXY, (endIndex - startIndex), srcImg.rows());} else {rect = new Rect(rowXY, startIndex, srcImg.cols(), (endIndex - startIndex));}rectList.add(rect);}}return rectList;}}

package com.acts.opencv.base;import java.io.File;
import java.util.Vector;import javax.servlet.http.HttpServletResponse;import net.sourceforge.tess4j.Tesseract;
import net.sourceforge.tess4j.TesseractException;import org.opencv.core.Core;
import org.opencv.core.Core.MinMaxLocResult;
import org.opencv.core.CvType;
import org.opencv.core.Mat;
import org.opencv.core.MatOfPoint;
import org.opencv.core.Point;
import org.opencv.core.Rect;
import org.opencv.core.Scalar;
import org.opencv.highgui.Highgui;
import org.opencv.imgproc.Imgproc;
import org.slf4j.Logger;
import org.slf4j.LoggerFactory;
import org.springframework.stereotype.Controller;
import org.springframework.web.bind.annotation.RequestMapping;import com.acts.opencv.common.utils.Constants;
import com.acts.opencv.common.utils.OpenCVUtil;
import com.acts.opencv.common.web.BaseController;@Controller
@RequestMapping(value = "page")
public class PageController extends BaseController {private static final Logger logger = LoggerFactory.getLogger(PageController.class);/*** 答题卡识别优化* 创建者 Songer* 创建时间  2018年3月23日*/@RequestMapping(value = "pageOCR")public void pageOCR(HttpServletResponse response, String imagefile, Integer ocrType) {System.loadLibrary(Core.NATIVE_LIBRARY_NAME);logger.info("\n 页码识别");String sourcePath = Constants.PATH + imagefile;logger.info("url==============" + sourcePath);// Mat sourceMat = Highgui.imread(sourcePath, Highgui.CV_LOAD_IMAGE_COLOR);// String destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "dtk0.png";// Highgui.imwrite(destPath, sourceMat);String result = "";if (ocrType == 1) {// tesseract ocr 识别方式result = getPageNoByTesseract(sourcePath);} else if (ocrType == 2) {// 轮廓识别result = getPageNoByContours(sourcePath);} else {result = getPageNoByTemplate(sourcePath);}renderString(response, result);}// public static void main(String[] args) {// long t1 = new Date().getTime();// try {// File imageFile = new File("D:\\test\\testImge\\t3.bmp");// Tesseract instance = new Tesseract(); // JNA Interface Mapping// instance.setLanguage("chi_sim");// String result = instance.doOCR(imageFile);// System.out.println("result=====" + result);// long t2 = new Date().getTime();// System.out.println((t2 - t1));// } catch (TesseractException e) {// e.printStackTrace();// }// }/*** 使用tesseract方式识别页码，注意tessdata放到tomcat的bin目录下* @Author 王嵩* @param filePath* @return String* @Date 2018年4月4日* 更新日志* 2018年4月4日 王嵩  首次创建**/public String getPageNoByTesseract(String filePath) {String result = "";try {File file = new File(filePath);Tesseract instance = new Tesseract(); // JNA Interface Mappinginstance.setLanguage("chi_sim");//result = instance.doOCR(file);logger.info("result====={}", result);} catch (TesseractException e) {e.printStackTrace();}return result;}/*** 使用轮廓识别页码* @Author 王嵩* @param filePath* @return String* @Date 2018年4月8日* 更新日志* 2018年4月8日 王嵩  首次创建**/public String getPageNoByContours(String filePath) {Mat source = Highgui.imread(filePath, Highgui.CV_LOAD_IMAGE_GRAYSCALE);System.out.println("row" + source.rows() + " col " + source.cols());// 二值化反转Mat bininv = new Mat(source.rows(), source.cols(), source.type());Imgproc.threshold(source, bininv, 170, 255, Imgproc.THRESH_BINARY_INV);// 腐蚀膨胀，用于消除噪点和干扰项Mat destination = OpenCVUtil.eroding(bininv);destination = OpenCVUtil.dilation(destination);// 轮廓识别Vector<MatOfPoint> contours = OpenCVUtil.findContours(destination);int pageSize = 0;// 原颜色图片加载，用于画出识别轮廓，实际开发不需要Mat image = Highgui.imread(filePath, Highgui.CV_LOAD_IMAGE_COLOR);for (int i = 0; i < contours.size(); i++) {Mat result = new Mat(destination.size(), CvType.CV_8UC3, new Scalar(255, 255, 255));Imgproc.drawContours(result, contours, i, new Scalar(0, 0, 255), 1);MatOfPoint mop = contours.get(i);// 获取轮廓面积double contArea = Math.abs(Imgproc.contourArea(mop, false));Rect r = Imgproc.boundingRect(mop);System.out.println("轮廓面积：" + contArea);if (contArea > 1200) {// 此处是根据轮廓面积// 红线画出识别的轮廓Core.rectangle(image, new Point(r.x, r.y), new Point(r.x + r.width, r.y+ r.height), new Scalar(0, 0, 255), 2);pageSize++;}}String destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "page0.png";Highgui.imwrite(destPath, image);System.out.println("页码为：" + pageSize);return pageSize + "";}/*** 使用模板匹配识别页码* @Author 王嵩* @param filePath* @return String* @Date 2018年4月8日* 更新日志* 2018年4月8日 王嵩  首次创建*/public String getPageNoByTemplate(String filePath) {String pageSize = "";Mat source = Highgui.imread(filePath, Highgui.CV_LOAD_IMAGE_GRAYSCALE);// 二值化反转Mat bininv = new Mat(source.rows(), source.cols(), source.type());Imgproc.threshold(source, bininv, 170, 255, Imgproc.THRESH_BINARY_INV);Vector<MatOfPoint> contours1 = OpenCVUtil.findContours(bininv);MatOfPoint mop = contours1.get(0);Rect rect = Imgproc.boundingRect(mop);Mat matchtemp = source.submat(rect.y, rect.y + rect.height, rect.x, rect.x + rect.width);String page_temp = Constants.PATH + Constants.DEST_IMAGE_PATH + "page_temp.png";Highgui.imwrite(page_temp, matchtemp);String pagePath = Constants.PATH + Constants.SOURCE_IMAGE_PATH + "shuzi.png";Mat pageimage = Highgui.imread(pagePath, Highgui.CV_LOAD_IMAGE_GRAYSCALE);int result_cols = pageimage.cols() - matchtemp.cols() + 1;int result_rows = pageimage.rows() - matchtemp.rows() + 1;Mat destination = new Mat(result_rows, result_cols, CvType.CV_32FC1);Imgproc.matchTemplate(pageimage, matchtemp, destination, Imgproc.TM_CCOEFF);// 矩阵归一化处理Core.normalize(destination, destination, 0, 255, Core.NORM_MINMAX, -1, new Mat());MinMaxLocResult minmaxLoc = Core.minMaxLoc(destination);Point matchLoc = minmaxLoc.maxLoc;Core.rectangle(pageimage, matchLoc, new Point(matchLoc.x + matchtemp.cols(), matchLoc.y + matchtemp.rows()),new Scalar(0), 2);System.out.println(matchLoc.x + "   " + matchLoc.y);pageSize = getPage(matchLoc.x) + "";String destPath = Constants.PATH + Constants.DEST_IMAGE_PATH + "page1.png";Highgui.imwrite(destPath, pageimage);return pageSize;}/*** 根据横坐标返回页码* @Author 王嵩* @param x* @return int* @Date 2018年4月8日* 更新日志* 2018年4月8日 王嵩  首次创建**/public int getPage(double x) {// 减去2像素，是因为shuzi.png外边框是预留了2像素的，因此匹配结果坐标为：2,2;72,2;142,2// Math.floor 返回不大于的最大整数return (int) Math.floor((x - 2) / 70) + 1;}public static void main(String[] args) {System.loadLibrary(Core.NATIVE_LIBRARY_NAME);Mat markmat1 = Highgui.imread("D:\\test\\abc\\1.png", Highgui.CV_LOAD_IMAGE_GRAYSCALE);Mat markmat2 = Highgui.imread("D:\\test\\abc\\2.png", Highgui.CV_LOAD_IMAGE_GRAYSCALE);Vector<MatOfPoint> contours1 = OpenCVUtil.findContours(markmat1);Vector<MatOfPoint> contours2 = OpenCVUtil.findContours(markmat2);Mat mat1 = getSimMark("D:\\test\\abc\\1.png");Mat mat2 = getSimMark("D:\\test\\abc\\2.png");double result1 = Imgproc.matchShapes(contours1.get(0), contours2.get(0), Imgproc.CV_CONTOURS_MATCH_I1, 0);double result2 = Imgproc.matchShapes(mat1, mat2, Imgproc.CV_CONTOURS_MATCH_I1, 0);System.out.println(result1);System.out.println(result2);}private static MatOfPoint getSimMark(String path) {Mat markmat = Highgui.imread(path, Highgui.CV_LOAD_IMAGE_GRAYSCALE);// Imgproc.threshold(markmat, markmat, 190, 255, Imgproc.THRESH_BINARY_INV);Vector<MatOfPoint> contours = new Vector<MatOfPoint>();Mat rsmat = new Mat();Imgproc.findContours(markmat, contours, rsmat, Imgproc.RETR_TREE, Imgproc.CHAIN_APPROX_SIMPLE, new Point());MatOfPoint markMop = new MatOfPoint();Mat result = new Mat(markmat.size(), CvType.CV_8U, new Scalar(255));Imgproc.drawContours(result, contours, 0, new Scalar(0), 1);String image1 = "d:\\test\\abc\\t.jpg";Highgui.imwrite(image1, result);for (int i = 0; i < contours.size(); i++) {markMop = contours.get(0);// MatOfPoint2f mat2f = new MatOfPoint2f();// MatOfPoint2f dstmat2f = new MatOfPoint2f();// markMop.convertTo(mat2f, CvType.CV_32FC1);// // 多边形逼近算法,减少轮廓的顶点，便于对比// Imgproc.approxPolyDP(mat2f, dstmat2f, markMop.total() * 0.02, true);// dstmat2f.convertTo(markMop, CvType.CV_32S);}return markMop;}
}

静态资源类

package com.acts.opencv.common.utils;import org.springframework.web.context.ContextLoader;/*** 常量 创建者 Songer 创建时间 2018年3月09日**/
public class Constants {public static final String CURRENT_USER = "UserInfo";public static final String WECHAT_USER = "weChatUserInfo";public static final String REFERENCE_CODE = "referenceCode";public static final String SUCCESS = "success";public static final String ERROR = "error";public static final String SF_FILE_SEPARATOR = System.getProperty("file.separator");// 文件分隔符public static final String SF_LINE_SEPARATOR = System.getProperty("line.separator");// 行分隔符public static final String SF_PATH_SEPARATOR = System.getProperty("path.separator");// 路径分隔符public static final String PATH = ContextLoader.getCurrentWebApplicationContext().getServletContext().getRealPath("/");/*** 文件*/public static final String SOURCE_IMAGE_PATH = Constants.SF_FILE_SEPARATOR + "statics"+ Constants.SF_FILE_SEPARATOR + "sourceimage" + Constants.SF_FILE_SEPARATOR;// 图片原地址public static final String DEST_IMAGE_PATH = Constants.SF_FILE_SEPARATOR + "statics" + Constants.SF_FILE_SEPARATOR+ "destimage" + Constants.SF_FILE_SEPARATOR;// 图片生成地址/*** 返回参数规范*//** 区分类型 1 -- 无错误，Code重复 */public static final String CODE_DUPLICATE = "1";/** 区分类型 2 -- 无错误，名称重复 */public static final String NAME_DUPLICATE = "2";/** 区分类型 3 -- 数量超出 */public static final String NUMBER_OVER = "3";/** 区分类型 0 -- 无错误，程序正常执行 */public static final String NO_ERROR = "0";/** 区分类型 -1 -- 无错误，返回结果为空 */public static final String NULL_POINTER = "-1";/** 区分类型 -2 -- 错误，参数不正确 */public static final String INCORRECT_PARAMETER = "-2";/** 区分类型 -3 -- 错误，程序执行错误 */public static final String PROGRAM_EXECUTION_ERROR = "-3";/** 区分类型 -5 -- 错误，数据已删除 */public static final String DATA_DELETED = "-5";/** 区分类型 -6 -- 错误，参数不一致（验证码） */public static final String DATA_NOT_SAME = "-6";/**json文件缺失  */public static final String NO_JSON_FILE = "-7";/*** 分页中可能用到的常量*/public static final Integer PAGE_SIZE=10;//一页共有十条内容}

opencv工具类

package com.acts.opencv.common.utils;import java.awt.image.BufferedImage;
import java.awt.image.DataBufferByte;
import java.io.IOException;
import java.util.Date;
import java.util.Vector;import org.opencv.core.Mat;
import org.opencv.core.MatOfByte;
import org.opencv.core.MatOfPoint;
import org.opencv.core.Point;
import org.opencv.core.Size;
import org.opencv.highgui.Highgui;
import org.opencv.imgproc.Imgproc;public class OpenCVUtil {public static BufferedImage covertMat2Buffer(Mat mat) throws IOException {long time1 = new Date().getTime();// Mat 转byte数组BufferedImage originalB = toBufferedImage(mat);long time3 = new Date().getTime();System.out.println("保存读取方法2转=" + (time3 - time1));return originalB;// ImageIO.write(originalB, "jpg", new File("D:\\test\\testImge\\ws2.jpg"));}public static byte[] covertMat2Byte(Mat mat) throws IOException {long time1 = new Date().getTime();// Mat 转byte数组byte[] return_buff = new byte[(int) (mat.total() * mat.channels())];Mat mat1 = new Mat();mat1.get(0, 0, return_buff);long time3 = new Date().getTime();System.out.println(mat.total() * mat.channels());System.out.println("保存读取方法2转=" + (time3 - time1));return return_buff;}public static byte[] covertMat2Byte1(Mat mat) throws IOException {long time1 = new Date().getTime();MatOfByte mob = new MatOfByte();Highgui.imencode(".jpg", mat, mob);long time3 = new Date().getTime();// System.out.println(mat.total() * mat.channels());System.out.println("Mat转byte[] 耗时=" + (time3 - time1));return mob.toArray();}public static BufferedImage toBufferedImage(Mat m) {int type = BufferedImage.TYPE_BYTE_GRAY;if (m.channels() > 1) {type = BufferedImage.TYPE_3BYTE_BGR;}int bufferSize = m.channels() * m.cols() * m.rows();byte[] b = new byte[bufferSize];m.get(0, 0, b); // get all the pixelsBufferedImage image = new BufferedImage(m.cols(), m.rows(), type);final byte[] targetPixels = ((DataBufferByte) image.getRaster().getDataBuffer()).getData();System.arraycopy(b, 0, targetPixels, 0, b.length);return image;}/*** 腐蚀膨胀是针对于白色区域来说的，腐蚀即腐蚀白色区域* 腐蚀算法（黑色区域变大）* @param source* @return*/public static Mat eroding(Mat source) {return eroding(source, 1);}public static Mat eroding(Mat source, double erosion_size) {Mat resultMat = new Mat(source.rows(), source.cols(), source.type());Mat element = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(2 * erosion_size + 1,2 * erosion_size + 1));Imgproc.erode(source, resultMat, element);return resultMat;}/*** 腐蚀膨胀是针对于白色区域来说的，膨胀是膨胀白色区域* 膨胀算法（白色区域变大）* @param source* @return*/public static Mat dilation(Mat source) {return dilation(source, 1);}/*** 腐蚀膨胀是针对于白色区域来说的，膨胀是膨胀白色区域* @Author 王嵩* @param source* @param dilationSize 膨胀因子2*x+1 里的x* @return Mat* @Date 2018年2月5日* 更新日志* 2018年2月5日 王嵩  首次创建**/public static Mat dilation(Mat source, double dilation_size) {Mat resultMat = new Mat(source.rows(), source.cols(), source.type());Mat element = Imgproc.getStructuringElement(Imgproc.MORPH_RECT, new Size(2 * dilation_size + 1,2 * dilation_size + 1));Imgproc.dilate(source, resultMat, element);return resultMat;}/*** 轮廓识别，使用最外轮廓发抽取轮廓RETR_EXTERNAL，轮廓识别方法为CHAIN_APPROX_SIMPLE* @param source 传入进来的图片Mat对象* @return 返回轮廓结果集*/public static Vector<MatOfPoint> findContours(Mat source) {Mat rs = new Mat();/*** 定义轮廓抽取模式*RETR_EXTERNAL:只检索最外面的轮廓;*RETR_LIST:检索所有的轮廓，并将其放入list中;*RETR_CCOMP:检索所有的轮廓，并将他们组织为两层:顶层是各部分的外部边界，第二层是空洞的边界;*RETR_TREE:检索所有的轮廓，并重构嵌套轮廓的整个层次。*/int mode = Imgproc.RETR_EXTERNAL;// int mode = Imgproc.RETR_TREE;/*** 定义轮廓识别方法* 边缘近似方法(除了RETR_RUNS使用内置的近似，其他模式均使用此设定的近似算法)。可取值如下:*CV_CHAIN_CODE:以Freeman链码的方式输出轮廓，所有其他方法输出多边形(顶点的序列)。*CHAIN_APPROX_NONE:将所有的连码点，转换成点。*CHAIN_APPROX_SIMPLE:压缩水平的、垂直的和斜的部分，也就是，函数只保留他们的终点部分。*CHAIN_APPROX_TC89_L1，CV_CHAIN_APPROX_TC89_KCOS:使用the flavors of Teh-Chin chain近似算法的一种。*LINK_RUNS:通过连接水平段的1，使用完全不同的边缘提取算法。使用CV_RETR_LIST检索模式能使用此方法。*/int method = Imgproc.CHAIN_APPROX_SIMPLE;Vector<MatOfPoint> contours = new Vector<MatOfPoint>();Imgproc.findContours(source, contours, rs, mode, method, new Point());return contours;}
}

图像识别——（java）opencv(答题卡识别过程优化)相关推荐

图像识别——（java使用opencv答题卡识别)
具体算法实现:具体算法实现--opencvhttps://blog.csdn.net/qq_39246466/article/details/123819795 1.安装: 配置linux或windo ...
基于EmguCv圆形答题卡识别的优化
主要对答题卡通用性方面进行了优化(可不限于5*5的答题卡,对图片规格进行统一,可自动计算轮廓间距),并对一些容易出现的bug进行了修复.还需要手动控制统一后的图片大小规格及测量需要检测到的轮廓大小范围 ...
基于 Java 的答题卡识别系统
这张答题卡想必伴随了大家的学生时代,不管是中考.高考.考研,都会用 2B 铅笔涂好卡上的红框框. 今天推荐一个基于 Spring MVC + OpenCV 的答题卡识别系统,OpenCV 是一个计算机 ...
opencv答题卡识别 (一)
背景:答题卡阅卷需要光标阅读机,有些小学校买不起光标阅读机. 主要开源库:opencv,版本3.0. 识别原理:把答题卡放在深色背景中,用查找轮廓定位好答题卡位置,用透视变换取出答题卡图像,根据位置判 ...
OpenCV 答题卡识别
1.预处理.轮廓检测 import cv2 import numpy as np # 正确答案 ANSWER_KEY = {0:1,1:4,2:0,3:3,4:1} def cv_show(name, ...
python OpenCV 答题卡识别判卷
完整代码: #导入工具包 import numpy as np import argparse import imutils import cv2# 设置参数 ap = argparse.Argume ...
用Python+OpenCV+PyQt开发的答题卡识别软件
用Python+OpenCV+PyQt开发的答题卡识别软件软件使用说明软件设计思路如何设置答案界面风格备注这是一个可以识别定制答题卡的软件,它可以根据用户自定的答案来进行识别,校对正误并统 ...
基于 SpringMvc＋OpenCV 实现的答题卡识别系统（附源码）
java_opencv 项目介绍 OpenCV是一个基于BSD许可(开源)发行的跨平台计算机视觉库,它提供了一系列图像处理和计算机视觉方面很多通用算法.是研究图像处理技术的一个很不错的工具.最初开始接 ...
基于 SpringMvc + OpenCV 实现的答题卡识别系统（附源码）
点击关注公众号,实用技术文章及时了解 java_opencv 项目介绍 OpenCV是一个基于BSD许可(开源)发行的跨平台计算机视觉库,它提供了一系列图像处理和计算机视觉方面很多通用算法.是研究图像 ...
答题卡识别任务--opencv python（附代码）
答题卡识别项目理论和源码来自唐宇迪opencv项目实战记一篇python-opencv 完成答题卡识别项目的学习笔记输入一张特定格式的答题卡图片(答题卡中题目数量和选项个数是固定的),能够输出 ...

图像识别——（java）opencv(答题卡识别过程优化)

图像识别——（java）opencv(答题卡识别过程优化)相关推荐

最新文章

热门文章