网站规划与建设课程设计,深圳哪里有可以做网站跳转的公司,长春专用网站建设,广东小程序系统开发最近有个需求#xff0c;希望识别图片上的虫子#xff0c;对于java来说#xff0c;图像识别不是很好做。在网上也搜索了很多#xff0c;很多的代码都是不完整#xff0c;或者下载下载报错#xff0c;有的写的很长看不懂。所以自己试着用java的opencv写了一段代码。发现识…最近有个需求希望识别图片上的虫子对于java来说图像识别不是很好做。在网上也搜索了很多很多的代码都是不完整或者下载下载报错有的写的很长看不懂。所以自己试着用java的opencv写了一段代码。发现识别的效果还不错下面把代码贴出来。有需要的可以参考。但是这里面有一些缺陷就是没有加入transformer和org.deeplearning4j对于复杂的场景识别的不是很准确。后期再更新把神经网络加上去。
一、POM.xml文件导入jar包
dependencygroupIdorg.bytedeco/groupIdartifactIdjavacv-platform/artifactIdversion1.5.10/version !-- 根据最新版本更新 --/dependencydependencygroupIdorg.opencv/groupIdartifactIdopencv/artifactIdversion4.9.0/version/dependencydependencygroupIdorg.deeplearning4j/groupIdartifactIddeeplearning4j-core/artifactIdversion1.0.0-M1.1/version/dependencydependencygroupIdorg.nd4j/groupIdartifactIdnd4j-native/artifactIdversion1.0.0-M2/version/dependencydependencygroupIdorg.nd4j/groupIdartifactIdnd4j-api/artifactIdversion1.0.0-M2/version/dependency
二、主要的处理步骤和逻辑代码
读取图片截取区域转换灰度二值化高斯中指滤波双边滤波去除线框干扰形态学操作轮廓检测
package org.example;import org.bytedeco.opencv.global.opencv_imgcodecs;
import org.bytedeco.opencv.global.opencv_imgproc;
import org.bytedeco.opencv.opencv_core.*;public class BugCounterTest {public static void main(String[] args) {// 读取图片文件Mat src opencv_imgcodecs.imread(C:\\Users\\HP\\Desktop\\aaaa.png);if (src.empty()) {System.out.println(Error: Cannot read image!);return;}// 截取感兴趣区域Rect roi new Rect(0, 180, 1300, 600);Mat croppedImage new Mat(src, roi);// 转换为灰度图像Mat gray new Mat();opencv_imgproc.cvtColor(croppedImage, gray, opencv_imgproc.COLOR_BGR2GRAY);// 二值化图像Mat binary new Mat();opencv_imgproc.threshold(gray, binary, 100, 255, opencv_imgproc.THRESH_BINARY_INV);//高斯模糊处理Mat blurredImage new Mat();opencv_imgproc.GaussianBlur(binary,blurredImage,new Size(5, 5),0);//中值滤波Mat medianFilteredImage new Mat();opencv_imgproc.medianBlur(blurredImage, medianFilteredImage, 5);// 双边滤波Mat bilateralFilteredImage new Mat();opencv_imgproc.bilateralFilter(medianFilteredImage, bilateralFilteredImage, 9, 75, 75);// 去除线框干扰Mat edgeImage new Mat();opencv_imgproc.Canny(bilateralFilteredImage, edgeImage, 50, 150); // 可调整参数// 形态学操作Mat kernel opencv_imgproc.getStructuringElement(opencv_imgproc.MORPH_RECT, new Size(3, 3));opencv_imgproc.dilate(edgeImage, edgeImage, kernel);opencv_imgproc.erode(edgeImage, edgeImage, kernel);// 轮廓检测MatVector contours new MatVector();Mat hierarchy new Mat();opencv_imgproc.findContours(edgeImage, contours, hierarchy, opencv_imgproc.RETR_EXTERNAL, opencv_imgproc.CHAIN_APPROX_SIMPLE);int blackPointsCount 0;// 在原始图像上绘制轮廓for (int i 0; i contours.size(); i) {Rect rect opencv_imgproc.boundingRect(contours.get(i));Scalar scalar new Scalar(0, 255, 0, 0);opencv_imgproc.rectangle(croppedImage, rect, scalar);if (rect.width() 1 rect.height() 1) {blackPointsCount;}}// 保存标记后的图像opencv_imgcodecs.imwrite(C:\\Users\\HP\\Desktop\\output.jpg, croppedImage);System.out.println(黑点数量 blackPointsCount);}
}对于复杂的图片识别有差距
