opencv源码地址:https://github.com/opencv/opencv
下载4.6.0版本:Releases · opencv/opencv · GitHub
从git上找到对应版本号下载zip文件
上传到Linux中并解压
unzip opencv-4.6.0.zip mv opencv-4.6.0 /cloud/ cd /cloud/opencv-4.6.0/ mkdir build && cd build
接下来就是需要编译源码,这时候需要cmake命令,如果Linux自带的cmake版本低不好使或者是本机并没有装cmake那需要提前准备一下
wget https://cmake.org/files/v3.9/cmake-3.9.2.tar.gz tar zxvf cmake-3.9.2.tar.gz mv cmake-3.9.2 /cloud/ cd /cloud/cmake-3.9.2/ ./configure make make install cmake -version # 检查一下版本,如果报错,可能是本机原来的cmake还有缓存,没替换过来需要执行以下 hash -r命令 # hash -r # 这时候如果显示版本号为3.9.2就对了
cmake执行编译源码过程中会用到java环境和ant,没有的话也需要提前准备
yum -y install ant ant -version
接下来编译源码
cd /cloud/opencv-4.6.0/build cmake -D BUILD_SHARED_LIBS=OFF -D BUILD_TEST=OFF ..
出现下面的ant: /bin/ant (ver 1.9.4)和java的信息说明命令执行成功过
-- Java: -- ant: /bin/ant (ver 1.9.4) -- JNI: /usr/java/jdk1.8.0_121/include /usr/java/jdk1.8.0_121/include/linux /usr/java/jdk1.8.0_121/include -- Java wrappers: YES -- Java tests: YES -- -- Install to: /usr/local -- ----------------------------------------------------------------- --
接下来执行make命令,执行完成后会出现对应的so包和jar包
# 8核cpu 就写 j8 make -j8 & make
生成的libopencv_java460.so包位置
cd /cloud/opencv-4.6.0/build/lib
生成的opencv-460.jar包位置
cd /cloud/opencv-4.6.0/build/bin
生成的lbpcascade_frontalface.xml文件位置
cd /cloud/opencv-4.6.0/build/java_test/res/raw
准备工作完成
spring集成把上一步的三个文件放到项目中,
因为我是maven管理的多模块项目,
人脸识别模块在人脸识别模块添加jar包,把jar放到
人脸识别模块/src/main/resources/lib下
如果没有lib那就新建一个lib文件夹
pom引入jar依赖
com.opencv opencv_java4601 system ${basedir}/src/main/resources/lib/opencv-460.jar
在build 的resources下忽略掉这个包,因为在war项目包中会引用,人脸识别模块就不要引用了
war模块src/main/resources ** @Controller public class FaceDetectorController { // 初始化人脸探测器 static CascadeClassifier faceDetector; static { boolean isWin = false; if (System.getProperty("os.name").toLowerCase().startsWith("win")) { isWin = true; } if (!isWin) { //路径不能包含中文 linux使用libopencv_java410.so String opencvDllName = null; try { opencvDllName = FaceDetectorController.class.getResource("/opencv/libopencv_java460.so").toURI().getRawPath(); } catch (Exception e) { e.printStackTrace(); } System.load(opencvDllName); // xml路径不能包含中文 String faceXml = null; try { faceXml = FaceDetectorController.class.getResource("/opencv/lbpcascade_frontalface.xml").toURI().getRawPath(); } catch (Exception e) { e.printStackTrace(); } faceDetector = new CascadeClassifier(faceXml); } } // 灰度化人脸 private Mat conv_Mat(String img) { Mat image0 = Imgcodecs.imread(img); Mat image1 = new Mat(); // 灰度化 Imgproc.cvtColor(image0, image1, Imgproc.COLOR_BGR2GRAY); // 探测人脸 MatOfRect faceDetections = new MatOfRect(); faceDetector.detectMultiScale(image1, faceDetections); // rect中人脸图片的范围 for (Rect rect : faceDetections.toArray()) { Mat face = new Mat(image1, rect); return face; } return null; } private double compare_image(String img_1, String img_2) { Mat mat_1 = conv_Mat(img_1); Mat mat_2 = conv_Mat(img_2); Mat hist_1 = new Mat(); Mat hist_2 = new Mat(); //颜色范围 MatOfFloat ranges = new MatOfFloat(0f, 256f); //直方图大小, 越大匹配越精确 (越慢) MatOfInt histSize = new MatOfInt(10000000); Imgproc.calcHist(Arrays.asList(mat_1), new MatOfInt(0), new Mat(), hist_1, histSize, ranges); Imgproc.calcHist(Arrays.asList(mat_2), new MatOfInt(0), new Mat(), hist_2, histSize, ranges); // CORREL 相关系数 return Imgproc.compareHist(hist_1, hist_2, Imgproc.CV_COMP_CORREL); } @RequestMapping("faceDetector") @ResponseBody public String faceDetector() { try { //图片路径不能包含中文 double compareHist = compare_image("1.jpg", "2.jpg"); if (compareHist > 0.6) { return "人脸对比成功"; } else { return "人脸不匹配"; } } catch (Exception e) { return "人脸对比失败"; } } }
在war模块的resources下新建opencv文件夹并把so文件和xml文件放进去
plugin修改
maven-war-plugin 3.2.0 ../face-service/src/main/resources/lib/ WEB-INF/lib ***.* ${project.basedir}/target/classes true opencv/*.* src/main/${package.environment} opencv/*.* ${project.basedir}/target/classes false



