* change readme.

* add more data to svm.
* change the svm train process, fix train data size, add new test data.
* close the svm train menu.

Author: liuruoze

README.md

@@ -62,6 +62,8 @@ EasyPR是一个开源的中文车牌识别系统，其目标是成为一个简
 
 关于本次改动的具体内容可以看博客中的[介绍](http://www.cnblogs.com/subconscious/p/5637735.html)。
 
+注意，目前1.4和1.5版的SVM训练好的文件在使用时会有问题，这个原因可能跟opencv3的实现改变有关。建议要训练SVM的话使用基于opencv2的1.3版。
+
 ### 跨平台
 
 目前除了windows平台以外，还有以下其他平台的EasyPR版本。一些平台的版本可能会暂时落后于主平台。
@@ -160,6 +162,18 @@ CPlate类包含了车牌的各种信息，其中重要的如下：
 
 plateMat代表车牌图像，rrect代表车牌的可旋转矩形位置，license代表车牌字符串，例如“蓝牌：苏EUK722”。
 
+这里说下如何去阅读如下图的识别结果。
+
+![EasyPR DetectResults](resources/doc/res/one_image_detect.jpg)
+
+第1行代表的是图片的文件名。
+
+第2行代表GroundTruth车牌，用后缀（g）表示。第3行代表EasyPR检测车牌，用后缀（d）表示。两者形成一个配对，第4行代表两者的字符差距。
+
+下面同上。本图片中有3个车牌，所有共有三个配对。最后的Recall等指标代表的是整幅图片的定位评价，考虑了三个配对的结果。
+
+有时检测车牌的部分会用“无车牌”与“No string”替代。“无车牌”代表“定位不成功”，“No string”代表“定位成功但字符分割失败”。
+
 ### 版权
 
 EasyPR的源代码与训练数据遵循Apache v2.0协议开源。

etc/main_menu

@@ -2,7 +2,7 @@
 EasyPR Option:
 1. 测试;
 2. 批量测试;
-3. SVM训练;
+3. SVM训练(暂时关闭，请用1.3版的);
 4. ANN训练;
 5. GDTS生成;
 6. 开发团队;

resources/image/extreme_test/Germen/MIE6907.JPG

@@ -2008,3 +2008,48 @@ Recall:84.2263%, Precise:70.6573%, Fscore:76.8474%.
 Recall:84.2263%, Precise:70.6573%, Fscore:76.8474%.
 0-error:60.4%,  1-error:71.6%,  Chinese-precise:73.2%  
 ��ʱ��:114��,  ƽ��ִ��ʱ��:0.445313��
+2016-07-05 21:37:51
+��ͼƬ��:256,  Plates count:297,  δ��λ����:40,  ��λ��:86.532%  
+Recall:78.8339%, Precise:53.3124%, Fscore:63.6086%.
+0-error:55.3571%,  1-error:64.2857%,  Chinese-precise:67.8571%  
+��ʱ��:114��,  ƽ��ִ��ʱ��:0.445313��
+2016-07-06 07:38:55
+��ͼƬ��:256,  Plates count:297,  δ��λ����:29,  ��λ��:90.2357%  
+Recall:84.2263%, Precise:70.6573%, Fscore:76.8474%.
+0-error:60.4%,  1-error:71.6%,  Chinese-precise:73.2%  
+��ʱ��:114��,  ƽ��ִ��ʱ��:0.445313��
+2016-07-06 07:46:53
+��ͼƬ��:256,  Plates count:297,  δ��λ����:29,  ��λ��:90.2357%  
+Recall:84.2263%, Precise:70.6573%, Fscore:76.8474%.
+0-error:60.4%,  1-error:71.6%,  Chinese-precise:73.2%  
+��ʱ��:122��,  ƽ��ִ��ʱ��:0.476563��
+2016-07-06 07:55:13
+��ͼƬ��:3,  Plates count:0,  δ��λ����:0,  ��λ��:-1.#IND%  
+Recall:0%, Precise:0%, Fscore:0%.
+0-error:0%,  1-error:0%,  Chinese-precise:100%  
+��ʱ��:3��,  ƽ��ִ��ʱ��:1��
+2016-07-06 07:55:42
+��ͼƬ��:3,  Plates count:0,  δ��λ����:0,  ��λ��:-1.#IND%  
+Recall:0%, Precise:0%, Fscore:0%.
+0-error:0%,  1-error:0%,  Chinese-precise:100%  
+��ʱ��:6��,  ƽ��ִ��ʱ��:2��
+2016-07-06 07:57:02
+��ͼƬ��:2,  Plates count:0,  δ��λ����:0,  ��λ��:-1.#IND%  
+Recall:0%, Precise:0%, Fscore:0%.
+0-error:0%,  1-error:0%,  Chinese-precise:100%  
+��ʱ��:1��,  ƽ��ִ��ʱ��:0.5��
+2016-07-06 07:57:45
+��ͼƬ��:4,  Plates count:0,  δ��λ����:0,  ��λ��:-1.#IND%  
+Recall:0%, Precise:0%, Fscore:0%.
+0-error:0%,  1-error:0%,  Chinese-precise:100%  
+��ʱ��:1��,  ƽ��ִ��ʱ��:0.25��
+2016-07-06 08:00:06
+��ͼƬ��:256,  Plates count:297,  δ��λ����:29,  ��λ��:90.2357%  
+Recall:84.2263%, Precise:70.6573%, Fscore:76.8474%.
+0-error:60.4%,  1-error:71.6%,  Chinese-precise:73.2%  
+��ʱ��:118��,  ƽ��ִ��ʱ��:0.460938��
+2016-07-06 08:27:36
+��ͼƬ��:256,  Plates count:297,  δ��λ����:29,  ��λ��:90.2357%  
+Recall:84.2263%, Precise:70.6573%, Fscore:76.8474%.
+0-error:60.4%,  1-error:71.6%,  Chinese-precise:73.2%  
+��ʱ��:123��,  ƽ��ִ��ʱ��:0.480469��

resources/image/extreme_test/Germen/MIN2931.JPG

@@ -36,6 +36,11 @@ namespace easypr {
     extractFeature(inMat, features);
 
     float response = svm_->predict(features);
+    /*std::cout << "response:" << response << std::endl;
+
+    float score = svm_->predict(features, noArray(), cv::ml::StatModel::Flags::RAW_OUTPUT);
+    std::cout << "score:" << score << std::endl;*/
+
     result = (int)response;
 
     return 0;
@@ -64,6 +69,8 @@ namespace easypr {
 
     float score = svm_->predict(features, noArray(), cv::ml::StatModel::Flags::RAW_OUTPUT);
 
+    //std::cout << "score:" << score << std::endl;
+
     // score is the distance of margin，below zero is plate, up is not
     // when score is below zero, the samll the value, the more possibliy to be a plate.
     plate.setPlateScore(score);

resources/image/extreme_test/Germen/MIN3479.JPG

@@ -743,12 +743,6 @@ int CPlateLocate::plateColorLocate(Mat src, vector<CPlate> &candPlates,
   candPlates.insert(candPlates.end(), plates_blue.begin(), plates_blue.end());
   candPlates.insert(candPlates.end(), plates_yellow.begin(), plates_yellow.end());
 
-  //for (auto plate : plates_blue)
-  //  candPlates.push_back(plate);
-
-  //for (auto plate : plates_yellow)
-  //  candPlates.push_back(plate);
-
   return 0;
 }
 
@@ -1044,6 +1038,7 @@ int CPlateLocate::plateLocate(Mat src, vector<Mat> &resultVec, int index) {
 
   plateColorLocate(src, all_result_Plates, index);
   plateSobelLocate(src, all_result_Plates, index);
+  plateMserLocate(src, all_result_Plates, index);
 
   for (size_t i = 0; i < all_result_Plates.size(); i++) {
     CPlate plate = all_result_Plates[i];

resources/image/extreme_test/Germen/MIN6791.JPG

@@ -27,7 +27,7 @@ void SvmTrain::train() {
   svm_->setC(1);
   svm_->setNu(0.1);
   svm_->setP(0.1);
-  svm_->setTermCriteria(cvTermCriteria(CV_TERMCRIT_ITER, 100000, 0.00001));
+  svm_->setTermCriteria(cvTermCriteria(CV_TERMCRIT_ITER, 20000, 0.0001));
 
   auto train_data = tdata();
 
@@ -43,6 +43,7 @@ void SvmTrain::train() {
   fprintf(stdout, ">> Training done. Time elapse: %ldms\n", end - start);
   fprintf(stdout, ">> Saving model file...\n");
   svm_->save(svm_xml_);
+
   fprintf(stdout, ">> Your SVM Model was saved to %s\n", svm_xml_);
   fprintf(stdout, ">> Testing...\n");
   this->test();
@@ -72,11 +73,11 @@ void SvmTrain::test() {
       continue;
     }
     cv::Mat feature;
-    getHistogramFeatures(image, feature);
-
-    //std::cout << "predict: " << result << std::endl;
+    getLBPFeatures(image, feature);
 
     auto predict = int(svm_->predict(feature));
+    std::cout << "predict: " << predict << std::endl;
+
     auto real = item.label;
     if (predict == kForward && real == kForward) ptrue_rtrue++;
     if (predict == kForward && real == kInverse) ptrue_rfalse++;
@@ -123,47 +124,37 @@ void SvmTrain::prepare() {
 
   char buffer[260] = {0};
 
-  sprintf(buffer, "%s/has", plates_folder_);
-  auto has_file_list = utils::getFiles(buffer);
-  std::random_shuffle(has_file_list.begin(), has_file_list.end());
+  sprintf(buffer, "%s/has/train", plates_folder_);
+  auto has_file_train_list = utils::getFiles(buffer);
+  std::random_shuffle(has_file_train_list.begin(), has_file_train_list.end());
 
-  sprintf(buffer, "%s/no", plates_folder_);
-  auto no_file_list = utils::getFiles(buffer);
-  std::random_shuffle(no_file_list.begin(), no_file_list.end());
+  sprintf(buffer, "%s/has/test", plates_folder_);
+  auto has_file_test_list = utils::getFiles(buffer);
+  std::random_shuffle(has_file_test_list.begin(), has_file_test_list.end());
 
-  auto has_num = has_file_list.size();
-  auto no_num = no_file_list.size();
+  sprintf(buffer, "%s/no/train", plates_folder_);
+  auto no_file_train_list = utils::getFiles(buffer);
+  std::random_shuffle(no_file_train_list.begin(), no_file_train_list.end());
+
+  sprintf(buffer, "%s/no/test", plates_folder_);
+  auto no_file_test_list = utils::getFiles(buffer);
+  std::random_shuffle(no_file_test_list.begin(), no_file_test_list.end());
 
   fprintf(stdout, ">> Collecting train data...\n");
 
-  auto has_for_train = static_cast<int>(has_num * kSvmPercentage);
-  auto no_for_train = static_cast<int>(no_num * kSvmPercentage);
+  for (auto file : has_file_train_list)
+    train_file_list_.push_back({ file, kForward });
 
-  // copy kSvmPercentage of has_file_list to train_file_list_
-  train_file_list_.reserve(has_for_train + no_for_train);
-  for (auto i = 0; i < has_for_train; i++) {
-    train_file_list_.push_back({has_file_list[i], kForward});
-  }
-  // copy kSvmPercentage of no_file_list to the end of train_file_list_
-  for (auto i = 0; i < no_for_train; i++) {
-    train_file_list_.push_back({no_file_list[i], kInverse});
-  }
+  for (auto file : no_file_train_list)
+    train_file_list_.push_back({ file, kInverse });
 
   fprintf(stdout, ">> Collecting test data...\n");
 
-  auto has_for_test = has_num - has_for_train;
-  auto no_for_test = no_num - no_for_train;
-
-  // copy the rest of has_file_list to the test_file_list_
-  test_file_list_.reserve(has_for_test + no_for_test);
-  for (size_t i = has_for_train; i < has_num; i++) {
-    test_file_list_.push_back({has_file_list[i], kForward});
-  }
+  for (auto file : has_file_test_list)
+    test_file_list_.push_back({ file, kForward });
 
-  // copy the rest of no_file_list to the end of the test_file_list_
-  for (size_t i = no_for_train; i < no_num; i++) {
-    test_file_list_.push_back({no_file_list[i], kInverse});
-  }
+  for (auto file : no_file_test_list)
+    test_file_list_.push_back({ file, kInverse });
 }
 
 cv::Ptr<cv::ml::TrainData> SvmTrain::tdata() {
@@ -179,7 +170,7 @@ cv::Ptr<cv::ml::TrainData> SvmTrain::tdata() {
       continue;
     }
     cv::Mat feature;
-    getHistogramFeatures(image, feature);
+    getLBPFeatures(image, feature);
     feature = feature.reshape(1, 1);
 
     samples.push_back(feature);

resources/image/extreme_test/Germen/MIN8233.JPG

@@ -165,6 +165,7 @@ namespace easypr {
 
         // get the ground truth and compare it with the detect list;
         vector<CPlate> plateVecGT;
+        bool hasGroundTruth = true;
 #pragma omp critical
         {
           map<string, vector<CPlate>>::iterator it;         
@@ -173,6 +174,7 @@ namespace easypr {
             plateVecGT = it->second;
           }
           else {
+            hasGroundTruth = false;
             img_ss << "No ground truth found!" << endl;
           }
         }
@@ -268,6 +270,16 @@ namespace easypr {
           Rect_<float> plateRect_d;
           calcSafeRect(platePos_d, src, plateRect_d);
 
+          if (!hasGroundTruth) {
+            string detectPlateLicense = plate_d.getPlateStr();
+            vector<string> spilt_plate = Utils::splitString(detectPlateLicense, ':');
+
+            size_t size = spilt_plate.size();
+            if (size == 2 && spilt_plate.at(1) != "") {
+              img_ss << detectPlateLicense << " (d)" << endl;
+            }
+          }
+        
           // remain
           //XMLNode rectangleNode = rectangleNodes.addChild("taggedRectangle");
           //RotatedRect rr = platePos_d;