优化按键位置检测算法

src/buildconfig.h

@@ -1,6 +1,7 @@
 //buildconfig.h 由 cmake 从 buildconfig.h.in 自动生成
 #pragma once
-#define LversionStr  L"0.1.3"
-#define versionStr "0.1.3"
+#define LversionStr  L"0.1.4"
+#define versionStr "0.1.4"
 constexpr int DEFAULT_KEY_THRESHOLD = 6;
 constexpr int DEFAULT_LOG_LEVEL = 4;
+constexpr int DEFAULT_HOUGH_CIRCLES_THRESHOLD = 60;

src/buildconfig.h.in

@@ -3,4 +3,5 @@
 #define LversionStr  L"${CMAKE_PROJECT_VERSION}"
 #define versionStr "${CMAKE_PROJECT_VERSION}"
 constexpr int DEFAULT_KEY_THRESHOLD = 6;
-constexpr int DEFAULT_LOG_LEVEL = 4;
+constexpr int DEFAULT_LOG_LEVEL = 4;
+constexpr int DEFAULT_HOUGH_CIRCLES_THRESHOLD = 60;

src/keyposfinder.cpp

@@ -1,11 +1,14 @@
 #include "keyposfinder.h"
+#include <opencv2/core/mat.hpp>
+#include <opencv2/imgproc.hpp>
 #include <sstream>
 #include <algorithm>
 #include <opencv2/opencv.hpp>
 #include "utils.h"
 
 
-void KeyPosFinder::begin(string filePath){
+void KeyPosFinder::begin(string filePath, int houghCirclesThreshold_){
+    houghCirclesThreshold = houghCirclesThreshold_;
     ui.addTask(L"检测按键位置");
     VideoCapture video(filePath);
     if(!video.isOpened()){
@@ -21,25 +24,30 @@ void KeyPosFinder::begin(string filePath){
     ui.logI(strBuf.str());
     strBuf.clear();
 
-    double scaleFactor = videoHeight/1080.0; //相比1080p的缩放系数
     Mat frame;
     Mat resizedFrame;
     if (logLevel == 5) namedWindow("frames");
     bool detectedGUIOpen = false;
+    //y缩放到480, x等比例
+    auto targetVideoHeight = 480;
+    auto targetVideoWidth = videoWidth * targetVideoHeight / videoHeight;
     // #pragma omp parallel
     while (minFramesCnt != vaildFramesCnt){
         //针对某些视频一开始根本没有打开界面的问题，在没有检测到界面时每隔5帧读取一帧，节省时间
         if(!video.read(frame)) break;
         if(!detectedGUIOpen) video.set(CAP_PROP_POS_FRAMES, video.get(CAP_PROP_POS_FRAMES) + 4); // 跳过4帧
-
-        resize(frame,resizedFrame,Size(),0.5,0.5);   //宽高缩小一半
-        resizedFrame = resizedFrame(Rect(Point(0,videoHeight/4),Point(videoWidth/2-1,videoHeight/2-1))); //切去上半部分, 保留下半部分
-        pyrMeanShiftFiltering(resizedFrame,resizedFrame,15*scaleFactor,100*scaleFactor);
+
+        resize(frame,resizedFrame,Size(targetVideoWidth,targetVideoHeight));
+        resizedFrame = resizedFrame(Rect(Point(0,targetVideoHeight/2),Point(targetVideoWidth-1,targetVideoHeight-1))); //切去上半部分, 保留下半部分
         cvtColor(resizedFrame,resizedFrame,COLOR_BGR2GRAY);
+        //equalizeHist(resizedFrame,resizedFrame);
+        auto blurredFrame = Mat();
+        GaussianBlur(resizedFrame, blurredFrame, Size(31,31), 2);
+        //pyrMeanShiftFiltering(resizedFrame,resizedFrame,15*scaleFactor,100*scaleFactor);
         //medianBlur(resizedFrame,resizedFrame,3);
         //Canny(resizedFrame,resizedFrame,80,255);
         vector<Vec3f> circles;
-        HoughCircles(resizedFrame,circles, HOUGH_GRADIENT,2,50*scaleFactor,200,55,0,60*scaleFactor);
+        HoughCircles(blurredFrame,circles, HOUGH_GRADIENT,2,50,200,houghCirclesThreshold,0,60);
         ui.updateMsg(L"找到的按键数量:" + to_wstring(circles.size()));
         if (!detectedGUIOpen && circles.size() > 10 ){
             detectedGUIOpen = true;
@@ -48,8 +56,8 @@ void KeyPosFinder::begin(string filePath){
 
         if (logLevel == 5) {
           for (auto c : circles)
-            circle(resizedFrame, Point(c[0], c[1]), c[2], Scalar(0, 255, 255), 3);
-          imshow("frames", resizedFrame);
+            circle(blurredFrame, Point(c[0], c[1]), c[2], Scalar(0, 255, 255), 3);
+          imshow("frames", blurredFrame);
           waitKey(1);   //必须加上,否则预览窗口灰屏卡住
         }
         //如果有21个圆，保存起来

src/keyposfinder.h

@@ -12,6 +12,7 @@ class KeyPosFinder {
 private:
   int vaildFramesCnt = 0;
   int minFramesCnt = 20;  
+  int houghCirclesThreshold;
   TermUI& ui;
   vector<vector<Vec3f>> vaildCircles; //3x7 = 21个按键
   array<Vec3f,21> result;
@@ -22,7 +23,7 @@ class KeyPosFinder {
 
 public:
 
-    void begin(string filePath);
+    void begin(string filePath, int houghCirclesThreshold_ = 60);
     array<Vec3f,21>&& getResult(){
       return move(result);
     }

src/main.cpp

@@ -38,11 +38,16 @@ int main(int argc, char* argv[]) {
   parser.add_argument("-o", "--output")
       .help("输出文件路径")
       .default_value(string("output.mid")); //存储const char*的std::any被any_cast到std::string时会抛异常
-
+  parser.add_argument("--key-threshold")
+      .help("设置按键位置检测的阈值, "
+            "如果输出的文件中有错音, 可以尝试增加这个值, 如果获取不到按键位置则减少")
+      .default_value(DEFAULT_HOUGH_CIRCLES_THRESHOLD)
+      .scan<'i', int>();
   //get a parameter which represent the threshold of the note finder
   parser.add_argument("-t", "--note-threshold")
       .help("设置音符检测的阈值, 如果输出的文件中有大量不存在的音符则增大这个值, 反之减少")
-      .default_value(DEFAULT_KEY_THRESHOLD);
+      .default_value(DEFAULT_KEY_THRESHOLD)
+      .scan<'i', int>();
   parser.add_argument("-c", "--notui")
       .help("不使用界面")
       .default_value(false)
@@ -97,11 +102,12 @@ ps:我不玩原神)123");
  // omp_set_num_threads(omp_get_num_procs());
 
   string filepath = parser.get<string>("input");
-  KeyPosFinder finder(ui); 
-  finder.begin(filepath);
+  KeyPosFinder finder(ui);
+  finder.begin(filepath, parser.get<int>("--key-threshold"));
   int startFrame = finder.getStartFrame();
   NoteDetector detector(ui);
-  detector.begin(filepath,finder.getResult(),parser.get<int>("--note-threshold"),startFrame);
+  detector.begin(filepath, finder.getResult(),
+                 parser.get<int>("--note-threshold"), startFrame);
   MidiWriter writer;
 
   //let the output file to be the input file except the extension changed to mid 

src/notedetector.cpp

@@ -71,8 +71,11 @@ void NoteDetector::begin(string filePath, array<Vec3f, 21> keyPos, int threshold
         frameQueue.pop();
         frameQueueMutex.unlock();
 
-        resize(frame,resizedFrame,Size(),0.5,0.5);   //宽高缩小一半,和获取位置时一样
-        resizedFrame = resizedFrame(Rect(Point(0,videoHeight/4),Point(videoWidth/2-1,videoHeight/2-1))); //切去上半部分, 保留下半部分
+        // y缩放到480, x等比例 ,和获取位置时一样
+        auto targetVideoHeight = 480;
+        auto targetVideoWidth = videoWidth * targetVideoHeight / videoHeight;
+        resize(frame,resizedFrame,Size(targetVideoWidth,targetVideoHeight));
+        resizedFrame = resizedFrame(Rect(Point(0,targetVideoHeight/2),Point(targetVideoWidth-1,targetVideoHeight-1))); //切去上半部分, 保留下半部分
         cvtColor(resizedFrame,resizedFrame,COLOR_BGR2GRAY); //只关心灰度
 
         //分别获取21个按钮区域的平均颜色