Merge branch 'neuer_ansatz_bildverarbeitung' of yasarba71520/Line-Following-Robot into master

AutonomousMode/Input/input.cpp

@@ -15,14 +15,14 @@ Mat Input::readFile(String filePath)
 {
     std::srand(std::time(0));
     // Read all .jpg files from the specified folder
-    std::string folder = filePath;
-    std::vector<std::string> filenames;
+    cv::String folder = filePath;
+    std::vector<cv::String> filenames;
     cv::glob(folder, filenames);
 
     // Random shuffle
     std::random_shuffle(filenames.begin(), filenames.end());
 
-    Mat image = imread(filenames[0], IMREAD_COLOR);
+    Mat image = imread(filePath, IMREAD_COLOR);
     
     if(image.empty())
     {
@@ -39,16 +39,19 @@ Mat Input::readWebcam()
     Mat image;
 
     if(!cap.isOpened()) {
-        cout << "Fehler";
+        cout << "Video capture not opened" << std::endl;
         return Mat();
     }
 
-    cap.read(image);
-    
+    if(!cap.grab()) {
+        cout << "Could not grab frame from camera" << std::endl;
+        return Mat();
+    }   
+    cap.retrieve(image);
     return image;  
 }
 
 void Input::freeWebcam()
 {
     this->cap.release();
-}
+}

AutonomousMode/Processing/processing.cpp

@@ -24,19 +24,37 @@ void Processing::processImage(Mat& inputPicture, int thresholdBinary, int gaussK
     // And one (the other one) to segment the lines.
     // No return value here as the input is passed by reference -> directly modified.
     cvtColor(inputPicture, inputPicture, COLOR_BGR2GRAY);
-    threshold(inputPicture, inputPicture, thresholdBinary, 255, THRESH_BINARY);
     GaussianBlur(inputPicture, inputPicture, Size(gaussKernelSize, gaussKernelSize), 0); 
-    Canny(inputPicture, inputPicture, thresholdCanny1, thresholdCanny2, apertureSizeCanny);
+    threshold(inputPicture, inputPicture, thresholdBinary, 255, THRESH_BINARY);
+    
+    //Perform a opening
+    Mat kernel(5,5, CV_8UC1,1);
+    morphologyEx(inputPicture, inputPicture, 2, kernel);
 }
 
-std::vector<Vec4i> Processing::calculateLineSegments(const Mat& inputPicture)
+FrameData Processing::calculateLineSegments(const Mat& inputPicture, const cv::Rect& roi)
 {
-    //See following link
-    //https://stackoverflow.com/questions/45322630/how-to-detect-lines-in-opencv
-    vector<Vec4i> lines;
-    VectorOfLines linesInVectors;
-    HoughLinesP(inputPicture, lines, 1, CV_PI/360, 150, 0, 250);
-    //lines = linesInVectors.findMiddleLine(lines);
-    
-    return lines;
+    FrameData data;
+    cv::findContours(inputPicture, data.contours, RETR_LIST, CHAIN_APPROX_SIMPLE);
+
+    //Delete the areas that are too small
+    auto iterator = data.contours.begin();
+    while(iterator != data.contours.end())
+    {
+        if (contourArea(*iterator) < 3500)
+        {
+            iterator = data.contours.erase(iterator);
+        }
+        else
+        {
+            Rect boundingBox = boundingRect(*iterator);
+            boundingBox.x += roi.x;
+            boundingBox.y += roi.y;
+            data.boundingBoxes.push_back(boundingBox);
+            data.middlePoints.push_back(Point(boundingBox.x+boundingBox.width/2, boundingBox.y+boundingBox.height/2));
+            data.leftEdges.push_back(Point(boundingBox.x, boundingBox.y+boundingBox.height/2));
+            ++iterator;
+        }
+    }
+    return data;
 }

AutonomousMode/Processing/processing.h

@@ -21,5 +21,5 @@ public:
 
     void processImage(Mat& inputPicture, int thresholdBinary, int gaussKernelSize, int thresholdCanny1, int thresholdCanny2, int apertureSizeCanny);
 
-    std::vector<Vec4i> calculateLineSegments(const Mat& inputPicture);
+    FrameData calculateLineSegments(const Mat& inputPicture, const cv::Rect& roi);
 };

AutonomousMode/Spielwiese/spielwiese.cpp

@@ -1,7 +1,8 @@
-#include <opencv2/core/utils/logger.hpp>
+//#include <opencv2/core/utils/logger.hpp>
 #include <opencv2/opencv.hpp>
 
 #include <iostream>
+#include <direct.h>
 
 #include <input.h>
 #include <processing.h>
@@ -34,27 +35,97 @@ void sweep_em_all(int thresholdBinary, int videoHeight, int videoWidth, int gaus
 
 void in_depth_processing_chain(int thresholdBinary, int videoHeight, int videoWidth, int gaussKernelSize, int thresholdCanny1, int thresholdCanny2, int apertureSizeCanny)
 {
-    std::string outputFolder = "C:\\Users\\User\\Desktop\\temp";
     Input input(videoHeight, videoWidth);
 
-    Mat image = input.readFile("C:\\Users\\User\\Desktop\\Studium\\02_Master_MSY\\2. Semester  Winter 22 23\\Projekt\\Line-Following-Robot\\Test_data");
-    imwrite(outputFolder + "\\01_input.jpg", image);
-    cvtColor(image, image, COLOR_BGR2GRAY);
-    imwrite(outputFolder + "\\02_color_convert.jpg", image);
-    GaussianBlur(image, image, Size(gaussKernelSize, gaussKernelSize), 0); 
-    imwrite(outputFolder + "\\03_gauss.jpg", image);
-    threshold(image, image, thresholdBinary, 255, THRESH_BINARY);
-    imwrite(outputFolder + "\\04_threshold.jpg", image);
-    Canny(image, image, thresholdCanny1, thresholdCanny2, apertureSizeCanny);
-    imwrite(outputFolder + "\\05_canny.jpg", image);
+    cv::String outputFolder = "C:\\Users\\tim-z\\Desktop\\temp";
+    cv::String inputFolder = "C:\\Users\\tim-z\\Desktop\\Studium\\02_Master MSY\\2. Semester Winter 2022 2023\\Projekt\\Repo\\Line-Following-Robot\\AutonomousMode\\Test_data";
 
+    std::vector<cv::String> filenames;
+    cv::glob(inputFolder, filenames);
+
+    //filenames.begin(), filenames.end()
+    int i = 0;
+    for(std::vector<cv::String>::iterator it = filenames.begin(); it != filenames.end(); it++)
+    {
+        std::string current_output = outputFolder + "\\" + to_string(i);
+        std::cout << current_output << std::endl;
+        const char* current_output_char = current_output.c_str();
+        _mkdir(current_output_char);
+
+        std::string inputFile = inputFolder + "\\image" + to_string(i+1) + ".jpeg";
+        Mat original_image = input.readFile(inputFile);
+        imwrite(current_output + "\\00_input.jpg", original_image);
+
+        Point roiOrigin(0, original_image.rows*(7.5/12.0));
+        Rect roi = Rect(roiOrigin.x, roiOrigin.y, original_image.cols, original_image.rows/12);
+
+        Mat image = original_image(roi);
+
+        imwrite(current_output + "\\01_roi.jpg", image);
+        cvtColor(image, image, COLOR_BGR2GRAY);
+        imwrite(current_output + "\\02_color_convert.jpg", image);
+        GaussianBlur(image, image, Size(gaussKernelSize, gaussKernelSize), 0); 
+        imwrite(current_output + "\\03_gauss.jpg", image);
+        threshold(image, image, thresholdBinary, 255, THRESH_BINARY);
+        imwrite(current_output + "\\04_threshold.jpg", image);
+
+        // Opening (reduces noise)
+        Mat kernel(5,5, CV_8UC1,1);
+        morphologyEx(image, image, 2, kernel);
+        imwrite(current_output + "\\05_opening.jpg", image);
+
+        //Canny(image, image, thresholdCanny1, thresholdCanny2, apertureSizeCanny);
+        //imwrite(outputFolder + "\\06_canny.jpg", image);
+
+        vector<vector<Point> > contours;
+        vector<Vec4i> hierarchy;
+        vector<Rect> vectorOfRects;
+        vector<Point> vectorOfLeftEdges;
+
+        findContours(image,contours, hierarchy, RETR_LIST, CHAIN_APPROX_SIMPLE);
+        int amountOfValidRects = 0;
+        for( int i = 0; i< contours.size(); i++ ) // iterate through each contour. 
+        {
+                double a = contourArea( contours[i],false);  //  Find the area of contour
+                if(a > 3500)
+                {
+                    drawContours(original_image, contours, i, Scalar(0,255,255), 1, 8, hierarchy, 0, roiOrigin);
+                    
+                    Rect currentBoundingRect = boundingRect(contours[i]);
+                    //Handle roi offset:
+                    currentBoundingRect.x += roiOrigin.x;
+                    currentBoundingRect.y += roiOrigin.y;
+                    vectorOfRects.push_back(currentBoundingRect);
+
+                    rectangle(original_image, currentBoundingRect, Scalar(0,255,0));
+                 
+                    // get mid-point of rect
+                    Point midRect = Point(currentBoundingRect.x+currentBoundingRect.width/2, currentBoundingRect.y+currentBoundingRect.height/2);
+
+                    // Draw middle as small rect instead of circle because for whatever reasons drawing a circle doesnt work.
+                    Rect testRect(Point(midRect.x-2, midRect.y-2), Point(midRect.x+2, midRect.y+2));
+                    rectangle(original_image, testRect, Scalar(0,0,255));
+                    // get the left edge of rect
+                    // used as offset as raspicam is not
+                    // mounted on mid of regbot
+                    Point leftEdge(currentBoundingRect.x, currentBoundingRect.y+currentBoundingRect.height/2);
+                    vectorOfLeftEdges.push_back(leftEdge);
+
+                    testRect = Rect(Point(leftEdge.x-2, leftEdge.y-2), Point(leftEdge.x+2, leftEdge.y+2));
+                    rectangle(original_image, testRect, Scalar(0,0,255));
+                    amountOfValidRects++;
+                }
+        }
+        imwrite(current_output + "\\06_contours.jpg", original_image);
+        i++;
+    }
 }
 
 
 int main(void)
 {
     //Disable opencv logging messages
-    cv::utils::logging::setLogLevel(cv::utils::logging::LOG_LEVEL_WARNING);
+    //cv::utils::logging::setLogLevel(cv::utils::logging::LOG_LEVEL_WARNING);
 
     const int thresholdBinary = 140;
     const int videoHeight = 720;

AutonomousMode/Utils/utils.h

@@ -60,4 +60,15 @@ class VectorOfLines{
     static double calcDistance(Point p0, Point p1);
     vector<Vec4i> findMiddleLine(vector<Vec4i> &lines);
 
+};
+
+class FrameData
+{
+public:
+    std::vector<std::vector<cv::Point>> contours;
+    std::vector<cv::Rect> boundingBoxes;
+    std::vector<cv::Point> leftEdges;
+    std::vector<cv::Point> middlePoints;
+
+    FrameData(): contours(), boundingBoxes(), leftEdges(), middlePoints() {}
 };

AutonomousMode/autonomous_mode_main.cpp

@@ -7,17 +7,19 @@ int main(void)
     cv::utils::logging::setLogLevel(cv::utils::logging::LOG_LEVEL_WARNING);
 
     const int thresholdBinary = 140;
-    const int videoHeight = 240;
-    const int videoWidth = 320;
+    const int videoHeight = 720;
+    const int videoWidth = 1280;
     const int gaussKernelSize = 21;
     const int thresholdCanny1 = 50;
     const int thresholdCanny2 = 100;
     const int apertureSizeCanny = 3;
 
     LFR lfr(videoHeight, videoWidth, thresholdBinary, gaussKernelSize, thresholdCanny1, thresholdCanny2, apertureSizeCanny);
+    lfr.saveOutputFlag = true;
+    lfr.outputFileName = "/home/pi/Line-Following-Robot/AutonomousMode/tmp/test.jpg";
     lfr.startLoop();
     //To end the video stream, write any char in the console.
     char a;
     std::cin >> a; 
     lfr.endLoop();
-}
+}

AutonomousMode/lfr.cpp

@@ -2,7 +2,7 @@
 
 
 LFR::LFR(int videoHeight, int videoWidth, int thresholdBinary, int gaussKernelSize, int thresholdCanny1, int thresholdCanny2, int apertureSizeCanny)
-    : iAmLooping(false), input(videoHeight, videoWidth), processing(), controlModule(), interpreter(), intersectionHandler()
+    : iAmLooping(false), input(videoHeight, videoWidth), processing(), controlModule(), interpreter(), intersectionHandler(), roi()
 {
     this->iAmLooping = false;
     this->thresholdBinary = thresholdBinary;
@@ -10,6 +10,14 @@ LFR::LFR(int videoHeight, int videoWidth, int thresholdBinary, int gaussKernelSi
     this->thresholdCanny1 = thresholdCanny1;
     this->thresholdCanny2 = thresholdCanny2;
     this->apertureSizeCanny = apertureSizeCanny;
+
+    this->videoFlag = false;
+    this->saveOutputFlag = false;
+    this->outputFileName = "";
+
+    cv::Point roiOrigin(0, videoHeight*(7.5/12.0));
+    roi = Rect(roiOrigin.x, roiOrigin.y, videoWidth, videoHeight/12);
+
 }
 
 LFR::~LFR()
@@ -22,21 +30,16 @@ LFR::~LFR()
 
 void LFR::loop()
 {
-    namedWindow("Display window");
+    if(this->videoFlag) {namedWindow("Display window");}
     while(iAmLooping)
     {
-        Mat image = input.readWebcam();
-        processing.processImage(image, this->thresholdBinary, this->gaussKernelSize, this->thresholdCanny1, thresholdCanny2, this->apertureSizeCanny);
-        std::vector<Vec4i> lines = processing.calculateLineSegments(image);
-        for( size_t i = 0; i < lines.size(); i++ )
-        {
-            line( image, Point(lines[i][0], lines[i][1]),
-            Point( lines[i][2], lines[i][3]), (0,0,255), 1, 8 );
-        }  
-        imshow("Display window", image);
-        char c = (char)waitKey(1);
+        Mat originalImage = input.readWebcam();
+        Mat processedImage = originalImage;
+        processing.processImage(processedImage, this->thresholdBinary, this->gaussKernelSize, this->thresholdCanny1, thresholdCanny2, this->apertureSizeCanny);
+        FrameData data = processing.calculateLineSegments(processedImage, this->roi);
+        this->provideOutput(processedImage);
     }
-    destroyWindow("Display window");
+    if(this->videoFlag) {destroyWindow("Display window");}
     input.freeWebcam();
 }
 
@@ -51,4 +54,17 @@ void LFR::endLoop()
     iAmLooping = false;
     this->loopThread.join();
     return;
-}
+}
+
+void LFR::provideOutput(const Mat& image)
+{
+    if(this->videoFlag)
+        {
+            imshow("Display window", image);
+            char c = (char)waitKey(1);
+        }
+        if (this->saveOutputFlag && !(this->outputFileName.empty()))
+        {
+            imwrite(this->outputFileName, image);        
+        }
+}

AutonomousMode/lfr.h

@@ -30,6 +30,9 @@ class LFR
     int thresholdCanny1;
     int thresholdCanny2;
     int apertureSizeCanny;
+    cv::Rect roi;
+
+    void provideOutput(const Mat& image);
 
 public:
 
@@ -39,5 +42,11 @@ public:
 
     void startLoop();
     void endLoop();
+
+    
+    bool videoFlag;
+    bool saveOutputFlag;
+
+    std::string outputFileName;
     
 };