""" ************************************************************************* // Example program using OpenCV library // python >3.7 - OpenCV 4.5 // @file e7c.py // @author Luis M. Jimenez // @date 2022 // // @brief Course: Computer Vision (1782) // Dept. of Systems Engineering and Automation // Automation, Robotics and Computer Vision Lab (ARVC) // http://arvc.umh.es // University Miguel Hernandez // // @note Description: // - Shows the use of feature points detectors/descriptors/matchers // // - Capture images from two cameras (decrease resolution) // - Detects featured points (ORB/SIFT/SURF) using common interface FeatureDetector // - Calculates Descriptor vector (ORB/SIFT/SURF) using common interface DescriptorExtractor // - Matches featured points between frames using Brute Force or FLANN algorithms // - shows the use of knnMatch, finds the 2 best matches for each keypoint (k=2) // - Select those matches with enough distance difference between these 2 matches (radio test) // - Draw Matches and detected points over captured image // ************************************************************************* """ # Import libraries import cv2 as cv import numpy as np import argparse # ----------------------------------------- # Global variables # ----------------------------------------- WINDOW_CAMERA1 = '(W1) Cameras (Feature Detector)' # window id CAMERA_ID = [0, 1] # camera ids FEATURE_TYPE = 'orb' # Default Feature Detector sift/surf/orb MATCHER_TYPE = 'flann' # Default Feature Matcher bf/flann # check command line parameters (camera id, feature, descriptor) parser = argparse.ArgumentParser(description='OpenCV example: Feature Points Detector') parser.add_argument('-c', dest='cameraID', type=int, default=CAMERA_ID[0], metavar='id', help='camera id') parser.add_argument('-f', dest='feature', type=str, default=FEATURE_TYPE, metavar='feature', help='sift | surf | orb') parser.add_argument('-m', dest='matcher', type=str, default=MATCHER_TYPE, metavar='matcher', help='bf (BruteForce) | flann') CAMERA_ID[0] = parser.parse_args().cameraID CAMERA_ID[1] = parser.parse_args().cameraID + 1 FEATURE_TYPE = parser.parse_args().feature.lower() # feature (lowercase) MATCHER_TYPE = parser.parse_args().matcher.lower() # matcher (lowercase) # ----------------------------------------- # Put here the code to Initialize objets # ----------------------------------------- # Open camera objects cameras = [] for id in CAMERA_ID: camera = cv.VideoCapture(id) if not camera.isOpened(): print("you need to connect a camera, sorry.") exit() cameras.append(camera) # lower resolution to speed feature extraction camera.set(cv.CAP_PROP_FRAME_WIDTH, 640) camera.set(cv.CAP_PROP_FRAME_HEIGHT, 480) # Getting camera resolution cameraWidth = int(camera.get(cv.CAP_PROP_FRAME_WIDTH)) cameraHeight = int(camera.get(cv.CAP_PROP_FRAME_HEIGHT)) # Creating visualization windows cv.namedWindow(WINDOW_CAMERA1, cv.WINDOW_AUTOSIZE) # Init FeaturePoint Detector if FEATURE_TYPE == 'sift': detector = cv.SIFT.create(nfeatures=100) elif FEATURE_TYPE == 'orb': detector = cv.ORB.create(nfeatures=100) elif FEATURE_TYPE == 'surf': detector = cv.xfeatures2d.SURF.create(hessianThreshold=400) else: detector = cv.SIFT.create(nfeatures=100) # Default detector # Detector information print(f"{FEATURE_TYPE=}") print(f"{detector.descriptorSize()=}") # Init Feature Matcher if MATCHER_TYPE == 'flann': if FEATURE_TYPE == 'orb': # Binary descriptor -> NORM_HAMMING index_params = dict(algorithm = 6, table_number = 6, key_size = 12, multi_probe_level = 3) matcher = cv.FlannBasedMatcher(index_params) else: matcher = cv.FlannBasedMatcher.create() else: # Brute Force Matcher if FEATURE_TYPE == 'orb': # Binary descriptor -> NORM_HAMMING matcher = cv.BFMatcher.create(normType=cv.NORM_HAMMING, crossCheck=True) else: # Scalar descriptor -> NORM_L2 matcher = cv.BFMatcher.create(normType=cv.NORM_L2, crossCheck=True) print(f"{MATCHER_TYPE=}") print(f"Capturing images from camera {CAMERA_ID} ({cameraWidth},{cameraHeight})") print("...Hit q/Q/Esc to exit.") # ----------------------------------------- # Main Loop # while there are images ... # ----------------------------------------- while True: # Capture frame-by-frame captures = [] for camera in cameras: ret, capture = camera.read() if ret: captures.append(capture) # if frame is read correctly ret is True if not ret: print("Can't receive frame (stream end?). Exiting ...") break # ----------------------------------------- # Put your image processing code here # ----------------------------------------- gray_images = [] for capture in captures: gray_image = cv.cvtColor(capture, cv.COLOR_BGR2GRAY) gray_images.append(gray_image) keypointsList = [] descriptorsList = [] for gray_image in gray_images: keypoints = detector.detect(gray_image) # Tuple of keypoints keypoints, descriptors = detector.compute(gray_image, keypoints) # Descriptor matrix (ndarray) keypointsList.append(keypoints) descriptorsList.append(descriptors) # knnMatch descriptors. matches = matcher.knnMatch(descriptorsList[0], descriptorsList[1], k=2) # list/tuple of k-tuple of matches # Apply ratio test to the two best matches goodMatches = [] for m1, m2 in matches: # tuple with two elements k==2 if m1.distance/m2.distance < 0.75: goodMatches.append(m1) # Sort matches in the order of their distance. matches = sorted(goodMatches, key=lambda x: x.distance) # ----------------------------------------- # Put your visualization code here # ----------------------------------------- # Draw first 20 matches. dispimage = cv.drawMatches(captures[0], keypointsList[0], captures[1], keypointsList[1], matches[:min(20,len(matches))], outImg=None, flags=cv.DRAW_MATCHES_FLAGS_DEFAULT) # Show #KeyPoints cv.putText(dispimage, f"Detected: {len(keypoints)} {FEATURE_TYPE.upper()} points - Matcher: {MATCHER_TYPE.upper()}", org=(5, 15), fontFace=cv.FONT_HERSHEY_DUPLEX, fontScale=0.4, color=(0, 0, 255), thickness=1, lineType=cv.LINE_AA) cv.imshow(WINDOW_CAMERA1, dispimage) # Display the resulting frame # check keystroke to exit (image window must be on focus) key = cv.pollKey() if key == ord('q') or key == ord('Q') or key == 27: break elif key == ord('s') or key == ord('S'): key = cv.waitKey() # Stop and wait for a keystroke # End while (main loop) # ----------------------------------------- # free windows and camera resources # ----------------------------------------- cv.destroyAllWindows() for camera in cameras: if camera.isOpened(): camera.release()