card_recognize.py

import copy

import cv2
import numpy as np

cap = cv2.VideoCapture(1)
width = 640
height = 480
cap.set(cv2.CAP_PROP_FRAME_WIDTH,width)
cap.set(cv2.CAP_PROP_FRAME_HEIGHT,height)

skip_pts = 1
fixed_width = int(width*0.73)
fixed_height = int(height*0.625)
coordinate = [[fixed_width, fixed_height],[fixed_width, 0],[] , [0, fixed_height],[0, 0]]
finish = False
while(cap.isOpened()):
    # get a frame
    ret, frame = cap.read()
    # create the rectangle ROI
    frame_copy = copy.deepcopy(frame)
    start_point = (int(width*0.156), int(height*0.209))
    cv2.rectangle(frame_copy, start_point, (start_point[0]+fixed_width, start_point[1]+fixed_height), (0, 255, 155), 5)
    roi = frame[start_point[1]:start_point[1]+fixed_height,start_point[0]:start_point[0]+fixed_width,:]
    ms = cv2.pyrMeanShiftFiltering(roi, 8, 20)
    canny = cv2.Canny(ms,100,180)
    # the point follow the Counter clockwise
    _, cnts, hierarchy = cv2.findContours(canny, cv2.RETR_TREE, cv2.CHAIN_APPROX_SIMPLE)
    cnts = sorted(cnts, key=cv2.contourArea, reverse=True)[:1]
    for j in range(len(cnts)):
        c = cnts[j]
        # approximate the contour
        peri = cv2.arcLength(c, True)
        approx = cv2.approxPolyDP(c, 0.01*peri, True)

        # if our approximated contour has four points, then
        # we can assume that we have found our screen
        # we assert the available area ratio bigger than 0.5
        if len(approx) == 4 and cv2.contourArea(approx.squeeze(1)) / (fixed_height*fixed_height) >= 0.5:
            compute_direction = [0,0,0,0]
            for i in range(4):
                delta = approx[(i+1)%4,0,:]-approx[i,0,:]
                delta_argmax = int(abs(delta[1]) > abs(delta[0]))
                # compute the line direction
                compute_direction[i] = ((delta_argmax+1)*(1 if delta[delta_argmax]>-delta[delta_argmax] else -1)+2)
            pts1 = approx.astype(np.float32)
            # compute the target warp coordinate
            pts2 = np.array([coordinate[x] for x in compute_direction],dtype=np.float32)
            M = cv2.getPerspectiveTransform(pts1, pts2)
            warp = cv2.warpPerspective(roi, M, (fixed_width, fixed_height))
            cv2.imshow("warp", warp)
            # finish = True

    cv2.imshow("capture", frame_copy)
    cv2.imshow("roi",roi)
    cv2.imshow("meanShift",ms)
    cv2.imshow("canny",canny)
    if cv2.waitKey(1) & 0xFF == ord('q'):
        break
    if finish:
        break
import matplotlib.pyplot as plt

cap.release()
cv2.destroyAllWindows()