Object Tracking

📕 Optional Flow

✍ 루카스 카나데

import numpy as np
import cv2

corner_track_params = dict(maxCorners = 10, qualityLevel = 0.3, minDistance=7, blockSize=7)

lk_params = dict(winSize = (200,200), maxLevel = 2, criteria = (cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 0.03))

cap = cv2.VideoCapture(0)

ret, prev_frame = cap.read()

prev_gray = cv2.cvtColor(prev_frame, cv2.COLOR_BGR2GRAY)

# 추적하고 싶은 점
prevPts = cv2.goodFeaturesToTrack(prev_gray, mask=None, **corner_track_params)

# mask는 실제 점을 표시하고 선을 그릴 예정 -> 점을 추적한다기 보다 선을 그리는 역할
mask = np.zeros_like(prev_frame)

while True:
    ret, frame = cap.read()
    
    frame_gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)

    nextPts, status, err = cv2.calcOpticalFlowPyrLK(prev_gray, frame_gray, prevPts, None, **lk_params) # 이 함수가 루카스 카나데 함수(LK)
    
    good_new = nextPts[status==1]
    good_prev = prevPts[status==1]
    
    for i, (new, prev) in enumerate(zip(good_new, good_prev)):
        x_new, y_new = new.ravel()
        x_prev, y_prev = prev.ravel()
        
        mask = cv2.line(mask, (x_new, y_new), (x_prev, y_prev), (0,255,0), 3)
        
        frame = cv2.circle(frame, (x_new, y_new), 8, (0,0,255), -1)
        
    img = cv2.add(frame, mask)
    cv2.imshow('tracking', img)
    
    k = cv2.waitKey(30) & 0xFF
    if k == 27:
        break
        
    prev_gray = frame_gray.copy()
    prevPts = good_new.reshape(-1,1,2)
    
cv2.destroyAllWindows()
cap.release()

---

✍ 군나르 파너백

import numpy as np
import cv2

cap = cv2.VideoCapture(0)

ret, frame1 = cap.read()

prvsImg = cv2.cvtColor(frame1, cv2.COLOR_BGR2GRAY)

hsv_mask = np.zeros_like(frame1)
hsv_mask[:,:, 1] = 255 # 채도를 최대치인 255로 설정

while True:
    ret, frame2 = cap.read()
    
    nextImg = cv2.cvtColor(frame2, cv2.COLOR_BGR2GRAY)
    
    flow = cv2.calcOpticalFlowFarneback(prvsImg, nextImg, None, 0.5, 3, 15, 3, 5, 1.2, 0) # 이게 디폴트 값
    
    mag, ang = cv2.cartToPolar(flow[:,:,0], flow[:,:,1], angleInDegrees=True)
    
    hsv_mask[:,:,0] = ang/2 # 360 ~ 180도 사이로 고정
    
    hsv_mask[:,:,2] = cv2.normalize(mag,None,0,255,cv2.NORM_MINMAX) # 0부터 255까지 정규화
    
    bgr = cv2.cvtColor(hsv_mask, cv2.COLOR_HSV2BGR)
    cv2.imshow('frame', bgr)
    
    k = cv2.waitKey(10) & 0xFF
    
    if k == 27:
        break
        
    prvsImg = nextImg
    
cap.release()
cv2.destroyAllWindows()

---

📗 MeanShift & CamShift

import cv2
import numpy as np

cap = cv2.VideoCapture(0)

ret, frame = cap.read()

# 초기 추적 윈도우 설정
face_cascade = cv2.CascadeClassifier('../Computer-Vision-with-Python/DATA/haarcascades/haarcascade_frontalface_default.xml')
face_rects = face_cascade.detectMultiScale(frame)


# 얼굴 하나만 추적
(face_x, face_y, w, h) = tuple(face_rects[0])
track_window = (face_x, face_y, w, h)

roi = frame[face_y:face_y+h, face_x:face_x+w] # 얼굴부분 선택

hsv_roi = cv2.cvtColor(roi, cv2.COLOR_BGR2HSV)

roi_hist = cv2.calcHist([hsv_roi], [0], None, [180], [0,180])

cv2.normalize(roi_hist, roi_hist, 0, 255, cv2.NORM_MINMAX)

term_crit = (cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 1)

while True:
    ret, frame = cap.read()
    
    if ret == True:
        hsv = cv2.cvtColor(frame, cv2.COLOR_BGR2HSV)
        
        dst = cv2.calcBackProject([hsv], [0], roi_hist, [0, 180], 1)
        
        ############# MeanShift ##############
        # MeanShift는 Box의 크기는 안변함
        '''
        ret, track_window = cv2.meanShift(dst, track_window, term_crit)
        
        x, y, w, h = track_window
        img2 = cv2.rectangle(frame, (x, y), (x+w, y+h), (0,0,255), 5)
        '''
        ######################################
        
        ############# CamShift ###############
        # CamShift는 크기도 변함
        ret, track_window = cv2.CamShift(dst, track_window, term_crit)
        
        pts = cv2.boxPoints(ret)
        pts = np.int0(pts)
        img2 = cv2.polylines(frame, [pts], True, (0,0,255), 5)
        
        ######################################
        
        cv2.imshow('img', img2)
        
        k = cv2.waitKey(1) & 0xFF
        
        if k == 27:
            break
            
    else:
        break
        
cv2.destroyAllWindows()
cap.release()

📘 Tracking APIs

import cv2

def ask_for_tracker():
    print("Welcome! What Tracker API would you like to use?")
    print("Enter 0 for BOOSTRING :")
    print("Enter 1 for MIL :")    
    print("Enter 2 for KCF :")    
    print("Enter 3 for TLD :")    
    print("Enter 4 for MEDIANFLOW :")
    choice = input("Please select your tracker : ")
    
    if choice == '0':
        tracker = cv2.TrackerBoosting_create()
    if choice == '1':
        tracker = cv2.TrackerMIL_create()
    if choice == '2':
        tracker = cv2.TrackerKCF_create()
    if choice == '3':
        tracker = cv2.TrackerTLD_create()
    if choice == '4':
        tracker = cv2.TrackerMedianFlow_create()
        
    
    return tracker
    
tracker = ask_for_tracker()
# <TrackerMedianFlow 000001A0T01523DI9> 뭐 이런 형식에서 이름만 가져오겠다는 의미
tracker_name = str(tracker).split()[0][1:]

cap = cv2.VideoCapture(0)

ret, frame = cap.read()

# 이 함수를 사용하면 바운딩 박스를 직접 그릴수 있어 관심영역을 지정할 수 있다.
roi = cv2.selectROI(frame, False)

# 트래커를 초기화시킨다(init)
ret = tracker.init(frame, roi)

while True:
    ret, frame = cap.read()
    
    # 업데이트를 해준다. 
    success, roi = tracker.update(frame)
    
    # roi는 직사각형 꼭짓점을 가리키는 4개의 부동 "소수점"을 나타냄.
    (x, y, w, h) = tuple(map(int, roi))
    
    if success:
        p1 = (x, y)
        p2 = (x+w, y+h)
        cv2.rectangle(frame, p1, p2, (0,255,0), 3)
    else:
        cv2.putText(frame, 'Failure to Detect Tracking!!', (100, 200), cv2.FONT_HERSHEY_SIMPLEX, 1, (0,0,255), 3)
        
    cv2.putText(frame, tracker_name, (20, 400), cv2.FONT_HERSHEY_SIMPLEX, 1, (0,255,0), 3)
    
    cv2.imshow(tracker_name, frame)
    
    k = cv2.waitKey(1) & 0xFF
    
    if k == 27:
        break
        
cap.release()
cv2.destroyAllWindows()