Computer Vision Day 2

97524 ワード

import cv2

ビデオキャプチャおよび表示(ビデオファイル)

cap = cv2.VideoCapture('./data/vtest.avi')


frame_size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)), int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))

# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 320)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 240)

print(frame_size)


while True:
    retval, frame = cap.read() # 프레임 캡처
    if not retval:
        break
        
    cv2.imshow("frame1", (frame))
    cv2.imshow("frame2", (255-frame))
    
    key = cv2.waitKey(25)
    if key == 27: #ESC
        break

if cap.isOpened():
    cap.release()

cv2.destroyAllWindows()

(768, 576)

ビデオキャプチャと表示(外部URL:droidcamから撮影したビデオ)

cap = cv2.VideoCapture('http://192.168.22.60:4747/mjpegfeed')


frame_size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)), int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))

# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 320)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 240)

print(frame_size)


while True:
    retval, frame = cap.read() # 프레임 캡처
    if not retval:
        break
        
    cv2.imshow("frame", (frame))
    
    key = cv2.waitKey(25)
    if key == 27: #ESC
        break

if cap.isOpened():
    cap.release()

cv2.destroyAllWindows()

(640, 480)

ビデオキャプチャと表示(YouTubeビデオ)

best.url

'https://r3---sn-nxwvob-pjoe.googlevideo.com/videoplayback?expire=1629357065&ei=qa8dYfK6EJKXgQOJj4iQBw&ip=210.126.217.236&id=o-AKMXatt3j318VG3m7DY89f3ysNe8gX4XycYJok6ZZdEe&itag=22&source=youtube&requiressl=yes&mh=zy&mm=31%2C29&mn=sn-nxwvob-pjoe%2Csn-oguelnl7&ms=au%2Crdu&mv=m&mvi=3&pl=18&gcr=kr&initcwndbps=156250&vprv=1&mime=video%2Fmp4&ns=EaCvty5y9I-bgB-5ClRtXDwG&cnr=14&ratebypass=yes&dur=687.194&lmt=1546490359129665&mt=1629335093&fvip=3&fexp=24001373%2C24007246&c=WEB&txp=5432432&n=n-K5tmqTAJYTxa0ts&sparams=expire%2Cei%2Cip%2Cid%2Citag%2Csource%2Crequiressl%2Cgcr%2Cvprv%2Cmime%2Cns%2Ccnr%2Cratebypass%2Cdur%2Clmt&sig=AOq0QJ8wRAIgcxgsG5hAWwGXssx3Fd35KDxGjrLCKUe9_PTX0XVRz9UCIGm_Wq-LAhgUl65f0p6x9tPDzo3Hvdh9NRIyKhyHVpZW&lsparams=mh%2Cmm%2Cmn%2Cms%2Cmv%2Cmvi%2Cpl%2Cinitcwndbps&lsig=AG3C_xAwRQIgBXid_hQS2Ytj9ptP-ZJrgBz2apLoCI1_7YYr-BA8TtACIQCLKEdpGpTNmp27dJdRoRn0FVfC57OgONrPeLycOZ4Neg%3D%3D'

import pafy

url = "https://www.youtube.com/watch?v=wE4JrlMGWm8&list=PLRNbTEZ7dhL28LZOlvn9PJV_pWCgULdpo"

video = pafy.new(url)

print('title = ', video.title)
print('rating = ', video.rating)
print('duration =', video.duration)

best = video.getbest()

print("best resolution = ", best.resolution)

cap = cv2.VideoCapture(best.url)

while True:
    retval, frame = cap.read() # 프레임 캡처
    if not retval:
        break
        
    cv2.imshow("frame", (frame))
    
    key = cv2.waitKey(25)
    if key == 27: #ESC
        break

if cap.isOpened():
    cap.release()

cv2.destroyAllWindows()

ビデオファイル(メインカメラから取得したビデオ)

を保存

cap = cv2.VideoCapture(0)

frame_size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)), int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))

# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 320)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 240)

print(frame_size)

fourcc = cv2.VideoWriter_fourcc(*'XVID')

fps = 20.0

out1 = cv2.VideoWriter('./out/record1.mp4', fourcc, fps, frame_size)
out2 = cv2.VideoWriter('./out/record2.mp4', fourcc, fps, frame_size, isColor=False)

while True:
    retval, frame = cap.read() # 프레임 캡처
    if not retval:
        break
        
    cv2.imshow("frame", frame)
    
    out1.write(frame)
    
    gray = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
    
#     edges = cv2.Canny(gray, 100, 200)
#     cv2.imshow("edges", edges)
    
    out2.write(gray)
    
    key = cv2.waitKey(25)
    if key == 27: #ESC
        break

if cap.isOpened():
    cap.release()

out1.release()
out2.release()
cv2.destroyAllWindows()

(640, 480)

ビデオファイルを保存
(1)Yourube動画

url = 'https://www.youtube.com/watch?v=wE4JrlMGWm8&list=PLRNbTEZ7dhL28LZOlvn9PJV_pWCgULdpo'
video = pafy.new(url)
print('title = ', video.title)
print('video.rating = ', video.rating)
print('video.duration = ', video.duration)

title =  [시즌1] 1편 1화 - 닌자고 레거시 - 스네이크군단의 부활1
video.rating =  4.7054715
video.duration =  00:11:27

best = video.getbest() 
print('best.resolution', best.resolution)

best.resolution 1280x720

cap=cv2.VideoCapture(best.url)

frame_size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)),
              int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))
print('frame_size =', frame_size)

frame_size = (1280, 720)

fourcc = cv2.VideoWriter_fourcc(*'XVID')
out = cv2.VideoWriter('./out/닌자고.mp4',fourcc, 20.0, frame_size)

while(True):
        retval, frame = cap.read()
        if not retval:
                break
        cv2.imshow('frame',frame)
        
        out.write(frame)

        key = cv2.waitKey(25)
        if key == 27: # Esc
                break
                
cap.release()
out.release()                
cv2.destroyAllWindows()

(2)ドローン動画

cap = cv2.VideoCapture('http://192.168.22.60:4747/mjpegfeed')

frame_size = (int(cap.get(cv2.CAP_PROP_FRAME_WIDTH)), int(cap.get(cv2.CAP_PROP_FRAME_HEIGHT)))

# cap.set(cv2.CAP_PROP_FRAME_WIDTH, 320)
# cap.set(cv2.CAP_PROP_FRAME_HEIGHT, 240)

print(frame_size)

(640, 480)

fourcc = cv2.VideoWriter_fourcc(*'XVID')

out = cv2.VideoWriter('./out/드로이드캠.mp4',fourcc, 20.0, frame_size)

while True:
    retval, frame = cap.read() # 프레임 캡처
    if not retval:
        break
        
    cv2.imshow("frame", (frame))
    out.write(frame)
    key = cv2.waitKey(25)
    if key == 27: #ESC
        break

if cap.isOpened():
    cap.release()
out.release()   
cv2.destroyAllWindows()

グラフィックとイベント処理

直線と矩形を描く

import numpy as np

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

pt1 = 100, 100
pt2 = 400, 400

pt3 = 0, 0
pt4 = 500, 0

pt5 = 0, 0
pt6 = 0, 500

b = np.random.randint(256)
g = np.random.randint(256)
r = np.random.randint(256)

cv2.rectangle(img, pt1, pt2, (r, g, b), 2)

cv2.line(img, pt3, pt4, (b, g, r), 5 )
cv2.line(img, pt5, pt6, (g, r, b), 5 )

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

直線と矩形の交点

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

x1, y1 = 100, 100
x2, y2 = 400, 400


cv2.rectangle(img, (x1, y1), (x2, y2), (r, g, b), 2)

x3, y3 = 120, 50
x4, y4 = 300, 500

cv2.line(img, (x3, y3), (x4, y4), (b, g, r), 2 )

imgRect = (x1, y1, x2-x1, y2-y1)
retval, rpt1, rpt2 = cv2.clipLine(imgRect, (x3, y3), (x4, y4) )

if retval:
    cv2.circle(img, rpt1, radius=10, color=(b, g, r), thickness=-1)
    cv2.circle(img, rpt2, radius=10, color=(b, g, r), thickness=-1)

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

rpt1, rpt2

((140, 100), (260, 399))

描画

元

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

cx = img.shape[0] // 2
cy = img.shape[1] // 2


for r in range(200, 0, -100):
    cv2.circle(img, (cx, cy), r, color=(255, 0, 0), thickness=5)

cv2.circle(img, (cx, cy), 50, color=(0, 0, 255), thickness=-1)

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

描画

楕円

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

cx = img.shape[0] // 2
cy = img.shape[1] // 2

size = 200, 100

cv2.ellipse(img, (cx, cy), size, 0, 0, 360, (0, 0, 255))
cv2.ellipse(img, (cx, cy), size, 45, 0, 360, (255, 0, 0))
cv2.ellipse(img, (cx, cy), size, 90, 0, 360, (0, 255, 0))

box = ((cx, cy), size, 0)
cv2.ellipse(img, box, (255, 0, 0), 5)

box = ((cx, cy), size, 45)
cv2.ellipse(img, box, (0, 0, 255), 5)

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

ポリゴン(長方形、三角形)

を描画

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255


pts1 = np.array([[100, 100], [200, 100], [200, 200], [100, 200]])
pts2 = np.array([[300, 200], [400, 100], [400, 200]])

cv2.polylines(img, [pts1, pts2], isClosed= True, color=(255, 0, 0))

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

ポリゴン(楕円内のポリゴン)

を描画

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

cx = img.shape[0] // 2
cy = img.shape[1] // 2

size = 200, 100

cv2.ellipse(img, (cx, cy), size, 0, 0, 360, (0, 0, 255))
pts1 = cv2.ellipse2Poly((cx, cy), size, 0, 0, 360, delta=45)

cv2.ellipse(img, (cx, cy), size, 45, 0, 360, (0, 0, 255))
pts2 = cv2.ellipse2Poly((cx, cy), size, 45, 0, 360, delta=45)


cv2.polylines(img, [pts1, pts2], isClosed= True, color=(255, 0, 0))


cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

ポリゴン(回転矩形)

を描画

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

cx = img.shape[0] // 2
cy = img.shape[1] // 2

size = 200, 200

for angle in range(0, 90, 10):
    box = ((cx, cy), size, angle)
    pts = cv2.boxPoints(box).astype(np.int32)
    
    b = np.random.randint(256)
    g = np.random.randint(256)
    r = np.random.randint(256)
    
    cv2.polylines(img, [pts], isClosed= True, color=(b, g, r))

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

ポリゴンパディング

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255


pts1 = np.array([[100, 100], [200, 100], [200, 200], [100, 200]])
pts2 = np.array([[300, 200], [400, 100], [400, 200]])

#cv2.fillConvexPoly(img, pts1, color=(255, 0, 0))

cv2.fillPoly(img, [pts1, pts2], color=[0, 0, 255])

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

実習

from IPython.display import Image
Image(filename='polygon.png', width=300)

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

pts1 = 50, 50
pts2 = 200, 100

pts3 = 50, 150
pts4 = 200, 200

pts5 = np.array([[300, 250], [350, 250], [350, 300], [400, 300], [400, 350], [300, 350]])

cx, cy = 150, 300
cx2, cy2 = 350, 100

size = 100, 50
size2 = 60, 30

cv2.rectangle(img, pts1, pts2, (0, 0, 255), 2)
cv2.rectangle(img, pts3, pts4, (0, 0, 50), -1)
cv2.ellipse(img, (cx, cy), size, 20, 0, 360, color=(0, 150, 0), thickness=2)
cv2.ellipse(img, (cx, cy), size2, 20, 0, 270, color=(235, 206, 135), thickness=-1)
cv2.circle(img, (cx2, cy2), 60, color=(139, 0, 0), thickness= 3)
cv2.circle(img, (cx2, cy2), 30, color=(235, 206, 135), thickness= -1)
cv2.polylines(img, [pts5], isClosed = True, color=(128, 0, 128), thickness=2)

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

文字列出力

img = np.zeros((512, 512, 3), dtype = np.uint8) + 255

text = 'OpenCV Programming'
org = (50, 100)

font = cv2.FONT_HERSHEY_SIMPLEX
cv2.putText(img, text, org, font, 1, (255, 0, 0), 2)

size, baseline = cv2.getTextSize(text, font, 1, 2)

cv2.rectangle(img, org, (org[0] + size[0], org[1] - size[1]), (0, 0, 255))
cv2.circle(img, org, 3, (0, 255, 0), 2)

cv2.imshow('img', img)
cv2.waitKey()
cv2.destroyAllWindows()

実践:カウントダウンビデオ

の作成

size2 = 300
fourcc = cv2.VideoWriter_fourcc(*'XVID') 
fps = 1
out = cv2.VideoWriter('./out/countdown.mp4', fourcc, fps, (512, 512))

for i in range(5, -1, -1):
    img = np.zeros((512, 512, 3), dtype=np.uint8) + 255 
    cx, cy = img.shape[0] // 2, img.shape[1] // 2
    text = str(i)
    font = cv2.FONT_HERSHEY_COMPLEX
    size, baseline = cv2.getTextSize(text, font, 8, 5)
    
    org = cx-(size[0]//2), cy+(size[1]//2)
    cv2.putText(img, text, org, font, 8, (0, 0, 0), 5)

    cv2.circle(img, (cx, cy), size2, (0, 255, 255), 5)
    size2 -= 40
    cv2.imshow('img', img)
    out.write(img)
    cv2.waitKey(1000)
    
cv2.waitKey(1000)
out.release()
cv2.destroyAllWindows()

import numpy as np
import cv2

img = np.zeros(shape=(512,512,3), dtype=np.uint8) + 255
cx, cy = img.shape[0]//2, img.shape[1]//2
fontFace = cv2.FONT_HERSHEY_TRIPLEX
fontScale=5

fourcc = cv2.VideoWriter_fourcc(*'XVID')
fps = 1
frame_size = (img.shape[0], img.shape[1])
out = cv2.VideoWriter('./out/countdown.mp4', fourcc, fps, frame_size)

# 카운트다운
for count in range(5, 0, -1):
    text = str(count)
    size, baseLine = cv2.getTextSize(text, fontFace, fontScale, 2)
    org = cx - size[0]//2, cy + size[1]//2
    cv2.putText(img, text, org, fontFace, fontScale, (255, 0, 0), 3)
    cv2.circle(img, (cx, cy), int(np.max(size)*count*0.5), (255, 255, 0), 4)
    cv2.imshow('img', img)
    out.write(img)
    cv2.waitKey(1000)
    img=np.zeros(shape=(512, 512, 3), dtype=np.uint8) + 255
    
cv2.destroyAllWindows()
out.release()

Reference

この問題について(Computer Vision Day 2), 我々は、より多くの情報をここで見つけました https://velog.io/@bbkyoo/Computer-Vision-Day-2-sb2p0ati

テキストは自由に共有またはコピーできます。ただし、このドキュメントのURLは参考URLとして残しておいてください。

Collection and Share based on the CC Protocol

[PROD Onboard-TIL]空間データベース

[基本]オブジェクト向けプログラミング