fluidity.py

# /usr/bin/python\
# Python to count and display the average number of objects per second moving in a video
# Assumption shown in parameters
# @Author: Kelton Temby 2014
# Copyright Keltronix 2014

# Simple opencv tool to quantify fluidity

# Input: a recorded video
# Output: a side-by-side video showing framerate

# Do simple frame subtraction 

# visualize framerate
# visualize framerate between last different frame

import cv2
import time
import sys

if (len(sys.argv)>1): # read input video from source
	fileName=sys.argv[1]
else:
	print("Please give an input parameter")	

# Create capture device
cam=cv2.VideoCapture(fileName)

videoWidth=int(cam.get(cv2.CAP_PROP_FRAME_WIDTH))
videoHeight=int(cam.get(cv2.CAP_PROP_FRAME_HEIGHT))
videoFramerate=int(cam.get(cv2.CAP_PROP_FPS))

print(f"base video framerate: {videoFramerate}")

# Define the codec and create VideoWriter object
fourcc = cv2.VideoWriter_fourcc(*'mp4v')
video = cv2.VideoWriter('AlgVideo.mp4',fourcc,videoFramerate,(videoWidth,videoHeight))

# Parameters 
vFPS = videoFramerate # assume video is 30fps 
f = 0
framesSinceChange = 1
maxFrame = videoFramerate*10 # let it analze just 10 seconds
minFrame = 1
binaryThresh = 10

# Calculate differential image
def diffImg(t0, t1, t2):
    d1 = cv2.absdiff(t2, t1)
    d2 = cv2.absdiff(t1, t0)
    return cv2.bitwise_and(d1, d2)

def fluidity(videoFramerate, framesSinceChange):
    # E.g. 2 frames since change, framerate 60 -> 60/2 = 30 fps
    fps = int(videoFramerate/(framesSinceChange))
    return fps 

# Initialize Display
winName = "Movement Indicator"
dfont=cv2.FONT_HERSHEY_DUPLEX
fontColor=(0,0,255,255)

while f < minFrame:
	s, img = cam.read()
	f +=1
	if not s:
		break

# Read three images first:
s, img = cam.read()
t_minus = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
t = cv2.cvtColor(cam.read()[1], cv2.COLOR_RGB2GRAY)
t_plus = cv2.cvtColor(cam.read()[1], cv2.COLOR_RGB2GRAY)

# Loop through all frames in video
while f < maxFrame : #True: #(cam.get(cv2.CV_CAP_PROP_POS_FRAME) <frames):
    tstart=time.time() # Get Time for FPS calculation

    # Read next image
    prev = img
    s, img = cam.read()
    if not s:
        break	

    # Shift the frames under analysis
    t_minus = t
    t = t_plus
    t_plus = cv2.cvtColor(img, cv2.COLOR_RGB2GRAY)
    f = f+1 # keep track of frame number

    dfI=diffImg(t_minus,t,t_plus) # Differential impage
    rt, gb=cv2.threshold(dfI,binaryThresh,255,cv2.THRESH_BINARY) # Threshold to clear noisy noise
    contours,hierarchy = cv2.findContours(gb,cv2.RETR_LIST,cv2.CHAIN_APPROX_SIMPLE) # simple contour detect

    # Did the image change?
    if contours:
        framesSinceChange = 1
    else:
        framesSinceChange = framesSinceChange + 1

    currentFluidity = fluidity(videoFramerate, framesSinceChange) # Get the real time fluidity 

    # Show the feed (will turn this off for headless mode)
    cv2.putText(prev, format(f"Static Frames: {framesSinceChange}"), (50,50),dfont,2,fontColor)
    cv2.putText(prev, format(f"Fluidity: {currentFluidity} fps"),(50,150),dfont,2,fontColor)
    cv2.drawContours(prev, contours,-1,(0,0,255),1)
    cv2.drawContours(dfI, contours,-1,(0,0,255),1)
    cv2.namedWindow(winName, cv2.WND_PROP_FULLSCREEN)          
    cv2.imshow(winName,prev)
    video.write(prev)

    # Allow nice closing
    key = cv2.waitKey(30)
    if key == 27:
        cv2.destroyWindow(winName)
        break

    # debugging step through
    if key == 32:
    	time.sleep(.2)
    
    tend=time.time()-tstart

cv2.destroyAllWindows()
video.release()
cam.release()