Snap photos with Sony API when there are significant differences in the preview

autophoto.py

from libsonyapi.camera import Camera
from libsonyapi.actions import Actions
import urllib.request
import cv2
import numpy as np
import time
import sys

DISPLAY = False
COMPARISON_DELAY = 3
TIME_STEP = 0.25
SHOT_SPACING = 1.0

KERNEL_SIZE = 0.042
WINDOW_NAME = "autophoto"
THRESHOLD = 25
COMPARISON_VALUE = 255
TRIGGER_THRESHOLD = 0.1

history = []

lastTime = 0
lastShot = 0
if DISPLAY:
    cv2.namedWindow(WINDOW_NAME, cv2.WINDOW_AUTOSIZE)

camera = Camera()  # create camera instance
#camera_info = camera.info()  # get camera camera_info
#print(camera_info)

print(camera.name)  # print name of camera
#print(camera.api_version)  # print api version of camera
#print(camera.services)
#print(camera.do("getAvailableApiList")['result'])
#camera.do(Actions.actTakePicture)  # take a picture
liveURL = camera.do(Actions.startLiveview)['result'][0]

stream = urllib.request.urlopen(liveURL)
frame = 1
buffer = bytearray()
lastJPEG = bytearray()

def snap():
    print("snap")
    camera.do(Actions.actTakePicture)
    
def getHistoric(t):
    for i in range(len(history)):
        if history[i][0] >= t - COMPARISON_DELAY:
            return history[i][1]
    if len(history):
        return history[-1][1]
    else:
        return None

def haveJPEG(jpeg):
    global lastTime, history, lastShot
    curTime = time.time()
    if curTime - lastTime < TIME_STEP:
        return
    lastTime = curTime
    nparr = np.frombuffer(jpeg, np.uint8)
    img = cv2.imdecode(nparr, cv2.IMREAD_COLOR)
    kernelSize = 1+2 * int(img.shape[0] * KERNEL_SIZE)
    blurred = cv2.GaussianBlur(img, (kernelSize, kernelSize), 0)
    prev = getHistoric(curTime)
    for i in range(len(history)):
        if history[i][0] >= curTime - COMPARISON_DELAY:
            del history[:i]
            break
    history.append((curTime,blurred))
    if curTime < lastShot + SHOT_SPACING:
        return
    if prev is not None:
        delta = cv2.threshold(cv2.absdiff(prev, blurred), THRESHOLD, COMPARISON_VALUE, cv2.THRESH_BINARY)[1]
        avg_color_per_row = np.average(delta, axis=0)
        avg_color = np.average(avg_color_per_row, axis=0)
        avg_difference = np.average(avg_color, axis=0)
        print("difference",avg_difference)
        if avg_difference >= TRIGGER_THRESHOLD:
            lastShot = curTime
            snap()
        if DISPLAY:
            cv2.resizeWindow(WINDOW_NAME, img.shape[1], img.shape[0])
            cv2.imshow(WINDOW_NAME, delta)
            if (cv2.waitKey(1) & 0xFF) == 27 or cv2.getWindowProperty(WINDOW_NAME, 0) == -1:
                cv2.destroyAllWindows()
                sys.exit(0)

def processStream(data):
    global buffer, lastJPEG
    if not lastJPEG:
        buffer += data
        try:
            offset = buffer.index(bytes((0xFF,0xD8)))
            lastJPEG = buffer[offset:]
            buffer = bytearray()
            data = bytearray()
        except ValueError:
            pass
    if lastJPEG:
        lastJPEG += data
        try:
            offset = lastJPEG.index(bytes((0xFF,0xD9)))
            buffer = lastJPEG[offset+2:]
            haveJPEG(lastJPEG[:offset+2])
            lastJPEG = None
        except ValueError:
            pass    

while True:
    data = stream.read(4096)
    processStream(bytearray(data))

#fNumber = camera.do(Actions.getFNumber)
#print(fNumber)  # prints fnumber

#camera.do(Actions.setFNumber, "5")  # set aperture to 5