a bit of work to try to detect blinking leds...

I'm temporarily abandoning this because the issue is that the camera
isn't a great sensor.  Issues w/ long shutter speed (blurs sampling
period), and sampling frequency makes this difficult.   I cannot
find api information about how to adjust the shutter speed in app,
though there does appear to be able to do this via other apps...

blinkled/sampledata/info.txt

@@ -0,0 +1,13 @@
+https://stackoverflow.com/questions/8877228/opencv-detect-blinking-lights
+
+ffmpeg -i PXL_20220806_231420638.mp4 -filter:v "crop=200:200:725:275" output.mp4
+
+PXL_20220806_231420638.mp4	725x275	200x200
+PXL_20220806_231644224.mp4	790x380	200x200
+
+Mon Aug  8 17:24:55 PDT 2022
+
+PXL_20220808_233318199.mp4
+
+captures w/ a delay of 35ms instead of previous 34ms (for 4224.mp4)
+also w/ a smaller resistor, 68Ω instead of 100Ω.

blinkled/sampledata/t.py

@@ -0,0 +1,126 @@
+import cv2
+import numpy as np
+import sys
+
+winname = 'Diff'
+
+def tovalues(values, histo):
+	return [ 0 if x < histo[1][1] else 2 if x > histo[1][2] else 1 for x in values ]
+
+def frameiter(fname):
+	#vc = cv2.VideoCapture('output.mp4')
+	#vc = cv2.VideoCapture('PXL_20220806_231420638.mp4')
+	#vc = cv2.VideoCapture('PXL_20220806_231644224.mp4')
+	#vc = cv2.VideoCapture('PXL_20220808_233318199.mp4')
+	vc = cv2.VideoCapture('PXL_20220809_162724414.mp4')
+
+	while True:
+		r, frame = vc.read()
+		if not r:
+			break
+
+		# base image processing
+		frame = cv2.cvtColor(frame, cv2.COLOR_RGB2GRAY, 1)
+
+		yield frame
+
+def pixtot(frame):
+	colsum = np.sum(frame, 0, np.float64)
+	colsum = np.sum(colsum, 0, np.float64)
+
+	return colsum
+
+def numpixels(frame):
+	return pixtot(frame) / 255
+
+sumhist = []
+cont = False
+#cont = True
+key = 0
+
+valuehist = []
+
+basefr = None
+for frame in frameiter('output.mp4'):
+	if basefr is None:
+		basefr = frame
+		continue
+
+	frdiff = cv2.absdiff(basefr, frame)
+
+	#hist = np.histogram(frdiff, bins=5)
+	#print(repr(hist))
+	#frdiff = frame - basefr
+
+	#frdiffbin = cv2.threshold(frdiff, 50, 100, cv2.THRESH_BINARY)
+	frdiffbin = cv2.compare(frdiff, 50, cv2.CMP_GE)
+	kernel = np.ones((4,4), np.uint8)
+	frdiffbin = cv2.erode(frdiffbin, kernel, iterations = 2)
+
+	numpix = numpixels(frdiffbin)
+	print('np:', repr(numpix))
+
+	if numpix:
+		# detected change
+		# get average pixel value
+		selected = cv2.bitwise_and(frame, frdiffbin)
+
+		avg = pixtot(selected) / numpix
+		print('avg:', avg)
+
+		valuehist.append(avg)
+
+		cv2.imshow('selected', selected)
+		cv2.moveWindow('Base Frame', 1200, 300)
+
+	if len(valuehist) > 6:
+		print(repr(sorted(valuehist)))
+		hist = np.histogram(valuehist, bins=3)
+		print('hist:', repr(hist))
+
+	#print(repr(frdiffbin))
+	#moments = cv2.moments(frdiffbin)
+	#((x, y), radius) = cv2.minEnclosingCircle(conts[0][0])
+	#x = int(x)
+	#y = int(y)
+	#radius = int(radius)
+	#circ = ((x, y), radius)
+
+	#white = (255,)
+
+	#print(repr(moments))
+	#frdiff = cv2.circle(frdiff, circ[0], circ[1], white)
+	cv2.imshow('Base Frame', basefr)
+	cv2.moveWindow('Base Frame', 300, 300)
+	cv2.imshow('Diff', frdiffbin)
+	cv2.moveWindow('Diff', 600, 300)
+	cv2.imshow('Next Frame', frame)
+	cv2.moveWindow('Next Frame', 900, 300)
+
+	# sum up the diffs
+	colsum = np.sum(frdiff, 0, np.float64)
+	colsum = np.sum(colsum, 0, np.float64)
+	colsum = np.sum(colsum, 0, np.float64)
+	sumhist.append(colsum)
+	print(repr(colsum))
+
+	if not cont:
+		key = cv2.waitKey(0)
+
+	if key == ord('q'):
+		break
+	elif key == ord('c'):
+		cont = True
+
+	basefr = frame
+
+print(len(sumhist))
+hist = np.histogram(sumhist, bins=5)
+print(repr(hist))
+print(len(hist))
+print(len(hist[1]))
+
+print(repr(valuehist))
+hist = np.histogram(valuehist, bins=3)
+print(repr(hist))
+print(repr(tovalues(valuehist, hist)))