CV_project/basicTracking.py at master · hvk3/CV_project · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
import cv2
import numpy as np
import os

def detectCameraShots(baseImages, differenceThreshold = 225000):
	prevFrame, currFrame = None, None
	cameraShots = []

	for image in sorted(os.listdir(baseImages)):
		frameCount = image.split('_')[0]
		if (int(frameCount) >= 2000):
			break
		currFrame = cv2.imread(baseImages + image)
		if (prevFrame is None):
			prevFrame = currFrame
			continue
		rows, cols = currFrame.shape[:2]

		HSVprevFrame = cv2.cvtColor(prevFrame, cv2.COLOR_BGR2HSV)
		HSVcurrFrame = cv2.cvtColor(currFrame, cv2.COLOR_BGR2HSV)

		HSVprevFrameQuads = [HSVprevFrame[ : rows / 2, : cols / 2], HSVprevFrame[ : rows / 2, cols / 2 : ], HSVprevFrame[ rows / 2 : , : cols / 2 ], HSVprevFrame[rows / 2 : , cols / 2 : ]]
		HSVcurrFrameQuads = [HSVcurrFrame[ : rows / 2, : cols / 2], HSVcurrFrame[ : rows / 2, cols / 2 : ], HSVcurrFrame[ rows / 2 : , : cols / 2 ], HSVcurrFrame[rows / 2 : , cols / 2 : ]]

		HSVprevFrameQuadsHist = [cv2.calcHist([HSVprevFrameQuads[i]], [0, 1], None, [30, 32], [0, 180, 0, 256]) for i in xrange(4)]
		HSVcurrFrameQuadsHist = [cv2.calcHist([HSVcurrFrameQuads[i]], [0, 1], None, [30, 32], [0, 180, 0, 256]) for i in xrange(4)]

		absDiff = 0.0
		for i in xrange(4):
			for j in xrange(30):
				for k in xrange(32):
					absDiff += abs(HSVprevFrameQuadsHist[i][j][k] - HSVcurrFrameQuadsHist[i][j][k])

		if (absDiff >= differenceThreshold):
			cameraShots.append(frameCount)

		prevFrame = currFrame
	return cameraShots

def getMaskedFrame(frame, groundTruthBoundingBox):
	grayFrame = cv2.cvtColor(frame, cv2.COLOR_BGR2GRAY)
	x, y, width, height = groundTruthBoundingBox
	mask = np.zeros(grayFrame.shape).astype('uint8')
	cv2.rectangle(mask, (x, y), (x + width, y + height), (255, 255, 255), thickness = -1)

	return np.bitwise_and(mask, grayFrame)

def initialDetections(frame, groundTruthBoundingBox, minDistance = 5):
	maskedGrayFrame = getMaskedFrame(frame, groundTruthBoundingBox)
	# cv2.imshow('image', maskedGrayFrame)
	# cv2.waitKey(0)
	# cv2.destroyAllWindows()
	STcorners = cv2.goodFeaturesToTrack(maskedGrayFrame, 500, 0.2, minDistance)

	return STcorners

def KLTtracker(baseImages, baseDetectedImages):
	featureParams = dict(maxCorners = 50, qualityLevel = 0.3, minDistance = 7, blockSize = 7)
	kltParams = dict( winSize  = (15, 15), maxLevel = 2, criteria = (cv2.TERM_CRITERIA_EPS | cv2.TERM_CRITERIA_COUNT, 10, 0.03))

	cameraShots = map(int, detectCameraShots(baseImages))
	prevFramePointsToTrack, currFramePointsToTrack = [], []
	prevFrame, currFrame = None, None
	groundTruthBoundingBoxes = {}

	for file in sorted(os.listdir(baseDetectedImages)):
		if ('txt' in file):
			frameCount = (file.split('_')[0])
			boundingBoxes = []
			f = open(baseDetectedImages + file, 'r')
			for line in f.readlines():
				line = map(int, line.strip().split(','))
				if (len(line) > 0):
					boundingBoxes.append(line)
			f.close()
			groundTruthBoundingBoxes[int(frameCount)] = boundingBoxes

	prevShotFrameCount, currShotFrameCount = -1, -1
	framesDetected = groundTruthBoundingBoxes.keys()
	import pdb;pdb.set_trace()

	prevShotFrameCount = min(framesDetected)
	shotBeginEnd = []
	for detectedFrame in framesDetected:
		if (detectedFrame in cameraShots):
			currShotFrameCount = detectedFrame
			shotBeginEnd.append([prevShotFrameCount, currShotFrameCount])
			prevShotFrameCount = currShotFrameCount

	for i in xrange(len(shotBeginEnd)):
		if (shotBeginEnd[i][1] - shotBeginEnd[i][0] < 60):
			stepSize = (shotBeginEnd[i][1] - shotBeginEnd[i][0]) / 6
		else:
			stepSize = 10

		STprevPoints, STcurrPoints = [], []
		for j in xrange(shotBeginEnd[i][0], shotBeginEnd[i][1] + 1, stepSize):

			if (j == shotBeginEnd[i][0]):
				imgName = str(j)
				while (len(imgName) < 5):
					imgName = '0' + imgName

				prevFrame = currFrame = cv2.imread(baseImages + imgName + '_groundTruth.bmp')
				for k in xrange(len(groundTruthBoundingBoxes[j])):
					STprevPoints.append(initialDetections(prevFrame, groundTruthBoundingBoxes[j][k]))
				continue

			if (j == shotBeginEnd[i][1]):
				prevFrame = None
				STprevPoints, STcurrPoints = [], []
				continue

			for k in xrange(len(STprevPoints)):
				currFrameGoodPoints, status, error = cv2.calcOpticalFlowPyrLK(prevFrame, currFrame, STprevPoints[k], None, **kltParams)
				prevFrameBestPoints = STprevPoints[k][status == 1]
				currFrameBestPoints = currFrameGoodPoints[status == 1]

				STprevPoints[k] = currFrameGoodPoints
				if (len(currFrameBestPoints) < 3 or len(prevFrameBestPoints) < 3):
					print j
					continue
				M = cv2.getAffineTransform(prevFrameBestPoints[:3], currFrameBestPoints[:3])
				print M
			prevFrame = currFrame