video_mark/video_analysis.py at master · TubmanProject/video_mark · 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
126
127
128
129
130
131
import cv2
import numpy as np
from scipy.stats import norm
import cvlib as cv
from cvlib.object_detection import draw_bbox


def frame_stream(file_name):
    vidcap = cv2.VideoCapture(file_name)
    success = True
    counter = 0
    while success:
        success, image = vidcap.read()
        if success:
            yield (image, vidcap.get(cv2.CAP_PROP_POS_MSEC), counter)
            counter += 1
    vidcap.release()


def iter_pairs(items):
    items_iter = iter(items)
    prev = next(items_iter)

    for item in items_iter:
        yield prev, item
        prev = item


def diff_stream(stream_of_frames):
    for ((im1, time1, counter1), (im2, time2, counter2)) in iter_pairs(stream_of_frames):
        yield {"diff": im1 - im2,
               "start_time": time1,
               "end_time": time2,
               "first_frame_index": counter1,
               "second_frame_index": counter2}


def moving_window_iterator(iterable, n):
    data = []
    iterator = iter(iterable)
    while len(data) < n:
        data.append(iterator.next())
    yield data
    for element in iterator:
        data = data[1:] + [element]
        yield data


def frame_diffs(file_name):
    return diff_stream(frame_stream(file_name))


def frame_moving_window(file_name, n):
    return moving_window_iterator(frame_diffs(file_name), n)


def moving_window_diff_norms(file_name, n):
    for window in frame_moving_window(file_name, n):
        diff_average = np.sum(frame_info["diff"].flatten() for frame_info in window) / float(n)
        yield {
            "norm": np.linalg.norm(diff_average),
            "start_time": window[0]["start_time"],
            "end_time": window[-1]["end_time"],
            "first_frame_index": window[0]["first_frame_index"],
            "final_frame_index": window[-1]["second_frame_index"]
        }


def find_unflattened_interesting_times(file_name, n, threshold):
    diff_norm_data = list(moving_window_diff_norms(file_name, n))
    diff_norms = [datum["norm"] for datum in diff_norm_data]
    mu = np.average(diff_norms)
    sigma2 = np.var(diff_norms)
    cut_off = norm.ppf(threshold, loc=mu, scale=np.sqrt(sigma2))
    return [{"start_time": datum["start_time"],
             "end_time": datum["end_time"],
             "first_frame_index": datum["first_frame_index"],
             "final_frame_index": datum["final_frame_index"]}
            for datum in diff_norm_data if datum["norm"] > cut_off]


def flatten_times(data):
    current_ff_info = None
    for frame_info in data:
        if current_ff_info is None:
            current_ff_info = frame_info
        else:
            if frame_info["first_frame_index"] <= current_ff_info["final_frame_index"]:
                current_ff_info["final_frame_index"] = frame_info["final_frame_index"]
                current_ff_info["end_time"] = frame_info["end_time"]
            else:
                yield current_ff_info
                current_ff_info = frame_info
    yield current_ff_info


def find_interesting_times(file_name, frames_in_window, percentage):
    return flatten_times(find_unflattened_interesting_times(file_name, frames_in_window, percentage))


def opencv_save_movie_with_frames(file_name, start_frame, end_frame):
    new_filename = "{start_frame}_{end_frame}_{file_name}".format(start_frame=start_frame,
                                                                  end_frame=end_frame,
                                                                  file_name=file_name)
    input_video = cv2.VideoCapture(file_name)
    for _ in range(start_frame):
        input_video.read()
    fourcc = cv2.VideoWriter_fourcc(*'mp4v')  # Be sure to use lower case
    width = int(input_video.get(cv2.CAP_PROP_FRAME_WIDTH))
    height = int(input_video.get(cv2.CAP_PROP_FRAME_HEIGHT))
    fps = input_video.get(cv2.CAP_PROP_FPS)
    output_video = cv2.VideoWriter(new_filename, fourcc, fps, (width, height), True)
    for _ in range(end_frame - start_frame):
        success, data = input_video.read()
        bbox, label, conf = cv.detect_common_objects(data)
        draw_bbox(data, bbox, label, conf)
        output_video.write(data)
    output_video.release()
    input_video.release()
    return new_filename


def save_interesting_times(file_name, frames_in_window=15, percentage=.90):
    for period in find_interesting_times(file_name, frames_in_window, percentage):
        print(opencv_save_movie_with_frames(file_name, period["first_frame_index"], period["final_frame_index"]))


# print(list(find_interesting_times('Surveillance_footage_of_crime_scene.mp4', 15, .90)))
# print(dir(ffmpeg))

save_interesting_times('Surveillance_footage_of_crime_scene.mp4', frames_in_window=15, percentage=.90)