【opencv】获取视频中的“黑白蓝绿”屏

【诉求】

长视频文件中&＃xff0c;检测异常画面帧出现的次数以及截图报错&＃xff0c;替代人工检查&＃xff08;附代码&＃xff09;

【学习】

方案一&＃xff1a;像素点&＃xff1a;(height, width, channels)
方案二&＃xff1a;二级图&＃xff1a;RGB/HSV颜色

> 颜色区间&＃xff1a;https://blog.csdn.net/qq78442761/article/details/83056346
# 定义灰色识别模型
lower_orange &＃61; np.array([0, 0, 100])
upper_orange &＃61; np.array([180, 43, 220])


> opencv导入:
import cv2
import numpy as np
from numpy import array


灰屏&＃xff1a;
1.RGB与HSB颜色&＃xff0c;灰色是一个区域范围
2.cv将灰度图像二值化
3.计算灰色轮廓的模型
4.根据张数计算图片发生的时刻

【实现顺序】

1. 传入待测文件&＃xff1a;mp4、mp3
2. 对视频文件进行切割/每帧&＃xff0c;变成对图片的处理
3. 关于如何获取文件时长
4. 关于获取到问题帧/区间&＃xff1a;通过像素点占比以及数组区间范围判定
5. 如何定位时间戳&＃xff08;相对时间点&＃xff09;
6. 自动化定位/对比、打印信息

【拓展运用】

一个回放文件里&＃xff0c;灰屏帧在的时间戳以及统计次数
复现设备概率性出现绿屏或者闪现绿屏等&＃xff0c;截图并获取全屏信息

【代码】方案一

def test_video_opencv(url):
start_time_1 &＃61; time.time()
# 将切割的文件进行cp和move
rm_cmd &＃61; "rm -rf ./tmp/*.png *.txt"
os.popen(rm_cmd).read()
pre_cmd1 &＃61; "mkdir tmp"
os.popen(pre_cmd1).read()
make_file_cmd &＃61; "touch video_tmp.txt"
os.popen(make_file_cmd).read()
file &＃61; open(&＃39;./video_tmp.txt&＃39;, &＃39;w&＃39;, encoding&＃61;&＃39;utf-8&＃39;)
# 将视频转化成数据帧-调试
start_time &＃61; time.time()
cmd &＃61; "ffmpeg -i " &＃43; url &＃43; " -r 2 -s 1080,960 -ss 00:00:00 ./tmp/%d.png"
print(cmd)
os.popen(cmd).read()
end_time &＃61; time.time()
file.write("切割耗时&＃xff1a;" &＃43; str(end_time-start_time) &＃43; "\n\n")
info_cmd &＃61; "ffprobe -v quiet -print_format json -show_format -show_streams " &＃43; url
data_json &＃61; os.popen(info_cmd).read()
# 获取文件时长
d &＃61; json.loads(data_json)
duration &＃61; d["format"]["duration"]
file.write("视频总时长:" &＃43; str(duration) &＃43; "\n\n")
pic_len &＃61; len(os.listdir("../tmp"))
file.write("切割视频图片个数:"&＃43;str(pic_len) &＃43; "\n\n")
start_time &＃61; time.time()
for png_num in range(1,pic_len):
img &＃61; cv2.imread("./tmp/" &＃43; str(png_num) &＃43; ".png")
# gray &＃61; cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) # 转换了灰度化
# ret, img &＃61; cv2.threshold(gray, 160, 255, cv2.THRESH_BINARY) # 将灰度图像二值化
# img &＃61; 255 - img
# 定义灰色识别模型
lower_orange &＃61; np.array([0, 0, 100])
upper_orange &＃61; np.array([180, 43, 220])
#RGB 转 HSV 颜色模型
hsv &＃61; cv2.cvtColor(img, cv2.COLOR_BGR2HSV)
# 低于lower_orange/高于upper_orange的值&＃xff0c;图像值为0
# 在此之间的值变成255&＃xff0c;即涂白色
mask &＃61; cv2.inRange(hsv, lower_orange, upper_orange)
# cv2.imshow(&＃39;image&＃39;, mask)
# cv2.waitKey(0)
binary &＃61; cv2.threshold(mask, 127, 255, cv2.THRESH_BINARY)[1] # 将灰度图像二值化
binary &＃61; cv2.dilate(binary, None, iterations&＃61;2) # 图像膨胀
# cv2的版本不同,findContours的返回值个数也不一样&＃xff0c;此处是为了兼容不同版本
if int(cv2.__version__[0]) > 2:
contours, _ &＃61; cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
else:
_, contours, _ &＃61; cv2.findContours(binary, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
# 3-求轮廓的面积
pic_sum &＃61; 0
space &＃61; img.shape[0] * img.shape[1]
for cts in contours:
pic_sum &＃43;&＃61; cv2.contourArea(cts)
print(pic_sum/space)
if pic_sum / space > 0.8: # 这里假定是80%&＃xff0c;
file.write("第 " &＃43; str(png_num) &＃43; " 张图片灰屏&＃xff0c;在第 " &＃43; str(float(duration)/pic_len * png_num) &＃43; " 秒&＃xff0c; 灰屏面积占比:" &＃43; str(pic_sum / space) &＃43; "\n\n")
end_time &＃61; time.time()
file.write("循环判断耗时:"&＃43; str(end_time - start_time) &＃43; "\n\n")
end_time_1 &＃61; time.time()
file.write("总耗时&＃xff1a;" &＃43; str(end_time_1 - start_time_1) &＃43; "\n\n")


【代码】方案二

def test_video_pixel(url):
# 将切割的文件进行cp和move
rm_cmd &＃61; "rm -rf ./tmp/*.png"
os.popen(rm_cmd).read()
pre_cmd1 &＃61; "mkdir tmp"
os.popen(pre_cmd1).read()
make_file_cmd &＃61; "touch video_tmp.txt"
os.popen(make_file_cmd).read()
file &＃61; open(&＃39;./video_tmp.txt&＃39;, &＃39;w&＃39;, encoding&＃61;&＃39;utf-8&＃39;)
# 将视频转化成数据帧-调试l
cmd &＃61; "ffmpeg -i " &＃43; url &＃43; " -r 1 -s 1080,960 -ss 00:00:00 ./tmp/%03d.png"
os.popen(cmd).read()
info_cmd &＃61; "ffprobe -v quiet -print_format json -show_format -show_streams " &＃43; url
data_json &＃61; os.popen(info_cmd).read()
# 获取文件时长
d &＃61; json.loads(data_json)
duration &＃61; d["format"]["duration"]
file.write("视频总时长:" &＃43; str(duration) &＃43; "\n\n")
pic_len &＃61; len(os.listdir("../tmp"))
# 循环将每张图片转换为numpy数组
for png_num in range(1,pic_len&＃43;1):
start_time &＃61; time.time()
im &＃61; array(Image.open("./tmp/00" &＃43; str(png_num) &＃43;".png").convert("RGB"))
image_arr &＃61; np.array(im)
img_red &＃61; image_arr[:, :, 0]
height, width &＃61; img_red.shape
pic_num &＃61; 0
gray_pic_sum &＃61; 0
black_pic_sum &＃61; 0
green_pic_sum &＃61; 0
for i in range(0, height):
for j in range(0, width):
# 判断每一点的像素值 &＃61;&＃61; 灰色&＃xff0c;有多少个
gray_pic_sum &＃43;&＃61; 1
black_pic_sum &＃43;&＃61; 1
green_pic_sum &＃43;&＃61; 1
if (image_arr[i, j,] &＃61;&＃61; [140, 143, 146]).all():
gray_pic_sum &＃61; gray_pic_sum &＃43; 1
if (image_arr[i, j,] &＃61;&＃61; [0, 0, 0]).all():
black_pic_sum &＃61; black_pic_sum &＃43; 1
if (image_arr[i, j,] &＃61;&＃61; [40, 220, 60]).all():
green_pic_sum &＃61; green_pic_sum &＃43; 1
file.write("第 " &＃43; str(png_num) &＃43; " 张图片&＃xff0c;一共 " &＃43; str(gray_pic_sum) &＃43; " 个灰色像素点&＃xff0c;共 " &＃43; str(gray_pic_sum) &＃43; " 个像素点 \n\n")
file.write("第 " &＃43; str(png_num) &＃43; " 张图片&＃xff0c;一共 " &＃43; str(black_pic_sum) &＃43; " 个黑色像素点&＃xff0c;共 " &＃43; str(black_pic_sum) &＃43; " 个像素点 \n\n")
file.write("第 " &＃43; str(png_num) &＃43; " 张图片&＃xff0c;一共 " &＃43; str(green_pic_sum) &＃43; " 个绿色像素点&＃xff0c;共 " &＃43; str(black_pic_sum) &＃43; " 个像素点 \n\n")
# 计算问题图片的时长
if pic_num/gray_pic_sum >&＃61; 0.2:
file.write("!! 第 " &＃43; str(png_num) &＃43; " 是灰屏问题帧&＃xff0c;时长是:" &＃43; str(float(duration)/pic_len * png_num) &＃43; " \n\n")
if pic_num/black_pic_sum >&＃61; 0.2:
file.write("!! 第 " &＃43; str(png_num) &＃43; " 是黑屏问题帧&＃xff0c;时长是:" &＃43; str(float(duration)/pic_len * png_num) &＃43; " \n\n")
if pic_num/green_pic_sum >&＃61; 0.2:
file.write("!! 第 " &＃43; str(png_num) &＃43; " 是绿屏问题帧&＃xff0c;时长是:" &＃43; str(float(duration)/pic_len * png_num) &＃43; " \n\n")
end_time &＃61; time.time()
file.write("该帧判断像素点耗时&＃xff1a;" &＃43; str(end_time - start_time) &＃43; "\n\n")


【效率问题】
方案一像素点判定&＃xff0c;因为需要将每张图片像素化&＃xff0c;建议是将1080X720P的图片裁剪到360X640P
方案二图片二级值&＃xff0c;处理成黑白&＃xff0c;明显效率更高&＃xff0c;推荐&＃xff5e;