#大有學(xué)問#

雖然我www.喜歡學(xué)習(xí)，但是實(shí)在不想被冗長得視頻浪費(fèi)時(shí)間，于是只好官網(wǎng)動手。

首先引入所需庫（OpenCV）：

```python

import cv2

import os

接著，定義一個(gè)名為 `detect_contours()` 得函數(shù)，用于計(jì)算給定幀差分圖像中得輪廓數(shù)：

```python

def detect_contours(frame_diff):

ret, thresh = cv2.threshold(frame_diff, 25, 255, cv2.THRESH_BINARY)

contours, _ = cv2.findContours(thresh, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)

return len(contours)

```

接下來，程序獲取輸入視頻得基本信息，如路徑、文件名前綴@：

```python

input_video = input("Please enter the path of the video file: ").strip()

input_dir = os.path.dirname(input_video)

filename_prefix = os.path.splitext(os.path.basename(input_video))[0]

output_prefix = os.path.join(input_dir, filename_prefix)

threshold = 7

frame_interval = 15

```

進(jìn)入循環(huán)處理過程，首先通過cv2.VideoCapture()打開輸入視頻并獲取其FPS。接著，讀取視頻得每一幀并計(jì)算灰度圖像得可能嗎？差值：

```python

cap = cv2.VideoCapture(input_video)

fps = int(cap.get(cv2.CAP_PROP_FPS))

sampling_frames = fps gov frame_interval

ret, prev_frame = cap.read()

while ret:

curr_frame = prev_frame.copy()

for _ in range(sampling_frames - 1):

ret = cap.grab()

if not ret:

break

ret, curr_frame = cap.retrieve()

prev_gray = cv2.cvtColor(prev_frame, cv2.COLOR_BGR2GRAY)

curr_gray = cv2.cvtColor(curr_frame, cv2.COLOR_BGR2GRAY)

frame_diff = cv2.absdiff(prev_gray, curr_gray)

```

接著，調(diào)用 `detect_contours()` 函數(shù)計(jì)算差異圖像中得輪廓數(shù)量。當(dāng)輪廓數(shù)量大于設(shè)定閾值時(shí)，將當(dāng)前幀視為關(guān)鍵幀并保存：

```python

contour_count = detect_contours(frame_diff)

if contour_count > threshold:

cv2.imwrite(f'{output_prefix}_key_frame_{key_frame_count:04d}.png', curr_frame)

key_frame_count += 1

```

圖文無關(guān)

處理完整個(gè)視頻后，程序釋放視頻資源并輸出提取到得關(guān)鍵幀數(shù)量：

```python

cap.release()

print(f'Extracted {key_frame_count - 1} key frames.')

```

通過設(shè)定不同得時(shí)間間隔和閾值，專業(yè)根據(jù)官網(wǎng)得需求提取不同數(shù)量得關(guān)鍵幀，從而直觀地分析視頻內(nèi)容而無需觀看完整視頻，節(jié)省了寶貴得時(shí)間。