import cv2
#读取视频
videoCapture = cv2.VideoCapture('./test.mp4')
#读一帧
success, frame = videoCapture.read()
#连续读帧，保存成图片
cnt = 0
while success:jpg_name = "./frame"+str(cnt)+".jpg"cv2.imwrite(jpg_name,frame)cnt += 1success, frame = videoCapture.read() #获取下一帧

def caption_region_extr(src_img, thresh_, v_cut_ratio_):#to grayimgray = src_imgif len(src_img.shape)==3 and src_img.shape[-1]==3:imgray = cv2.cvtColor(src_img,cv2.COLOR_BGR2GRAY)#binaryth,img_bn = cv2.threshold(imgray,thresh_,255,cv2.THRESH_BINARY)#vertical cutcrop_start = int(v_cut_ratio_ * img_bn.shape[0])crop_end = img_bn.shape[0]v_cut_img = img_bn[crop_start:crop_end,:]#horizontal cuth_left = 0h_right = 0for i in range(v_cut_img.shape[1]):if np.any(v_cut_img[:,i]>0):h_left = ibreakfor i in range(v_cut_img.shape[1]-1,-1,-1):if np.any(v_cut_img[:,i]>0):h_right = ibreakh_cut_img = np.zeros(1)if (h_right-h_left)>20:#expand a littleh_left = max(h_left - 10,0)h_right = min(h_right + 10, v_cut_img.shape[1]-1)h_cut_img = v_cut_img[:,h_left:h_right+1]return h_cut_img

import pytesseract
text = pytesseract.image_to_string(cap_img)

def Equal_(region_a_, region_b_, thresh_):if region_a_.shape != region_b_.shape:return Falsea = region_a_.reshape(-1).astype(np.float64)b = region_b_.reshape(-1).astype(np.float64)a_norm = np.linalg.norm(a)b_norm = np.linalg.norm(b)similiarity = np.dot(a, b.T)/(a_norm * b_norm) dist = 1. - similiarityif dist>thresh_:return Falseelse:return True

import cv2
import numpy as np
import pytesseractdef caption_region_extr(src_img, thresh_, v_cut_ratio_):#to grayimgray = src_imgif len(src_img.shape)==3 and src_img.shape[-1]==3:imgray = cv2.cvtColor(src_img,cv2.COLOR_BGR2GRAY)#binaryth,img_bn = cv2.threshold(imgray,thresh_,255,cv2.THRESH_BINARY)#vertical cutcrop_start = int(v_cut_ratio_ * img_bn.shape[0])crop_end = img_bn.shape[0]v_cut_img = img_bn[crop_start:crop_end,:]#horizontal cuth_left = 0h_right = 0for i in range(v_cut_img.shape[1]):if np.any(v_cut_img[:,i]>0):h_left = ibreakfor i in range(v_cut_img.shape[1]-1,-1,-1):if np.any(v_cut_img[:,i]>0):h_right = ibreakh_cut_img = np.zeros(1)if (h_right-h_left)>20:#expand a littleh_left = max(h_left - 10,0)h_right = min(h_right + 10, v_cut_img.shape[1]-1)h_cut_img = v_cut_img[:,h_left:h_right+1]return h_cut_imgdef Equal_(region_a_, region_b_, thresh_):if region_a_.shape != region_b_.shape:return Falsea = region_a_.reshape(-1).astype(np.float64)b = region_b_.reshape(-1).astype(np.float64)a_norm = np.linalg.norm(a)b_norm = np.linalg.norm(b)similiarity = np.dot(a, b.T)/(a_norm * b_norm) dist = 1. - similiarityif dist>thresh_:return Falseelse:return True#获得视频的格式
videoCapture = cv2.VideoCapture('./test.mp4')#获得码率及尺寸
# fps = videoCapture.get(cv2.CAP_PROP_FPS)
# size = (int(videoCapture.get(cv2.CAP_PROP_FRAME_WIDTH)),
#  int(videoCapture.get(cv2.CAP_PROP_FRAME_HEIGHT)))
# fNUMS = videoCapture.get(cv2.CAP_PROP_FRAME_COUNT)
# print(size)
# print(fps)#读帧
success, frame = videoCapture.read()name_cnt = 0crop_ratio_ = 0.9
pre_cap_region = np.zeros(1)
ocr_=list()b_reference_ = Falsewhile success:# cv2.imshow('windows', frame) #显示# cv2.waitKey(int(1000/fps)) #延迟cap_region = caption_region_extr(frame, 200, crop_ratio_)#first captionif (len(pre_cap_region) == 1) and (len(cap_region.shape) != 1):pre_cap_region = cap_regionif b_reference_:img_name = "zimu"+str(name_cnt)+".jpg"cv2.imwrite(img_name, pre_cap_region)name_cnt += 1text = pytesseract.image_to_string(pre_cap_region)ocr_.append(text)if len(cap_region.shape) != 1:if False == Equal_(cap_region,pre_cap_region,0.1):if b_reference_:img_name = "zimu"+str(name_cnt)+".jpg"cv2.imwrite(img_name, cap_region)name_cnt += 1text = pytesseract.image_to_string(cap_region)if text!=ocr_[-1]:ocr_.append(text)pre_cap_region = cap_regionsuccess, frame = videoCapture.read() #获取下一帧videoCapture.release()with open("result.txt","w") as wf:for line in ocr_:wf.writelines(line+"n")