生活随笔
收集整理的這篇文章主要介紹了
Python在视频处理上的优势有哪些
小編覺得挺不錯的,現(xiàn)在分享給大家,幫大家做個參考.
今天與大家分享一些Python處理視頻的一下經驗，視頻的處理和圖片的處理類似，只不過視頻處理需要連續(xù)處理一系列圖片。
一般有兩種視頻源，一種是直接從硬盤加載視頻，另一種是獲取攝像頭視頻。
一共這幾個模塊：
class videoReader 讀取視頻
class videoFramesExtractor(videoReader):繼承了讀取視頻，主要是用來限制讀取視頻中的哪些幀，并保存。
read_excel_single(excel_path,event_instance): 處理Excel中的數(shù)組，找到事件的幀數(shù)范圍：
Excel內容是這樣的，需要的只是第二列的幀范圍，另外第二列中的這些值還有重復的所以需要去除重復：
每一個范圍，例如:23450:23461之內的幀要存放在一個文件夾里。
?
?
[python]?view plaincopy
</pre><pre?name="code"?class="python">??import?cv2.cv?as?cv??import?os??import?cv2??import?numpy??import?Image??import?xlrd????class?videoReader:??????frame_count=0??????def?__init__(self,videoPath):??????????self.videoPath=videoPath??????def?video_init(self):??????????self.capture?=?cv.CaptureFromFile(self.videoPath)??????????self.win_name?=?"test"??????????cv.NamedWindow(self.win_name,?cv.CV_WINDOW_AUTOSIZE)????????????????def?read(self):??????????self.video_init()??????????while?1:??????????????self.frame_count+=1??????????????image?=?cv.QueryFrame(self.capture?)???????????????????????????cv.ShowImage(self.win_name,?image)??????????????print?self.frame_count??????????????c?=?cv.WaitKey(10)??????????????if?c?==?27:??????????????????break????????????????????????cv.DestroyWindow(self.win_name)??????class?videoFramesExtractor(videoReader):??????save_path='d:/'??????def?__init__(self,videoPath,frameSpan,image_prefix):????????videoReader.__init__(self,videoPath)??????????self.frameSpan=frameSpan??????????self.image_prefix=image_prefix????????????????def?read(self):??????????capture?=?cv2.VideoCapture(self.videoPath)??????????win_name?=?"test"??????????cv.NamedWindow(win_name,?cv.CV_WINDOW_AUTOSIZE)??????????success,frame?=?capture.read()????????????????????init_row=0??????????while?success:????????????????????????????self.frame_count+=1??????????????success,frame?=?capture.read()??????????????????????????time_duration=self.frameSpan[init_row][1]-self.frameSpan[init_row][0]+1??????????????if?self.frame_count>=self.frameSpan[init_row][0]?and?\????????????????self.frame_count<=self.frameSpan[init_row][1]?:?????????????????????????????????????????????????????????????????????????????self.image_save(frame,init_row,time_duration)??????????????????????????????????????????????????????????if?self.frame_count==self.frameSpan[init_row][1]:??????????????????????init_row+=1??????????????????if?init_row==len(self.frameSpan):??????????????????????cv.DestroyWindow(win_name)??????????????????????return?0????????????????????????????????c?=?cv.WaitKey(10)??????????????if?c?==?27:??????????????????break????????????????????????cv.DestroyWindow(win_name)??????def?image_save(self,frame,init_row,time_duration):??????????save_path=self.make_folder(init_row,time_duration)????????????????????????????????????????frame=cv2.resize(frame,(300,300))????????cv2.imwrite(?'%s//%s.jpg'?%?(save_path,self.frame_count),?frame)??????????????????def?make_folder(self,init_row,time_duration):??????????temp=?self.image_prefix+'_'+str(init_row)+'_'+str(time_duration)????????new_path?=?os.path.join(self.save_path,temp)??????????if?not?os.path.isdir(new_path):??????????????os.makedirs(new_path)??????????return?new_path????????????????def?read_excel_single(excel_path,event_instance):????????????????????data=[]??????????data=xlrd.open_workbook(excel_path)??????????????????table?=?data.sheets()[0]??????????????????nrows?=?table.nrows??????????rowmsg=[]??????????for?i?in?xrange(nrows):??????????????start_frame,end_frame=table.row_values(i)[1].split(':')??????????????rowmsg.append([int(start_frame),int(end_frame),table.row_values(i)[3]])????????????rowmsg.sort(lambda?x,y:cmp(x[0],y[0]))??????????????????event_id=[]??????????event_id=event_instance??????????????????????????????????event=[]??????????for?item?in?xrange(len(rowmsg)):??????????????if?rowmsg[item][2]==event_id:????????????????event.append(rowmsg[item])??????????if?event==[]:??????????????print?("there?is?no?%s?occurs"?%?event_instance)??????????????return?0??????????????????????????????????????????cur_pos=1??????????new_rowmsg=[]??????????pre_s=event[0][0]??????????pre_e=event[0][1]??????????cur_s=event[cur_pos][0]??????????cur_e=event[cur_pos][1]????????????while?cur_pos<len(event)-1:???????????????while?not?(cur_s>pre_e):??????????????????if?cur_s>pre_s:??????????????????????cur_s=pre_s??????????????????if?cur_e<pre_e:??????????????????????cur_e=pre_e??????????????????still_pos=[cur_s,cur_e,event_id]??????????????????cur_pos+=1??????????????????if?cur_pos==len(event)-1:??????????????????????break????????????????????????????????????????????????pre_s=cur_s??????????????????pre_e=cur_e??????????????????cur_s=event[cur_pos][0]??????????????????cur_e=event[cur_pos][1]??????????????new_rowmsg.append(still_pos)??????????????if?cur_pos==len(event)-1:??????????????????break????????????????????cur_pos+=1??????????????pre_s=cur_s??????????????pre_e=cur_e??????????????cur_s=event[cur_pos][0]??????????????cur_e=event[cur_pos][1]??????????????????????return?new_rowmsg?????????????????????????excel_source='E:/08ann/dev_20071101/LGW_20071101_E1_CAM1.mpeg.xlsx'??event_instance='CellToEar'????row_msg=read_excel_single(excel_source,event_instance)??new_row=[?item[0:2]?for?item?in?row_msg?]?????filename?=?"G:/TrecvidData/08/DEV/LGW_20071101_E1_CAM1.mpg"????????????????????videoread=videoFramesExtractor(filename,new_row,'CellToEar')??videoread.read()??
有幾點發(fā)現(xiàn)：
cv2.的函數(shù)輸出的數(shù)據(jù)類型基本都是numpy，而cv.不行。所以盡量在Python中使用cv2.的函數(shù)。
通過numpy轉換后的圖片會在顏色上有些失真，不過影響不大，如果是灰度圖的話則完全沒有影響。
總結
以上是生活随笔為你收集整理的Python在视频处理上的优势有哪些的全部內容，希望文章能夠幫你解決所遇到的問題。
如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。
