python 视频处理，提取视频相关帧，读取Excel

一共这几个模块：

class videoReader 读取视频

class videoFramesExtractor(videoReader):继承了读取视频，主要是用来限制读取视频中的哪些帧，并保存。

read_excel_single(excel_path,event_instance): 处理Excel中的数组，找到事件的帧数范围：

Excel内容是这样的，需要的只是第二列的帧范围，另外第二列中的这些值还有重复的所以需要去除重复：

每一个范围，例如:23450:23461之内的帧要存放在一个文件夹里。

</pre><pre name="code" class="python"># -*- coding: cp936 -*-

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):#image_prefix=event_name
        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()
            #cv2.imshow(win_name,frame) #显示照片浪费时间
            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=Image.fromarray(frame)
        #frame.resize((300,300))
        #frame.save('%s//%s.jpg' % (save_path,self.frame_count))
        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)#事件名称_事件顺序_
        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)
        #read the first sheets
        table = data.sheets()[0]
        #read the num fo cols
        nrows = table.nrows
        rowmsg=[]#用来存储帧的范围，及事件的类别，描述。

        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]))#按照数字大小排序~
        
        event_id=[]
        event_id=event_instance
        #print event_id

            
        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
        #去除重复
                      
        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()

有几点发现：

cv2.的函数输出的数据类型基本都是numpy，而cv.不行。所以尽量在python中使用cv2.的函数。

通过numpy转换后的图片会在颜色上有些失真，不过影响不大，如果是灰度图的话则完全没有影响。