将RTSP流录制为mp4文件
录制程序要继续添加新功能：模拟电视，板卡发送出来的是rtsp流（h264视频+alaw(pcma)音频）。
由于之前做过将rtp流（h264视频+aac音频）录制合成mp4文件（参见http://www.cnblogs.com/chutianyao/archive/2012/04/13/2446140.html），很自然的就决定将其合成为mp4文件。
但是有些不同：
（1）需要解析RTSP协议。研究了一下RFC2326，发现也不是很复杂。
rtsp分控制流和数据流：控制流就是客户端向服务端发送控制命令，包括查看节目信息、播放、停止节目等，一般是通过TCP协议通信的；数据流就是服务端将音视频数据发送到指定的地址、端口上，我们的音频和视频单独发送到两个不同的端口上，采用的是UDP协议。采用TCP或UDP，在RTSP协议中并没有明确规定，可以根据实际情况确定。
控制流采用的是HTTP文本协议，比较简单、方便调试，这个RTSP协议中也没有规定必须使用HTTP，不过一般都是采用HTTP来实现的。
大致步骤：
1. 客户端连接rtsp服务器，发送option方法；服务器返回可用的方法，通常有DESCRIBE,SETUP,PLAY,TEARDOWN等，由于板卡端的rtsp服务程序也是我们自己实现的，可以确保已经实现了这些方法，因此客户端就没有进行检查了；
2. 客户端发送DESCRIBE方法，服务器返回RTSP流的相关信息，包括video stream,audio stream的个数、码率、分辨率等参数信息；
3. 根据返回的参数信息，客户端决定要播放哪些video stream,audio stream，发送SETUP方法；
我们的RTSP流为：一个alaw audio 和一个h264 video，需要指定音视频数据分别发送到哪个端口上，通过下面的代码来构造发送消息：
1 int RTSP::Set_Setup() 2 { 3 int nRet = -1; 4 int m_nIndex = 0; 5 6 if (m_pBuf != NULL) 7 { 8 // if (m_pContentBase == NULL) 9 // { 10 // sprintf(m_pBuf, "SETUP %s/%s %s\r\n", m_strUrl.c_str(), m_pMedia->p_control, RTSP_VERSSION); 11 // } 12 // else 13 // { 14 // sprintf(m_pBuf, "SETUP %s%s %s\r\n", m_pContentBase, m_pMedia->p_control, RTSP_VERSSION); 15 // } 16 // printf("m_pContentBase:%s\n", m_pContentBase); 17 // printf("m_strUrl:%s\n", m_strUrl.c_str()); 18 // printf("m_pMedia->p_control:%s\n", m_pMedia->p_control); 19 // printf("m_pBuf:%s\n", m_pBuf); 20 sprintf(m_pBuf, "SETUP %s %s\r\n", m_pMedia->p_control, RTSP_VERSSION); 21 22 m_nIndex = strlen(m_pBuf); 23 sprintf(m_pBuf + m_nIndex, "CSeq: %d\r\n", m_nSeqNum); 24 m_nIndex = strlen(m_pBuf); 25 26 if (m_pMedia->i_media_type == VIDEO) 27 { 28 GetVideoPort(); 29 sprintf(m_pBuf + m_nIndex, "Transport: %s;%s;client_port=%d-%d\r\n", "RTP/AVP", "unicast", m_nVideoPort, m_nVideoPort + 1); 30 m_nIndex = strlen(m_pBuf); 31 } 32 else if (m_pMedia->i_media_type == AUDIO) 33 { 34 GetAudioPort(); 35 sprintf(m_pBuf + m_nIndex, "Transport: %s;%s;client_port=%d-%d\r\n", "RTP/AVP", "unicast", m_nAudioPort, m_nAudioPort + 1); 36 m_nIndex = strlen(m_pBuf); 37 } 38 39 if (m_pSession[0] != 0) 40 { 41 sprintf(m_pBuf + m_nIndex, "Session: %s\r\n", m_pSession); 42 m_nIndex = strlen(m_pBuf); 43 } 44 45 sprintf(m_pBuf + m_nIndex, "User-Agent: %s\r\n", USER_AGENT_STR); 46 m_nIndex = strlen(m_pBuf); 47 sprintf(m_pBuf + m_nIndex, "\r\n"); 48 m_nIndex = strlen(m_pBuf); 49 m_nBufSize = m_nIndex; 50 51 nRet = 0; 52 } 53 54 return nRet; 55 }
4. SETUP成功之后，通过PLAY命令就可以进行播放了：
1 int RTSP::Set_Play() 2 { 3 int nRet = -1; 4 int m_nIndex = 0; 5 6 if (m_pBuf != NULL) 7 { 8 sprintf(m_pBuf, "PLAY %s %s\r\n", m_strUrl.c_str(), RTSP_VERSSION); 9 m_nIndex = strlen(m_pBuf); 10 sprintf(m_pBuf + m_nIndex, "CSeq: %d\r\n", m_nSeqNum); 11 m_nIndex = strlen(m_pBuf); 12 sprintf(m_pBuf + m_nIndex, "Session: %s\r\n", m_pSession); 13 m_nIndex = strlen(m_pBuf); 14 sprintf(m_pBuf + m_nIndex, "Range: npt=0.000-\r\n"); 15 m_nIndex = strlen(m_pBuf); 16 sprintf(m_pBuf + m_nIndex, "User-Agent: %s\r\n", USER_AGENT_STR); 17 m_nIndex = strlen(m_pBuf); 18 sprintf(m_pBuf + m_nIndex, "\r\n"); 19 m_nIndex = strlen(m_pBuf); 20 m_nBufSize = m_nIndex; 21 22 nRet = 0; 23 } 24 25 return nRet; 26 }
这样我们就可以在刚才指定的端口上接收UDP的音视频数据了。
更详细的可以参考rtsp协议的实现。
（2）合成MP4.
我们已经知道音视频格式分别为:alaw(pcma), h264;查看文档发现，mp4v2刚好支持这两种格式，剩下就很简单了：
1 bool COutputATV::CreateMp4File(string filename) 2 { 3 m_Mp4File = MP4CreateEx(filename.c_str()); 4 if (m_Mp4File == MP4_INVALID_FILE_HANDLE) 5 { 6 return false; 7 } 8 9 MP4SetTimeScale(m_Mp4File, 90000); 10 m_nVideoTrack = MP4AddH264VideoTrack(m_Mp4File, 11 90000, //timescale 12 3214, //sample duration:/*(90000 / 25)*/ 13 /* NOTICE: 14 * why 3214? read the commets below. 15 */ 16 320, //width: 17 240, //height: 18 0x64, //sps[1] AVCProfileIndication 19 0x00, //sps[2] profile_compat 20 0x1f, //sps[3] AVCLevelIndication 21 3); // 4 bytes length before each NAL unit 22 if (m_nVideoTrack == MP4_INVALID_TRACK_ID) 23 { 24 LOG(LOG_TYPE_ERROR, "CreateMp4File():MP4AddH264VideoTrack() failed."); 25 return false; 26 } 27 MP4SetVideoProfileLevel(m_Mp4File, 0x7F); 28 29 m_nAudioTrack = MP4AddALawAudioTrack(m_Mp4File, 30 8000, //timescale 31 500); //sampleDuration. 32 /* NOTICE: 33 * in standard release of mp4v2 library(v1.9.1, and trunk-r479),the function MP4AddALawAudioTrack() does not specify the 3rd param: 34 * 'sampleDuration', it calculate a fixed duration value with the following formula: 35 * uint32_t fixedSampleDuration = (timeScale * 20)/1000; // 20mSec/Sample 36 * please read the source code of MP4AddALawAudioTrack(). 37 * they can do it in this way because RFC3551 defines PCMA(a-law) as 20msec per sample, so the duration is a fixed value, please read RFC 38 * 3551:http://www.ietf.org/rfc/rfc3551.txt 39 * but, the souce boards' we used does not follow the RFC specifition, we found the sample duration value is 500. 40 * (why the param is 500? every rtp packet contains a timestamp, the duration is the difference of two samples(not rtp packets), the same as 41 * h264 tracks in rtp). SO: 42 * I modified the declarion of MP4AddALawAudioTrack(), add the 3rd param:'sampleDuration', to pass the actual duration value,I also modified 43 * the implmention of MP4AddALawAudioTrack(). 44 * 45 * as a result: 46 * *************************** IMPORTANT *************************** 47 * when distribute the Record software, you MUST use the mp4v2 library distribute with it, 48 * please DO NOT use the standard release download from network! 49 * *********************************************************************************** 50 * 51 * we use the default value of duration when creating mp4 file, we will modify it later when begin to write the first two samples with its 52 * actual value. 53 * 54 * Added by:Zhengfeng Rao. 55 * 2012-05-08 56 */ 57 58 MP4SetTrackIntegerProperty(m_Mp4File, 59 m_nAudioTrack, 60 "mdia.minf.stbl.stsd.alaw.channels", 61 1); 62 63 if (m_nAudioTrack == MP4_INVALID_TRACK_ID) 64 { 65 LOG(LOG_TYPE_ERROR, "CreateMp4File():MP4AddAudioTrack() failed."); 66 return false; 67 } 68 MP4SetAudioProfileLevel(m_Mp4File, 0x02); 69 70 return true; 71 }
写音视频数据：
1 void COutputATV::DecodeRtp(unsigned char *pbuf, int datalength) 2 { 3 if((pbuf == NULL) || (datalength <= 0)) 4 { 5 return; 6 } 7 8 rtp_header_t rtp_header; 9 char cType = pbuf[0]; 10 11 //the 1st byte indicate the node is audio/video, it's added by the input thread, so we need to remove it. 12 pbuf += 1; 13 datalength -= 1; 14 int i_header_size = GetRtpHeader(&rtp_header, pbuf, datalength); 15 16 if(i_header_size <=0 ) 17 { 18 LOG(LOG_TYPE_ERROR, "COutputATV::DecodeRtp() Invalid header size:%d", i_header_size); 19 return; 20 } 21 22 if(cType == 'A') 23 { 24 if (rtp_header.i_pt == 0x8)//AUDIO 25 { 26 int i_size = datalength - i_header_size; 27 if (m_nAudioTimeStamp == 0) 28 { 29 m_nAudioTimeStamp = rtp_header.i_timestamp; 30 } 31 32 if (m_nAudioTimeStamp != rtp_header.i_timestamp)//got a frame 33 { 34 MP4WriteSample(m_Mp4File, m_nAudioTrack, m_pAudioFrame, m_nAudioFrameIndex); 35 m_nAudioFrameIndex = 0; 36 37 m_nAudioTimeStamp = rtp_header.i_timestamp; 38 memcpy(m_pAudioFrame + m_nAudioFrameIndex, pbuf + i_header_size, i_size); 39 m_nAudioFrameIndex += i_size; 40 } 41 else 42 { 43 memcpy(m_pAudioFrame + m_nAudioFrameIndex, pbuf + i_header_size, i_size); 44 m_nAudioFrameIndex += i_size; 45 } 46 } 47 else 48 { 49 //INVALID packet. 50 } 51 } 52 else if(cType == 'V') 53 { 54 if (rtp_header.i_pt == 0x60)// VIDEO 55 { 56 char p_save_buf[4096] = {0}; 57 int i_size = RtpToH264(pbuf, datalength, p_save_buf, &m_nNaluOkFlag, &m_nLastPktNum); 58 if(i_size <= 0) 59 { 60 DumpFrame(pbuf, datalength); 61 LOG_PERIOD(LOG_TYPE_WARN, "RtpToH264() Illegal packet, igonred. datalength = %d, i_size = %d", datalength-1, i_size); 62 return; 63 } 64 65 if (m_nVideoTimeStamp == 0) 66 { 67 m_nVideoTimeStamp = rtp_header.i_timestamp; 68 69 m_nVideoFrameIndex = 0; 70 memcpy(m_pVideoFrame + m_nVideoFrameIndex, p_save_buf, i_size); 71 m_nVideoFrameIndex += i_size; 72 } 73 74 if (m_nVideoTimeStamp != rtp_header.i_timestamp || p_save_buf[12] == 0x78) 75 { 76 if (m_nVideoFrameIndex >= 4) 77 { 78 unsigned int* p = (unsigned int*) (&m_pVideoFrame[0]); 79 *p = htonl(m_nVideoFrameIndex - 4); 80 81 MP4WriteSample(m_Mp4File, m_nVideoTrack, m_pVideoFrame, m_nVideoFrameIndex, MP4_INVALID_DURATION, 0, 1); 82 //DumpFrame(m_pVideoFrame, m_nVideoFrameIndex); 83 } 84 85 m_nVideoFrameIndex = 0; 86 m_nVideoTimeStamp = rtp_header.i_timestamp; 87 memcpy(m_pVideoFrame + m_nVideoFrameIndex, p_save_buf, i_size); 88 m_nVideoFrameIndex += i_size; 89 } 90 else 91 { 92 //printf("2.3.3*************i_size:%d, m_nVideoFrameIndex:%d\n", i_size, m_nVideoFrameIndex); 93 memcpy(m_pVideoFrame + m_nVideoFrameIndex, p_save_buf, i_size); 94 m_nVideoFrameIndex += i_size; 95 } 96 } 97 else 98 { 99 //INVALID packet. 100 } 101 } 102 else 103 { 104 //INVALID packet. 105 } 106 }
需要说明的是：
libmp4v2通过MP4AddALawAudioTrack(mp4file, timescale,sampleDuration)添加alaw音频时,第三个参数sampleDuration是我自己修改libmp4v2库添加的。
因为libmp4v2中 MP4AddALawAudioTrack接口为：MP4AddALawAudioTrack(mp4file, timescale)，sampleDuration是通过如下公式计算得到的：
uint32_t fixedSampleDuration = (timeScale * 20)/1000; // 20mSec/Sample
而这计算出来的值，并不符合我们的实际情况，所以我添加了这第三个参数，可以自己指定sample duration。
