1.1 海思3518 H264编码
作者：互联网
海思的SDK里其实有H264编码的sample，但因为要匹配很多东西，代码有点复杂，让初学都感到有点混乱。我根据sample自己修改了一下代码，从最简单的情况(确定摄像头类型，只选一种尺寸的图片，只用一个通道)来说明海思HI3518是怎么编码为H264的。
先把源代码下载下来，再分析程序。
下载链接：https://download.csdn.net/download/zhanshenrui/10324766
首先从main函数开始。
//./myvenc int main(int argc, char *argv[])//main() { HI_S32 s32Ret; signal(SIGINT, SAMPLE_VENC_HandleSig);//ctrl+c,delete signal(SIGTERM, SAMPLE_VENC_HandleSig);//shell命令kill缺省产生这个信号. s32Ret = H264_Venc();<span class="hljs-keyword">if</span>(s32Ret==HI_SUCCESS) <span class="hljs-built_in">printf</span>(<span class="hljs-string">"normally\n"</span>); <span class="hljs-keyword">else</span> <span class="hljs-built_in">printf</span>(<span class="hljs-string">"unnormally\n"</span>); <span class="hljs-keyword">return</span> <span class="hljs-number">-1</span>;
}
main函数里先是用signal定义了两个信号来中断程序的运行，在板子上输入./myvenc 执行程序，按ctrl+c来中断程序，然后再往下就是H264编码函数H264_Venc()。
在H264_Venc()函数里先去初始化MPI系统，然后根据摄像头设置参数。我这里是AR030摄像头，其拍摄图片大小为720P,编码为H264,为了简单只输出一路。在初始化MPI系统时要计算视频缓存池VB_CONF_S的大小，然后填充VB_CONF_S结构体。填充完VB_CONF_S结构体后再用HI_MPI_VB_SetConf设置，最后初始化视频视频缓存池HI_MPI_VB_Init。这里注意的是基本上每个函数调用后都会判断返回值并作出错处理，后面很多函数也是这样处理的，这样利于程序排错。
//为了简单点，编码类型就选择PT_H264，图片大小就选择PIC_HD720，通道也只用1路s32ChnNum=1
PAYLOAD_TYPE_E enPayLoad=PT_H264;//264编码
PIC_SIZE_E enSize=PIC_HD720;//摄像头拍摄图片的大小，这里只用720P
HI_S32 s32ChnNum=1;//支持一路摄像
HI_S32 s32Ret=HI_FAILURE;<span class="hljs-comment">/****************************************** mpp system init. ******************************************/</span> HI_MPI_SYS_Exit(); HI_MPI_VB_Exit(); VB_CONF_S stVbConf;<span class="hljs-comment">//缓存池参数结构体</span> HI_U32 u32BlkSize;<span class="hljs-comment">//一张图片占多少字节</span> <span class="hljs-built_in">memset</span>(&stVbConf,<span class="hljs-number">0</span>,<span class="hljs-keyword">sizeof</span>(VB_CONF_S)); <span class="hljs-comment">//根据制式，图片大小，图片格式及对齐方式确定图片缓存大小</span> <span class="hljs-comment">//这里用NTSC,720P,YUV420,64字节对齐</span> u32BlkSize=SAMPLE_COMM_SYS_CalcPicVbBlkSize(gs_enNorm,enSize, PIXEL_FORMAT_YUV_SEMIPLANAR_420, SAMPLE_SYS_ALIGN_WIDTH); <span class="hljs-built_in">printf</span>(<span class="hljs-string">"u32BlkSize=%d\n"</span>,u32BlkSize); stVbConf.u32MaxPoolCnt = <span class="hljs-number">128</span>;<span class="hljs-comment">//用默认值，3518默认是128</span> stVbConf.astCommPool[<span class="hljs-number">0</span>].u32BlkSize = u32BlkSize; stVbConf.astCommPool[<span class="hljs-number">0</span>].u32BlkCnt = g_u32BlkCnt; s32Ret = HI_MPI_VB_SetConf(&stVbConf);<span class="hljs-comment">//设置 MPP 视频缓存池属性</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VB_SetConf failed!\n"</span>); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VB_Init();<span class="hljs-comment">//初始化 MPP 视频缓存池</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VB_Init failed!\n"</span>); <span class="hljs-keyword">return</span> HI_FAILURE; }</code></pre><span></span>视频缓存池初始化后就是初始化MPI系统。海思模块化了各个功能，我们想实现哪一种产品就按这种产品的框架来实现各个模块功能，而实现模块化功能也很简单，基本套路就是定义一个结构体，给结构体赋值，然后设置就可以了，像前面缓存池模块就是先定一个VB_CONF_S结构体，然后对结体体赋值，然后调用函数设置进去，最后初始化。同理，初始化MPI系统也要先定义一个结构体再赋值然后设置最后初始化，这种思路在linux里用得很多。<p></p><p><span style="line-height:28px;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs"><span class="hljs-comment">//定义结构体，对结构体赋值，然后设置，最后初始化</span> MPP_SYS_CONF_S stSysConf = {<span class="hljs-number">0</span>}; stSysConf.u32AlignWidth = SAMPLE_SYS_ALIGN_WIDTH; s32Ret = HI_MPI_SYS_SetConf(&stSysConf);<span class="hljs-comment">//配置系统控制参数</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_SYS_SetConf failed\n"</span>); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_SYS_Init();<span class="hljs-comment">//初始化 MPP 系统</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_SYS_Init failed!\n"</span>); <span class="hljs-keyword">return</span> HI_FAILURE; }</code></pre>系统初始化完后还要打开硬件，摄像头用的是MIPI，应用层只需要用open打开MIPI,用ioctl设置<p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs">HI_S32 fdmipi; <span class="hljs-keyword">combo_dev_attr_t</span> *pstcomboDevAttr = <span class="hljs-literal">NULL</span>; fdmipi = open(<span class="hljs-string">"/dev/hi_mipi"</span>, O_RDWR); <span class="hljs-keyword">if</span> (fdmipi < <span class="hljs-number">0</span>) { <span class="hljs-built_in">printf</span>(<span class="hljs-string">"warning: open hi_mipi dev failed\n"</span>); <span class="hljs-keyword">return</span> <span class="hljs-number">-1</span>; } pstcomboDevAttr = &MIPI_CMOS3V3_ATTR;<span class="hljs-comment">//AR0130模组属性</span> <span class="hljs-keyword">if</span> (ioctl(fdmipi, HI_MIPI_SET_DEV_ATTR, pstcomboDevAttr)) { <span class="hljs-built_in">printf</span>(<span class="hljs-string">"set mipi attr failed\n"</span>); close(fdmipi); <span class="hljs-keyword">return</span> HI_FAILURE; } close(fdmipi);</code></pre>根据海思SDK里的文档介绍，H264编码过程应该是：系统初始化--VI--VPSS-VENC.系统初始化前面已完成，现在要做的就是实现输入模块VI.VI模块有通道，所以还要设置通道参数。具体解释和代码如下,不再做过多解释：<p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs">HI_S32 i,ViChn=<span class="hljs-number">0</span>;<span class="hljs-comment">//只有一路输出，所以ViChn=0</span> <span class="hljs-comment">//设置输入配置参数</span> SAMPLE_VI_CONFIG_S stViConfig = {<span class="hljs-number">0</span>}; stViConfig.enViMode = APTINA_AR0130_DC_720P_30FPS;<span class="hljs-comment">//摄像头是AR0130模组</span> stViConfig.enRotate = ROTATE_NONE;<span class="hljs-comment">//不翻转</span> stViConfig.enNorm = VIDEO_ENCODING_MODE_AUTO;<span class="hljs-comment">//编码模式自动</span> stViConfig.enViChnSet = VI_CHN_SET_NORMAL;<span class="hljs-comment">//普通</span> stViConfig.enWDRMode = WDR_MODE_NONE;<span class="hljs-comment">//不设置宽动态</span> VI_DEV_ATTR_S stViDevAttr; VI_DEV ViDev=<span class="hljs-number">0</span>;<span class="hljs-comment">//只有一个摄像头设备，所以设备序号为0</span> ISP_DEV s32IspDev=<span class="hljs-number">0</span>;<span class="hljs-comment">//同样，ISP序号也为0</span> <span class="hljs-built_in">memset</span>(&stViDevAttr,<span class="hljs-number">0</span>,<span class="hljs-keyword">sizeof</span>(stViDevAttr)); <span class="hljs-built_in">memcpy</span>(&stViDevAttr,&DEV_ATTR_9M034_DC_720P_BASE,<span class="hljs-keyword">sizeof</span>(stViDevAttr)); <span class="hljs-comment">/****************************************** step 1: mipi configure ******************************************/</span> <span class="hljs-comment">/* s32Ret = SAMPLE_COMM_VI_StartMIPI(&stViConfig); if (HI_SUCCESS != s32Ret) { SAMPLE_PRT("%s: MIPI init failed!\n", __FUNCTION__); return HI_FAILURE; } */</span> <span class="hljs-comment">/****************************************** step 2: configure sensor and ISP (include WDR mode). note: you can jump over this step, if you do not use Hi3516A interal isp. //虽然说了不用，但还要需要，因为后面HI_MPI_ISP_GetWDRMode函数会调用ISP_CHECK_MEM_INIT //检测内存，所以还是在这里使用 ******************************************/</span> s32Ret = SAMPLE_COMM_ISP_Init(stViConfig.enWDRMode); <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"%s: Sensor init failed!\n"</span>, __FUNCTION__); <span class="hljs-keyword">return</span> HI_FAILURE; } <span class="hljs-comment">/****************************************** step 3: run isp thread note: you can jump over this step, if you do not use Hi3516A interal isp. ******************************************/</span> s32Ret = SAMPLE_COMM_ISP_Run(); <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"%s: ISP init failed!\n"</span>, __FUNCTION__); <span class="hljs-comment">/* disable videv */</span> <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret=HI_MPI_VI_SetDevAttr(ViDev, &stViDevAttr);<span class="hljs-comment">//只有一个设备，所以设备号为0</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VI_SetDevAttr failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } ISP_WDR_MODE_S stWdrMode; s32Ret = HI_MPI_ISP_GetWDRMode(s32IspDev, &stWdrMode); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_ISP_GetWDRMode failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } VI_WDR_ATTR_S stWdrAttr; stWdrAttr.enWDRMode = stWdrMode.enWDRMode; stWdrAttr.bCompress = HI_FALSE; s32Ret = HI_MPI_VI_SetWDRAttr(ViDev, &stWdrAttr); <span class="hljs-keyword">if</span> (s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VI_SetWDRAttr failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VI_EnableDev(ViDev); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VI_EnableDev failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } RECT_S stCapRect; SIZE_S stTargetSize; stCapRect.s32X = <span class="hljs-number">0</span>; stCapRect.s32Y = <span class="hljs-number">0</span>; stCapRect.u32Width = <span class="hljs-number">1280</span>; stCapRect.u32Height = <span class="hljs-number">720</span>;<span class="hljs-comment">//AR030输出1280*720图像</span> stTargetSize.u32Width = stCapRect.u32Width; stTargetSize.u32Height = stCapRect.u32Height; <span class="hljs-comment">//设置通道属性</span> VI_CHN_ATTR_S stChnAttr; <span class="hljs-built_in">memcpy</span>(&stChnAttr.stCapRect, &stCapRect, <span class="hljs-keyword">sizeof</span>(RECT_S)); stChnAttr.enCapSel = VI_CAPSEL_BOTH; stChnAttr.stDestSize.u32Width = stTargetSize.u32Width; stChnAttr.stDestSize.u32Height = stTargetSize.u32Height; stChnAttr.enPixFormat = PIXEL_FORMAT_YUV_SEMIPLANAR_420; stChnAttr.bMirror = HI_FALSE; stChnAttr.bFlip = HI_FALSE; stChnAttr.s32SrcFrameRate = <span class="hljs-number">-1</span>; stChnAttr.s32DstFrameRate = <span class="hljs-number">-1</span>; stChnAttr.enCompressMode = COMPRESS_MODE_NONE; s32Ret = HI_MPI_VI_SetChnAttr(ViChn, &stChnAttr);<span class="hljs-comment">//设置通道属性</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"in HI_MPI_VI_SetChnAttr failed with %#x!\n"</span>, s32Ret); SAMPLE_COMM_ISP_Stop(); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VI_EnableChn(ViChn);<span class="hljs-comment">//使能通道</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VI_EnableChn failed with %#x!\n"</span>, s32Ret); SAMPLE_COMM_ISP_Stop(); <span class="hljs-keyword">return</span> HI_FAILURE; }</code></pre>VI模块设置完以后就要设置VPSS模块，VPSS模块有组，通道概念，所以设置VPSS模块也要设置VPSS组、VPSS通道。<p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;">设置VPSS组过程如下：定义结构体，给结构体赋值，设置(创建的同时就设置了)，启动</span></span></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs">SIZE_S stSize; stSize.u32Width = <span class="hljs-number">1280</span>; stSize.u32Height = <span class="hljs-number">720</span>; VPSS_GRP_ATTR_S stVpssGrpAttr; VPSS_NR_PARAM_U unNrParam = {{<span class="hljs-number">0</span>}}; VPSS_GRP VpssGrp=<span class="hljs-number">0</span>;<span class="hljs-comment">//只有一个设备，序号为0</span> stVpssGrpAttr.u32MaxW = stSize.u32Width; stVpssGrpAttr.u32MaxH = stSize.u32Height; stVpssGrpAttr.bIeEn = HI_FALSE; stVpssGrpAttr.bNrEn = HI_TRUE; stVpssGrpAttr.bHistEn = HI_FALSE; stVpssGrpAttr.bDciEn = HI_FALSE; stVpssGrpAttr.enDieMode = VPSS_DIE_MODE_NODIE; stVpssGrpAttr.enPixFmt = PIXEL_FORMAT_YUV_SEMIPLANAR_420; s32Ret = HI_MPI_VPSS_CreateGrp(VpssGrp, &stVpssGrpAttr);<span class="hljs-comment">//创建组的同时将组的属性设置进去</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VPSS_CreateGrp failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_GetNRParam(VpssGrp, &unNrParam); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VPSS_GetNRParam failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_SetNRParam(VpssGrp, &unNrParam); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VPSS_SetNRParam failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_StartGrp(VpssGrp);<span class="hljs-comment">//启动VPSS组</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VPSS_StartGrp failed with %#x\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; }</code></pre>VPSS还要与VI绑定，以便VI的数据直接输入VPSS，不需要我们手动从VI复制数据到VPSS,绑定时将绑定源设置为VI的哪<br style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;">一个通道，目的源设置为VPSS的一个通道</span><pre><code class="language-cpp hljs">SAMPLE_VI_PARAM_S stViParam; stViParam.s32ViDevCnt = <span class="hljs-number">1</span>; stViParam.s32ViDevInterval = <span class="hljs-number">1</span>; stViParam.s32ViChnCnt = <span class="hljs-number">1</span>; stViParam.s32ViChnInterval = <span class="hljs-number">1</span>; MPP_CHN_S stSrcChn; MPP_CHN_S stDestChn; stSrcChn.enModId = HI_ID_VIU; stSrcChn.s32DevId = <span class="hljs-number">0</span>; stSrcChn.s32ChnId = ViChn; stDestChn.enModId = HI_ID_VPSS; stDestChn.s32DevId = VpssGrp; stDestChn.s32ChnId = <span class="hljs-number">0</span>; s32Ret = HI_MPI_SYS_Bind(&stSrcChn, &stDestChn);<span class="hljs-comment">//将VI的0通道与VPSS的组的0通道绑定起来</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; }</code></pre>前面设置了VPSS组，将VI通道与VPSS组里的0通道绑定起来，但还未设置0通道，下面设置0通道.还是老套路，定义结构体，给结构体赋值，将结构体设置进去，开启通道。<br><pre><code class="language-cpp hljs"><span class="hljs-comment">//设置VPSS通道，只有一个通道，所以VpssChn=0</span> VPSS_CHN VpssChn=<span class="hljs-number">0</span>; VPSS_CHN_ATTR_S stVpssChnAttr; VPSS_CHN_MODE_S stVpssChnMode; stVpssChnMode.enChnMode = VPSS_CHN_MODE_USER; stVpssChnMode.bDouble = HI_FALSE; stVpssChnMode.enPixelFormat = PIXEL_FORMAT_YUV_SEMIPLANAR_420; stVpssChnMode.u32Width = stSize.u32Width; stVpssChnMode.u32Height = stSize.u32Height; stVpssChnMode.enCompressMode = COMPRESS_MODE_SEG; <span class="hljs-built_in">memset</span>(&stVpssChnAttr, <span class="hljs-number">0</span>, <span class="hljs-keyword">sizeof</span>(stVpssChnAttr)); stVpssChnAttr.s32SrcFrameRate = <span class="hljs-number">-1</span>; stVpssChnAttr.s32DstFrameRate = <span class="hljs-number">-1</span>; s32Ret = HI_MPI_VPSS_SetChnAttr(VpssGrp, VpssChn, &stVpssChnAttr); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VPSS_SetChnAttr failed with %#x\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_SetChnMode(VpssGrp, VpssChn, &stVpssChnMode); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"%s failed with %#x\n"</span>, __FUNCTION__, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_EnableChn(VpssGrp, VpssChn); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VPSS_EnableChn failed with %#x\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; }</code></pre>上面将VPSS设置完成，按照框架，下一个应该设置VENC模块，跟VPSS一样，老套路设置，然后绑定。<p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs"><span class="hljs-comment">//定义结构体，给结构体赋值，创建通道同时将结构体设置进去</span> SAMPLE_RC_E enRcMode= SAMPLE_RC_CBR; VENC_CHN VencChn=<span class="hljs-number">0</span>; VENC_CHN_ATTR_S stVencChnAttr; VENC_ATTR_H264_S stH264Attr; VENC_ATTR_H264_CBR_S stH264Cbr; stVencChnAttr.stVeAttr.enType =enPayLoad; stH264Attr.u32MaxPicWidth =stSize.u32Width; stH264Attr.u32MaxPicHeight = stSize.u32Height; stH264Attr.u32PicWidth = stSize.u32Width;<span class="hljs-comment">/*the picture width*/</span> stH264Attr.u32PicHeight = stSize.u32Height;<span class="hljs-comment">/*the picture height*/</span> stH264Attr.u32BufSize = stSize.u32Width * stSize.u32Height;<span class="hljs-comment">/*stream buffer size*/</span> stH264Attr.u32Profile = <span class="hljs-number">0</span>;<span class="hljs-comment">/*0: baseline; 1:MP; 2:HP; 3:svc_t */</span> stH264Attr.bByFrame = HI_TRUE;<span class="hljs-comment">/*get stream mode is slice mode or frame mode?*/</span> stH264Attr.u32BFrameNum = <span class="hljs-number">0</span>;<span class="hljs-comment">/* 0: not support B frame; >=1: number of B frames */</span> stH264Attr.u32RefNum = <span class="hljs-number">1</span>;<span class="hljs-comment">/* 0: default; number of refrence frame*/</span> <span class="hljs-built_in">memcpy</span>(&stVencChnAttr.stVeAttr.stAttrH264e, &stH264Attr, <span class="hljs-keyword">sizeof</span>(VENC_ATTR_H264_S)); stVencChnAttr.stRcAttr.enRcMode = VENC_RC_MODE_H264CBR; stH264Cbr.u32Gop = (VIDEO_ENCODING_MODE_PAL== gs_enNorm)?<span class="hljs-number">25</span>:<span class="hljs-number">30</span>; stH264Cbr.u32StatTime = <span class="hljs-number">1</span>; <span class="hljs-comment">/* stream rate statics time(s) */</span> stH264Cbr.u32SrcFrmRate = (VIDEO_ENCODING_MODE_PAL== gs_enNorm)?<span class="hljs-number">25</span>:<span class="hljs-number">30</span>;<span class="hljs-comment">/* input (vi) frame rate */</span> stH264Cbr.fr32DstFrmRate = (VIDEO_ENCODING_MODE_PAL== gs_enNorm)?<span class="hljs-number">25</span>:<span class="hljs-number">30</span>;<span class="hljs-comment">/* target frame rate */</span> stH264Cbr.u32BitRate = <span class="hljs-number">1024</span>*<span class="hljs-number">2</span>; stH264Cbr.u32FluctuateLevel = <span class="hljs-number">0</span>; <span class="hljs-comment">/* average bit rate */</span> <span class="hljs-built_in">memcpy</span>(&stVencChnAttr.stRcAttr.stAttrH264Cbr, &stH264Cbr, <span class="hljs-keyword">sizeof</span>(VENC_ATTR_H264_CBR_S)); s32Ret = HI_MPI_VENC_CreateChn(VencChn, &stVencChnAttr);<span class="hljs-comment">//创建通道同时将结构体设置进去</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VENC_CreateChn [%d] faild with %#x!\n"</span>,VencChn, s32Ret); <span class="hljs-keyword">return</span> s32Ret; }</code></pre>设置完以后，开启接收图片并将VPSS通道与VENC通道绑定起来。VPSS通道作为源通道，VENC通道作为目的通道<p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs"><span class="hljs-comment">//将VPSS通道作为源通道与VENC通道作为目的通道绑定起来</span> stSrcChn.enModId = HI_ID_VPSS; stSrcChn.s32DevId = VpssGrp; stSrcChn.s32ChnId = VpssChn; stDestChn.enModId = HI_ID_VENC; stDestChn.s32DevId = <span class="hljs-number">0</span>; stDestChn.s32ChnId = VencChn; s32Ret=HI_MPI_SYS_Bind(&stSrcChn, &stDestChn); <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; }</code></pre>这样，整个系统我们就设置好了，下一步要操作地是开始接收H264码流并将码流保存起来。这里用到多线程，用pthread_create创建个线程函数，在线程函数里用select多路IO复用来获取码流并保存。<pre><code class="language-cpp hljs">gs_stPara.bThreadStart = HI_TRUE; gs_stPara.s32Cnt = s32ChnNum; pthread_create(&gs_VencPid, <span class="hljs-number">0</span>, VENC_GetVencStreamProc, (HI_VOID*)&gs_stPara); <span class="hljs-built_in">printf</span>(<span class="hljs-string">"please press twice ENTER to exit this sample\n"</span>);</code></pre>gs_stPara是自己定义的一个结构体，里面的成员变量bThreadStart为真时表示不断读取保存H264码流，当接连有两次按下键盘时或有ctrl+c按下时，这个成员变量会为假，从而会停止VENC_GetVencStreamProc线程函数。<span style="color:rgb(79,79,79);line-height:28px;text-align:justify;">VENC_GetVencStreamProc线程函数可以说是获取H264码流的最关键的函数，我们稍会再说，先按正常流程说下后面代码的作用。后面的代码如下：</span><p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs">getchar(); getchar(); <span class="hljs-keyword">if</span> (HI_TRUE == gs_stPara.bThreadStart) { gs_stPara.bThreadStart = HI_FALSE; pthread_join(gs_VencPid, <span class="hljs-number">0</span>); }</code></pre>getchar（）函数的执行模式是阻塞式的，当需要接收字符流的时候，当前线程(main函数所在线程)就会被挂起，其后的所有代码均要等待用户输入回车表示输入完毕后，线程才会被调度执行余下的代码，即进入if让gs_stPara.bThreadStart 为假，同时调用pthread_join(gs_VencPid, 0)函数，main函数又会进入阻塞模式，等待线程函数<span style="color:rgb(79,79,79);line-height:28px;text-align:justify;">VENC_GetVencStreamProc结束。<span style="color:rgb(79,79,79);line-height:28px;text-align:justify;">线程函数</span><span style="color:rgb(79,79,79);line-height:28px;">VENC_GetVencStreamProc结束后，CPU又进入main函数执行<span style="color:rgb(79,79,79);line-height:28px;text-align:justify;">pthread_join(gs_VencPid, 0)后的代码，可以猜测后面的代码就是收尾工作，应该与前面的步骤动作相反，前面绑定通道，后面就要解绑，前面创建开启通道，后面就要停止销毁通道，前面初始化， 后面就要去初始化，前面按顺序设置MPI_SYS,VI,VPSS,VENC,后面就要反过来按顺序设置VENC,VPSS,VI,MPI_SYS,看代码的确如此：</span></span></span><p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;"></span></span></span></span></p><pre><code class="language-cpp hljs">stSrcChn.enModId = HI_ID_VPSS; stSrcChn.s32DevId = VpssGrp; stSrcChn.s32ChnId = VpssChn; stDestChn.enModId = HI_ID_VENC; stDestChn.s32DevId = <span class="hljs-number">0</span>; stDestChn.s32ChnId = VencChn; s32Ret = HI_MPI_SYS_UnBind(&stSrcChn, &stDestChn);<span class="hljs-comment">//先解绑VPSS和VENC的绑定，后面再解绑VPSS和VI的绑定</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VENC_StopRecvPic(VencChn);<span class="hljs-comment">//前面开启了接收图片，这里就要停止</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VENC_StopRecvPic vechn[%d] failed with %#x!\n"</span>,VencChn, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } <span class="hljs-comment">/****************************************** Distroy Venc Channel ******************************************/</span> s32Ret = HI_MPI_VENC_DestroyChn(VencChn); <span class="hljs-comment">//前面创建了VENC通道，这里就要销毁VENC通道，而且最先销毁，后面再销毁VI通道</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VENC_DestroyChn vechn[%d] failed with %#x!\n"</span>,VencChn, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } stSrcChn.enModId = HI_ID_VIU; stSrcChn.s32DevId = ViDev; stSrcChn.s32ChnId = ViChn; stDestChn.enModId = HI_ID_VPSS; stDestChn.s32DevId = VpssGrp; stDestChn.s32ChnId = <span class="hljs-number">0</span>; s32Ret = HI_MPI_SYS_UnBind(&stSrcChn, &stDestChn); <span class="hljs-comment">//后面再解绑VPSS和VI的绑定,在解绑VPSS和VENC的绑定之后，顺序与绑定顺序相反</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"failed with %#x!\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_DisableChn(VpssGrp, VpssChn);<span class="hljs-comment">//失能VPSS组和VPSS通道</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"%s failed with %#x\n"</span>, __FUNCTION__, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_StopGrp(VpssGrp);<span class="hljs-comment">//停止VPSS组</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"%s failed with %#x\n"</span>, __FUNCTION__, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VPSS_DestroyGrp(VpssGrp);<span class="hljs-comment">//销毁组</span> <span class="hljs-keyword">if</span> (s32Ret != HI_SUCCESS) { SAMPLE_PRT(<span class="hljs-string">"%s failed with %#x\n"</span>, __FUNCTION__, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VI_DisableChn(ViChn);<span class="hljs-comment">//失能通道</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VI_DisableChn failed with %#x\n"</span>,s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } s32Ret = HI_MPI_VI_DisableDev(ViDev);<span class="hljs-comment">//失能VI设备即摄像头</span> <span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret) { SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VI_DisableDev failed with %#x\n"</span>, s32Ret); <span class="hljs-keyword">return</span> HI_FAILURE; } SAMPLE_COMM_ISP_Stop();<span class="hljs-comment">//停止ISP</span> HI_MPI_SYS_Exit();<span class="hljs-comment">//去初始化</span> HI_MPI_VB_Exit(); <span class="hljs-comment">//由以上几个步骤可以看出，设置与销毁顺序相反，而且是成对出现</span> <span class="hljs-keyword">return</span> HI_SUCCESS;</code></pre>至此，H264_Venc函数分析完，现在我们来重点分析线程函数VENC_GetVencStreamProc是怎么获取码流并保存到一个文件里的。<p></p><p><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"><span style="color:rgb(79,79,79);line-height:28px;text-align:justify;"></span></span></p><pre><code class="language-cpp hljs"><span class="hljs-function">HI_VOID* <span class="hljs-title">VENC_GetVencStreamProc</span><span class="hljs-params">(HI_VOID *p)</span>
{
SAMPLE_VENC_GETSTREAM_PARA_S *pstPara;
HI_S32 s32ChnTotal;
VENC_CHN VencChn;
HI_S32 s32Ret;
PAYLOAD_TYPE_E enPayLoadType[VENC_MAX_CHN_NUM];
HI_CHAR aszFileName[VENC_MAX_CHN_NUM][64];
VENC_CHN_ATTR_S stVencChnAttr;
struct timeval TimeoutVal;
fd_set read_fds;
HI_S32 VencFd[VENC_MAX_CHN_NUM],maxfd;
VENC_STREAM_S stStream;
VENC_CHN_STAT_S stStat;
FILE *pFile[VENC_MAX_CHN_NUM];
HI_S32 i;
pstPara = (SAMPLE_VENC_GETSTREAM_PARA_S*)p;
s32ChnTotal = pstPara->s32Cnt;<span class="hljs-comment">//pstPara->s32Cnt是由参数传进来的，为1</span>
<span class="hljs-keyword">for</span> (i = <span class="hljs-number">0</span>; i < s32ChnTotal; i++)
{
VencChn = i;
s32Ret = HI_MPI_VENC_GetChnAttr(VencChn, &stVencChnAttr);
<span class="hljs-comment">//这里是为了取得是什么编码类型，以便确定保存文件的后缀名</span>
<span class="hljs-comment">//比如这里是H264编码，所以保存文件的后缀后就是.h264</span>
<span class="hljs-keyword">if</span>(s32Ret != HI_SUCCESS)
{
SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VENC_GetChnAttr chn[%d] failed with %#x!\n"</span>, \
VencChn, s32Ret);
<span class="hljs-keyword">return</span> <span class="hljs-literal">NULL</span>;
}
enPayLoadType[i] = stVencChnAttr.stVeAttr.enType;
gettime(filename);<span class="hljs-comment">//获取时间如20171101085639为文件名</span>
<span class="hljs-built_in">sprintf</span>(aszFileName[i], <span class="hljs-string">"%s_%d%s"</span>,filename, i, <span class="hljs-string">".h264"</span>);<span class="hljs-comment">//20171101085639.h264</span>
pFile[i] = fopen(aszFileName[i], <span class="hljs-string">"wb"</span>);
<span class="hljs-keyword">if</span> (!pFile[i])
{
SAMPLE_PRT(<span class="hljs-string">"open file[%s] failed!\n"</span>, aszFileName[i]);
<span class="hljs-keyword">return</span> <span class="hljs-literal">NULL</span>;
}
VencFd[i] = HI_MPI_VENC_GetFd(i);<span class="hljs-comment">//获取文件句柄，以便后面能用select来IO复用</span>
<span class="hljs-keyword">if</span> (VencFd[i] < <span class="hljs-number">0</span>)
{
SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VENC_GetFd failed with %#x!\n"</span>, VencFd[i]);
<span class="hljs-keyword">return</span> <span class="hljs-literal">NULL</span>;
}
<span class="hljs-keyword">if</span> (maxfd <= VencFd[i])
{
maxfd = VencFd[i];
}
}
<span class="hljs-keyword">while</span> (HI_TRUE == pstPara->bThreadStart)
<span class="hljs-comment">//当main函数所在的线程接收到两个键盘字符或ctrl+c时，pstPara->bThreadStart会为假，跳出while循环</span>
<span class="hljs-comment">//然后往下执行关闭前面打开的文件，执行完这个VENC_GetVencStreamProc线程函数，线程结束</span>
{
<span class="hljs-comment">/*IO复用4步骤
1.清空文件集合FD_ZERO(&read_fds);
2.将文件加入文件集合FD_SET(VencFd[i], &read_fds);
3.设置超时时间并用select(maxfd + 1, &read_fds, NULL, NULL, &TimeoutVal)来等待文件状态有变化唤醒线程
或超时唤醒
4.FD_ISSET查询文件状态是否有变化，有变化则处理
*/</span>
FD_ZERO(&read_fds);
<span class="hljs-keyword">for</span> (i = <span class="hljs-number">0</span>; i < s32ChnTotal; i++)
{
FD_SET(VencFd[i], &read_fds);
}
TimeoutVal.tv_sec = <span class="hljs-number">2</span>;
TimeoutVal.tv_usec = <span class="hljs-number">0</span>;
s32Ret = select(maxfd + <span class="hljs-number">1</span>, &read_fds, <span class="hljs-literal">NULL</span>, <span class="hljs-literal">NULL</span>, &TimeoutVal);
<span class="hljs-keyword">if</span> (s32Ret < <span class="hljs-number">0</span>)
{
SAMPLE_PRT(<span class="hljs-string">"select failed!\n"</span>);
<span class="hljs-keyword">break</span>;
}
<span class="hljs-keyword">else</span> <span class="hljs-keyword">if</span> (s32Ret == <span class="hljs-number">0</span>)
{
SAMPLE_PRT(<span class="hljs-string">"get venc stream time out, exit thread\n"</span>);
<span class="hljs-keyword">continue</span>;
}
<span class="hljs-keyword">else</span>
{
<span class="hljs-keyword">for</span> (i = <span class="hljs-number">0</span>; i < s32ChnTotal; i++)
{
<span class="hljs-keyword">if</span> (FD_ISSET(VencFd[i], &read_fds))
{
<span class="hljs-built_in">memset</span>(&stStream, <span class="hljs-number">0</span>, <span class="hljs-keyword">sizeof</span>(stStream));
s32Ret = HI_MPI_VENC_Query(i, &stStat);<span class="hljs-comment">//查询是否有码流，并将码流信息填充到stStat结构体中</span>
<span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret)
{
SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VENC_Query chn[%d] failed with %#x!\n"</span>, i, s32Ret);
<span class="hljs-keyword">break</span>;
}
<span class="hljs-keyword">if</span>(<span class="hljs-number">0</span> == stStat.u32CurPacks)
{
SAMPLE_PRT(<span class="hljs-string">"NOTE: Current frame is NULL!\n"</span>);
<span class="hljs-keyword">continue</span>;
}
stStream.pstPack = (VENC_PACK_S*)<span class="hljs-built_in">malloc</span>(<span class="hljs-keyword">sizeof</span>(VENC_PACK_S) * stStat.u32CurPacks);
<span class="hljs-comment">//分配内存以便保存码流包数据</span>
<span class="hljs-keyword">if</span> (<span class="hljs-literal">NULL</span> == stStream.pstPack)
{
SAMPLE_PRT(<span class="hljs-string">"malloc stream pack failed!\n"</span>);
<span class="hljs-keyword">break</span>;
}
stStream.u32PackCount = stStat.u32CurPacks;
<span class="hljs-comment">//printf("stStream.u32PackCount=%d\n",stStream.u32PackCount);</span>
s32Ret = HI_MPI_VENC_GetStream(i, &stStream, HI_TRUE);<span class="hljs-comment">//获取码流数据并保存到stStream结构体中</span>
<span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret)
{
<span class="hljs-built_in">free</span>(stStream.pstPack);<span class="hljs-comment">//获取失败则要释放前面分配的内存，否则会造成内存溢出</span>
stStream.pstPack = <span class="hljs-literal">NULL</span>;
SAMPLE_PRT(<span class="hljs-string">"HI_MPI_VENC_GetStream failed with %#x!\n"</span>, s32Ret);
<span class="hljs-keyword">break</span>;
}
HI_S32 u32PackIndex;
<span class="hljs-keyword">for</span> (u32PackIndex= <span class="hljs-number">0</span>;u32PackIndex < stStream.u32PackCount; u32PackIndex++)
{
fwrite( stStream.pstPack[u32PackIndex].pu8Addr+stStream.pstPack[u32PackIndex].u32Offset,\
stStream.pstPack[u32PackIndex].u32Len- stStream.pstPack[u32PackIndex].u32Offset, \
<span class="hljs-number">1</span>, pFile[i]);
fflush(pFile[i]);
<span class="hljs-meta">#<span class="hljs-meta-keyword">if</span> 1</span>
<span class="hljs-built_in">printf</span>(<span class="hljs-string">"stStream.u32PackCount=%d,stStream.pstPack[%d].pu8Addr=0x%08x,\
stStream.pstPack[%d].u32Offset=%d,stStream.pstPack[%d].u32Len=%d\n"</span>,\
stStream.u32PackCount,u32PackIndex,stStream.pstPack[u32PackIndex].pu8Addr,\
u32PackIndex,stStream.pstPack[u32PackIndex].u32Offset,u32PackIndex,stStream.pstPack[u32PackIndex].u32Len);
<span class="hljs-comment">//添加打印信息，查看保存码流内容的内存是怎么样的</span>
<span class="hljs-meta">#<span class="hljs-meta-keyword">endif</span></span>
}
s32Ret = HI_MPI_VENC_ReleaseStream(i, &stStream);<span class="hljs-comment">//保存后要释放码流</span>
<span class="hljs-keyword">if</span> (HI_SUCCESS != s32Ret)
{
<span class="hljs-built_in">free</span>(stStream.pstPack);<span class="hljs-comment">//获取失败则要释放前面分配的内存，否则会造成内存溢出</span>
stStream.pstPack = <span class="hljs-literal">NULL</span>;
<span class="hljs-keyword">break</span>;
}
<span class="hljs-built_in">free</span>(stStream.pstPack);<span class="hljs-comment">//释放码流后，也要释放分配的内存，避免内存溢出</span>
stStream.pstPack = <span class="hljs-literal">NULL</span>;
}
}
}
}
<span class="hljs-keyword">for</span> (i = <span class="hljs-number">0</span>; i < s32ChnTotal; i++)
{
fclose(pFile[i]);
}
<span class="hljs-keyword">return</span> <span class="hljs-literal">NULL</span>;
}线程函数里先通过HI_MPI_VENC_GetFd取得文件句柄，然后在while循环里不断清空读文件集合FD_ZERO(&read_fds);将文件句柄加入读文件集合FD_SET(VencFd[i], &read_fds);用select来休眠线程，有数据或超时又唤醒线程select(maxfd + 1, &read_fds, NULL, NULL, &TimeoutVal);用FD_ISSET(VencFd[i], &read_fds)来查询是哪个文件句柄有数据以便处理，然后用HI_MPI_VENC_Query查询码流统计信息以便分配内存，HI_MPI_VENC_GetStream来取得码流包数据，fwrite将码流包数据保存到文件中，码流包处理后要HI_MPI_VENC_ReleaseStream释放掉码流，释放掉申请的内存。
综合整个程序，在创建线程函数前都是对系统设置，按照海思SDK里的文档步骤设置就行，基本上就是定义结构体，给结构体赋值，然后设置进系统。到后面创建线程后就有两个线程函数了，一个main函数所在线程，一个VENC_GetVencStreamProc函数所在线程。main函数所在线程创建VENC_GetVencStreamProc线程后就调用getchar();函数进入休眠状态，只有当键盘有输入时getchar()函数有返回值时才会唤醒main函数所在线程，所以CPU就切换到VENC_GetVencStreamProc线程，VENC_GetVencStreamProc线程在调用select函数时，如果读文件集合read_fds有要读的内容则会往下执行，如果没有VENC_GetVencStreamProc线程也会休眠，CPU会切换到别的线程，当read_fds有要读的内容内容时或超时VENC_GetVencStreamProc线程会唤醒，往下执行去进行读码流的处理过程。
程序写完后，在sample文件夹里创建个子文件夹如myvenc,然后将这个文件保存到myvenc里，从sample\venc里复制makefile到sample\myvenc里，这个makefile都不要我们改些什么东西，只要位置放对了位置(跟源代码文件在同一个目录里)就直接可以用，然后make clean,make编译，如果没问题的话会生成myvenc可执行文件，将这个可执行文件拷贝到HI3518开发板(可通过SD卡从PC拷贝到开发板，也可挂载根文件系统直接复制过去，方法多样)，然后执行./myvenc就会运行程序，过一会就会在myvenc同级目录下生成一个以时间为名的，后缀名为h264的文件，这个文件就是H264码流保存的文件，将这文件从Hi3518里复制到PC机上，用VLC播放器就可以播放出摄像头拍摄的内容。
编译运行程序后生成了h264文件，这个文件及程序运行时打印的输出内容有助于我们分析h264,下一篇我们将分析h264以便后期怎么将h264数据打包到rtp中进行rtsp传送数据。
这个程序比较简单，只是为了让大家了解海思芯片是怎么进行多媒体开发，所以很多功能没有写成函数，感觉内容很多，但避免了sample里复杂的函数调用，有助于将了解各个数据结构及海思SDK各个函数的用法。程序从前往后看，流程一目了然，各个数据结构、函数对应哪个模块很清楚，对入门很有好处。我将代码上传到github,欢迎朋友们下载，一起讨论学习。
最后看一下程序运行效果，用VLC打开H264文件，可以看到摄像头拍摄界面：
项目源代码及工具:Hi3518编码H264
标签：3518,s32Ret,H264,MPI,SAMPLE,HI,VENC,VPSS,海思 来源： https://blog.csdn.net/weixin_44893022/article/details/94604950