Android RTMP 投屏直播推流实现

2021-03-26 本文已影响0人蒋斌文

简介

观看手游直播时，我们观众端看到的是选手的屏幕上的内容，这是如何实现的呢？这篇博客将手写一个录屏直播Demo，实现类似手游投屏直播的效果

获取屏幕数据很简单，Android 系统有提供对应的服务，难点在于传输数据到直播服务器，我们使用 RtmpDump 来传输 Rtmp 数据，由于 RtmpDump 使用 C 语言实现，我们还需要用到 NDK 开发，单单用 Java 无法实现哈，当年没有使用开发实现的RTMP协议的开源库，所以在RTMP协议包上，使用了RtmpDump C库。

实现效果

基本流程

获取录屏数据
对数据进行 h264 编码
Rtmp 数据包
上传到直播服务器推流地址

RTMP直播实现流程

获取录屏数据

MediaProjection 视频采集 SDK中的接口

A token granting applications the ability to capture screen contents and/or record system audio. The exact capabilities granted depend on the type of MediaProjection.

A screen capture session can be started through MediaProjectionManager.createScreenCaptureIntent(). This grants the ability to capture screen contents, but not system audio.

@Override
    protected void onActivityResult(int requestCode, int resultCode, Intent data) {
        super.onActivityResult(requestCode, resultCode, data);
        if (requestCode == 100 && resultCode == Activity.RESULT_OK) {
            if (editText.getText()!=null&&!TextUtils.isEmpty(editText.getText().toString())) {
                url=editText.getText().toString();
                Log.i("tuch", "url: "+url);
            }
            Log.i(TAG, " url:"+url);
            mediaProjection = mediaProjectionManager.getMediaProjection(resultCode, data);
        }
    }

    public void startLive(View view) {
        this.mediaProjectionManager = (MediaProjectionManager)getSystemService(Context.MEDIA_PROJECTION_SERVICE);
        Intent captureIntent = mediaProjectionManager.createScreenCaptureIntent();
        startActivityForResult(captureIntent, 100);
    }

mediaProjectionManager.createScreenCaptureIntent

获取录屏数据

通过 Intent 获取到 MediaProjectionService，继而获取到 Mediaprojection 的 VirtualCanvas，我们录屏的原始数据就是从中得来的

public VirtualDisplay createVirtualDisplay (String name,                 int width,                 int height,                 int dpi,                 int flags,                 Surface surface,                 VirtualDisplay.Callback callback,                 Handler handler)

Creates a VirtualDisplay to capture the contents of the screen.

createVirtualDisplay API

mediaProjection--->产生录屏数据

对数据进行 h264 编码

通过 MediaProjection 获取到的 YUV 裸数据，我们先需要对其进行 h264 编码，此时我们使用原生 MediaCodec 进行硬编码,本例子是Demo工程，没有考虑编码兼容性的问题，直接使用 MediaCodec。MediaCodec的使用参考 MediaCodec 介绍

public void startLive(MediaProjection mediaProjection) {
        this.mediaProjection = mediaProjection;

        MediaFormat format = MediaFormat.createVideoFormat(MediaFormat.MIMETYPE_VIDEO_AVC,
                720,
                1280);
        format.setInteger(MediaFormat.KEY_COLOR_FORMAT,
                MediaCodecInfo.CodecCapabilities.COLOR_FormatSurface);
        //码率，帧率，分辨率，关键帧间隔
        format.setInteger(MediaFormat.KEY_BIT_RATE, 400_000);
        format.setInteger(MediaFormat.KEY_FRAME_RATE, 15);
        format.setInteger(MediaFormat.KEY_I_FRAME_INTERVAL, 1);
        try {
            mediaCodec = MediaCodec.createEncoderByType(MediaFormat.MIMETYPE_VIDEO_AVC);//手机
            mediaCodec.configure(format, null, null,
                    MediaCodec.CONFIGURE_FLAG_ENCODE);
            Surface surface = mediaCodec.createInputSurface();

            virtualDisplay = mediaProjection.createVirtualDisplay(
                    "screen-codec",
                    720, 1280, 1,
                    DisplayManager.VIRTUAL_DISPLAY_FLAG_PUBLIC,
                    surface, null, null);

        } catch (IOException e) {
            e.printStackTrace();
        }
        LiveTaskManager.getInstance().execute(this);
    }

Surface surface = mediaCodec.createInputSurface();从编码器创建一个画布, 画布上的图像会被编码器自动编码, 调用createVirtualDisplay创建虚拟显示器VirtualDisplay ，即会将手机屏幕镜像到虚拟显示器上。在createVirtualDisplay时，需要传递一个Surface(画布)。需要获取图像数据即可从这个Surface中读取。

VirtualDisplay-Surface-MediaCodec

配置完成后，从MediaCodec中获取数据

    @Override
    public void run() {
        isLiving = true;
        mediaCodec.start();
        MediaCodec.BufferInfo bufferInfo = new MediaCodec.BufferInfo();

        while (isLiving){
            //若时间差大于 2 s，则通知编码器，生成 I 帧
            if (System.currentTimeMillis() - timeStamp >= 2000){
                // Bundle 通知 Dsp
                Bundle msgBundle = new Bundle();
                msgBundle.putInt(MediaCodec.PARAMETER_KEY_REQUEST_SYNC_FRAME,0);
                mediaCodec.setParameters(msgBundle);
                timeStamp = System.currentTimeMillis();
            }
            // 接下来就是 MediaCodec 常规操作，获取 Buffer 可用索引，这里不需要获取输出索引，内部已经操作了
            int outputBufferIndex = mediaCodec.dequeueOutputBuffer(bufferInfo,100_000);
            if (outputBufferIndex >=0){
                // 获取到了
                ByteBuffer byteBuffer = mediaCodec.getOutputBuffer(outputBufferIndex);
                byte[] outData = new byte[bufferInfo.size];
                byteBuffer.get(outData);
            }
        }

VideoCodec线程的润方法中，不断的判int outputBufferIndex = mediaCodec.dequeueOutputBuffer(bufferInfo,100_000);是否有编码完成的H264数据，为什么是h264编码的，在配置的时候，MediaFormat.MIMETYPE_VIDEO_AVC参数定义的！

此时可以把好的数据，以byte的方式写入文件查看

image-20210326175405628

可以看到sps pps I帧等关键信息

也可以通过ffplay命令进行播放

ffplay -i codec.h264

此时，我们获得了编码好的 h264 数据，接下来封装 Rtmp数据包。

LIBRTMP

C语言开源RTMP库，封装 Socket 建立TCP通信，并实现了RTMP数据的收发。

RTMPDump

rtmpdump is a toolkit for RTMP streams. All forms of RTMP are supported, including rtmp://, rtmpt://, rtmpe://, rtmpte://, and rtmps://.

License: GPLv2
Copyright (C) 2009 Andrej Stepanchuk
Copyright (C) 2010-2011 Howard Chu

Download the source:

git clone git://git.ffmpeg.org/rtmpdump

The latest release is 2.4 which you can check out from git. Aside from various minor bugfixes since 2.3, RTMPE type 9 handshakes are now supported.

使用第三方库 Rtmpdump 来实现推流到直播服务器，由于 Rtmpdump 的代码量不是很多，我们直接拷贝源代码到 Android 的 cpp 文件

#定义宏  如果代码中定义了 #defind NO_CRYPTO
#就表示不适用ssl，不支持rtmps。我们这里不支持ssl
set(CMAKE_C_FLAGS "${CMAKE_C_FLAGS} -DNO_CRYPTO")

# 把当前目录下所有得文件 变成一个 SOURCE变量表示
aux_source_directory(. SOURCE)
# 编译成librtmp.a 静态库，编译源文件引用${SOURCE}获取
add_library(rtmp STATIC ${SOURCE})

工程cmake引入编译好的rtmp静态库

# 加入子文件夹
add_subdirectory(librtmp)
# 链接引入
target_link_libraries( # Specifies the target library.
                       native-lib
                       # Links the target library to the log library
                       # included in the NDK.
                       ${log-lib}
                         rtmp)

RtmpDump 的使用

RtmpDump

连接服务器

RTMP_Init(RTMP *r) 初始化
RTMP_EnableWrite(RTMP *r) 配置开启数据写入
RTMP_Connect(RTMP *r, RTMPPacket *cp)
RTMP_ConnectStream(RTMP *r, int seekTime)

发送数据

RTMPPacket_Alloc(RTMPPacket *p, int nSize)
RTMP_SendPacket(RTMP *r, RTMPPacket *packet, int queue)
RTMPPacket_Free(RTMPPacket *p)

Rtmp 协议关键帧协议格式

image

Rtmp 协议非关键帧协议格式

image-20210326195045243

SPS PPS数据包

sps-pps

启动配置好的SRS推流服务器

SRS推流服务器

./objs/srs -c conf/rtmp.conf
lsof -i :1935

连接直播服务器

这一步中，需要预先准备直播推流地址，然后实现 native 方法

extern "C"
JNIEXPORT jboolean JNICALL
Java_com_lecture_rtmtscreenlive_ScreenLive_connect(JNIEnv *env, jobject thiz, jstring url_) {

    // 首先 Java 的转成 C 的字符串，不然无法使用
    const char *url = env->GetStringUTFChars(url_, 0);
    int ret;
    do {
        live = (Live *) malloc(sizeof(Live));
        memset(live, 0, sizeof(Live));
        live->rtmp = RTMP_Alloc();// Rtmp 申请内存
        RTMP_Init(live->rtmp);
        live->rtmp->Link.timeout = 10;// 设置 rtmp 初始化参数，比如超时时间、url
        LOGI("connect %s", url);
        if (!(ret = RTMP_SetupURL(live->rtmp, (char *) url))) break;
        RTMP_EnableWrite(live->rtmp);// 开启 Rtmp 写入
        LOGI("RTMP_Connect");
        if (!(ret = RTMP_Connect(live->rtmp, 0))) break;
        LOGI("RTMP_ConnectStream ");
        if (!(ret = RTMP_ConnectStream(live->rtmp, 0))) break;
        LOGI("connect success");
    } while (0);
    if (!ret && live) {
        free(live);
        live = nullptr;
    }

    env->ReleaseStringUTFChars(url_, url);
    return ret;

}

2021-03-26 20:25:33.202 9139-9211/com.lecture.rtmtscreenlive I/DDDDDD: connect rtmp://192.168.10.224/live/livestream
2021-03-26 20:25:33.202 9139-9211/com.lecture.rtmtscreenlive I/DDDDDD: RTMP_Connect
2021-03-26 20:25:33.422 9139-9211/com.lecture.rtmtscreenlive I/DDDDDD: RTMP_ConnectStream 
2021-03-26 20:25:33.562 9139-9211/com.lecture.rtmtscreenlive I/DDDDDD: connect success
2021-03-26 20:25:34.152 9139-9222/com.lecture.rtmtscreenlive I/------>dddd<---------: run: -2
2021-03-26 20:25:34.152 9139-9222/com.lecture.rtmtscreenlive I/------>dddd<---------: run: 0

服务器连接成功，现在开始发送RTMP视频数据

视频数据封包

RTMPPacket *createVideoPackage(int8_t *buf, int len, const long tms, Live *live) {
//    分隔符被抛弃了      --buf指的是651
    buf += 4;
    len -= 4;
    int body_size = len + 9;
    RTMPPacket *packet = (RTMPPacket *) malloc(sizeof(RTMPPacket));
    RTMPPacket_Alloc(packet, len + 9);

    packet->m_body[0] = 0x27;
    if (buf[0] == 0x65) { //关键帧
        packet->m_body[0] = 0x17;
        LOGI("发送关键帧 data");
    }
    packet->m_body[1] = 0x01;
    packet->m_body[2] = 0x00;
    packet->m_body[3] = 0x00;
    packet->m_body[4] = 0x00;
    //长度
    packet->m_body[5] = (len >> 24) & 0xff;
    packet->m_body[6] = (len >> 16) & 0xff;
    packet->m_body[7] = (len >> 8) & 0xff;
    packet->m_body[8] = (len) & 0xff;

    //数据
    memcpy(&packet->m_body[9], buf, len);


    packet->m_packetType = RTMP_PACKET_TYPE_VIDEO;
    packet->m_nBodySize = body_size;
    packet->m_nChannel = 0x04;
    packet->m_nTimeStamp = tms;
    packet->m_hasAbsTimestamp = 0;
    packet->m_headerType = RTMP_PACKET_SIZE_LARGE;
    packet->m_nInfoField2 = live->rtmp->m_stream_id;
    return packet;

}

Sps pps 封包

RTMPPacket *createVideoPackage(Live *live) {
    int body_size = 13 + live->sps_len + 3 + live->pps_len;
    RTMPPacket *packet = (RTMPPacket *) malloc(sizeof(RTMPPacket));
    RTMPPacket_Alloc(packet, body_size);
    int i = 0;
    //AVC sequence header 与IDR一样
    packet->m_body[i++] = 0x17;
    //AVC sequence header 设置为0x00
    packet->m_body[i++] = 0x00;
    //CompositionTime
    packet->m_body[i++] = 0x00;
    packet->m_body[i++] = 0x00;
    packet->m_body[i++] = 0x00;
    //AVC sequence header
    packet->m_body[i++] = 0x01;   //configurationVersion 版本号 1
    packet->m_body[i++] = live->sps[1]; //profile 如baseline、main、 high

    packet->m_body[i++] = live->sps[2]; //profile_compatibility 兼容性
    packet->m_body[i++] = live->sps[3]; //profile level
    packet->m_body[i++] = 0xFF; // reserved（111111） + lengthSizeMinusOne（2位 nal 长度） 总是0xff
    //sps
    packet->m_body[i++] = 0xE1; //reserved（111） + lengthSizeMinusOne（5位 sps 个数） 总是0xe1
    //sps length 2字节
    packet->m_body[i++] = (live->sps_len >> 8) & 0xff; //第0个字节
    packet->m_body[i++] = live->sps_len & 0xff;        //第1个字节
    memcpy(&packet->m_body[i], live->sps, live->sps_len);
    i += live->sps_len;

    /*pps*/
    packet->m_body[i++] = 0x01; //pps number
    //pps length
    packet->m_body[i++] = (live->pps_len >> 8) & 0xff;
    packet->m_body[i++] = live->pps_len & 0xff;
    memcpy(&packet->m_body[i], live->pps, live->pps_len);

    packet->m_packetType = RTMP_PACKET_TYPE_VIDEO;
    packet->m_nBodySize = body_size;
    packet->m_nChannel = 0x04;
    packet->m_nTimeStamp = 0;
    packet->m_hasAbsTimestamp = 0;
    packet->m_headerType = RTMP_PACKET_SIZE_LARGE;
    packet->m_nInfoField2 = live->rtmp->m_stream_id;
    return packet;
}

int sendPacket(RTMPPacket *packet) {
    int r = RTMP_SendPacket(live->rtmp, packet, 1);
    RTMPPacket_Free(packet);
    free(packet);
    return r;
}

void prepareVideo(int8_t *data, int len, Live *live) {
    for (int i = 0; i < len; i++) {
        //0x00 0x00 0x00 0x01
        if (i + 4 < len) {
            if (data[i] == 0x00 && data[i + 1] == 0x00
                && data[i + 2] == 0x00
                && data[i + 3] == 0x01) {
                //0x00 0x00 0x00 0x01 7 sps 0x00 0x00 0x00 0x01 8 pps
                //将sps pps分开
                //找到pps
                if (data[i + 4] == 0x68) {
                    //去掉界定符
                    live->sps_len = i - 4;
                    live->sps = static_cast<int8_t *>(malloc(live->sps_len));
                    memcpy(live->sps, data + 4, live->sps_len);

                    live->pps_len = len - (4 + live->sps_len) - 4;
                    live->pps = static_cast<int8_t *>(malloc(live->pps_len));
                    memcpy(live->pps, data + 4 + live->sps_len + 4, live->pps_len);
                    LOGI("sps:%d pps:%d", live->sps_len, live->pps_len);
                    break;
                }
            }
        }
    }
}

测试

ffplay rtmp://192.168.10.224/live/livestream

image-20210326211259263

获取音频数据

AudioRecord 采集

//录音工具类  采样位数 通道数   采样评率   固定了   设备没关系  录音 数据一样的
            minBufferSize = AudioRecord.getMinBufferSize(44100,
                    AudioFormat.CHANNEL_IN_MONO,
                    AudioFormat.ENCODING_PCM_16BIT);
            audioRecord = new AudioRecord(
                    MediaRecorder.AudioSource.MIC, 44100,
                    AudioFormat.CHANNEL_IN_MONO,
                    AudioFormat.ENCODING_PCM_16BIT, minBufferSize);

使用MediaCodec对采集的PCM数据编码

 MediaFormat format = MediaFormat.createAudioFormat(MediaFormat.MIMETYPE_AUDIO_AAC, 44100, 1);
        //录音质量
        format.setInteger(MediaFormat.KEY_AAC_PROFILE, MediaCodecInfo.CodecProfileLevel
                .AACObjectLC);
        //一秒的码率 aac
        format.setInteger(MediaFormat.KEY_BIT_RATE, 64_000);

mediaCodec = MediaCodec.createEncoderByType(MediaFormat.MIMETYPE_AUDIO_AAC);
            mediaCodec.configure(format, null, null, MediaCodec.CONFIGURE_FLAG_ENCODE);
            mediaCodec.start();

麦克风的数据读取出来 pcm

 audioRecord.startRecording();
//        容器 固定
        byte[] buffer = new byte[minBufferSize];
//            麦克风的数据读取出来   pcm   buffer  aac
            int len = audioRecord.read(buffer, 0, buffer.length);

获取经过mediaCodec编码的aac数据

int index = mediaCodec.dequeueInputBuffer(0);
if (index >= 0) {
    ByteBuffer inputBuffer = mediaCodec.getInputBuffer(index);
    inputBuffer.clear();
    inputBuffer.put(buffer, 0, len);
    //填充数据后再加入队列
    mediaCodec.queueInputBuffer(index, 0, len,
            System.nanoTime() / 1000, 0);
}

RTMP 包中封装的音视频数据流，其实和FLV/tag封装音频和视频数据的方式是相同的，所以我们只需要按照FLV格式封装音视频即可。

image-20210326215407401

RTMPPacket *createAudioPacket(int8_t *buf, const int len, const int type, const long tms,
                              Live *live) {

//    组装音频包  两个字节    是固定的   af    如果是第一次发  你就是 01       如果后面   00  或者是 01  aac
    int body_size = len + 2;
    RTMPPacket *packet = (RTMPPacket *) malloc(sizeof(RTMPPacket));
    RTMPPacket_Alloc(packet, body_size);
//         音频头
    packet->m_body[0] = 0xAF;
    if (type == 1) {
//        头
        packet->m_body[1] = 0x00;
    }else{
        packet->m_body[1] = 0x01;
    }
    memcpy(&packet->m_body[2], buf, len);
    packet->m_packetType = RTMP_PACKET_TYPE_AUDIO;
    packet->m_nChannel = 0x05;
    packet->m_nBodySize = body_size;
    packet->m_nTimeStamp = tms;
    packet->m_hasAbsTimestamp = 0;
    packet->m_headerType = RTMP_PACKET_SIZE_LARGE;
    packet->m_nInfoField2 = live->rtmp->m_stream_id;
    return packet;
}

代码GitHub