OpenGL ES_手把手教你打造VR全景播放器

酷走天涯

发布于 2018-09-14 14:38:09

2K0

发布于 2018-09-14 14:38:09

文章被收录于专栏：每日一篇技术文章

实战2中，详细介绍了多屏显示的原理和实现过程,今天我们继续我们的OpenGL 旅程！技术再牛逼也要学习!

学习是一件开心的额事情

学习目标

打造全景视频，以及VR 眼镜专用的双屏显示框架!

你应该知道的

网络截图

全景显示的原理通俗的将，好比红色区域就是你的手机屏幕，当你旋转手机的时候，我们球体向相反的方向旋转，这样，你就可以看到球体上的画面了.

准备工作

找一个全景视频，添加到项目中去。

实现步骤 1.创建一个球体模型 2.获取视频数据的每一帧数据转换成RGB 格式，渲染到球体上 3.通过手势的变换，改变球体模型视图矩阵值 4.如果是VR模式,则通过角度传感器获取用户的行为，调整视图矩阵。

实现了那些功能

支持普通视频播放
支持全景视频播放
支持VR 双屏显示模式
支持快进，快退
支持播放，暂停
支持暂停广告功能

核心代码讲解

如果你想要和我一样,能够从零开始把代码敲出来，请确保自己有OpenGL ES 2.0 的基础知识和 GLSL 的简单基本知识，如果你不具备这方面的知识，没关系，我已经写好了OpenGL学习教程和GLSL教程,请移步开始学习。下面开始我们的内容讲解.

视频采集

<p>工程中的两个文件 XJVRPlayerViewController.h和XJVRPlayerController.m主要负责视频数据采集,界面布局在XJVRPlayerViewController中可以更改,主要使用AVFoundation框架这部分内容今天咱不讲解,后面我会写关于视频采集的教程</p>

模型创建 a.全景播放器生成球体的顶点坐标和纹理坐标 b.普通播放器生成长方形的顶点坐标和纹理坐标两个生成函数在OSShere.h中
将数据加载到GPU中去 // 加载顶点索引数据 glGenBuffers(1, &_indexBuffer); glBindBuffer(GL_ELEMENT_ARRAY_BUFFER, _indexBuffer); glBufferData(GL_ELEMENT_ARRAY_BUFFER, _numIndices*sizeof(GLushort), _indices, GL_STATIC_DRAW); // 加载顶点坐标 glGenBuffers(1, &_vertexBuffer); glBindBuffer(GL_ARRAY_BUFFER, _vertexBuffer); glBufferData(GL_ARRAY_BUFFER, numVertices*strideNum*sizeof(GLfloat), _vertices, GL_STATIC_DRAW); glEnableVertexAttribArray(GLKVertexAttribPosition); glVertexAttribPointer(GLKVertexAttribPosition, strideNum, GL_FLOAT, GL_FALSE, strideNum*sizeof(GLfloat), NULL); //加载纹理坐标 glGenBuffers(1, &_textureCoordBuffer); glBindBuffer(GL_ARRAY_BUFFER, _textureCoordBuffer); glBufferData(GL_ARRAY_BUFFER, sizeof(GLfloat)*2*numVertices, _texCoords, GL_DYNAMIC_DRAW); glEnableVertexAttribArray(GLKVertexAttribTexCoord0); glVertexAttribPointer(GLKVertexAttribTexCoord0, 2, GL_FLOAT, GL_FALSE, 2*sizeof(GLfloat), NULL);

以上函数的具体用法，在之前的教程中都讲过,这里就不赘述了。

着色器程序我把着色器分为两种类型，一种是渲染全景视频的，一种是渲染普通视频的，两个没有多大区别，只是在全景着色器中添加了一个视图转换矩阵 (全景着色器:ShadePanorama,普通着色器:ShaderNormal) 下面给出的是全景的着色器的代码: a.顶点着色器 attribute vec4 position; // 顶点坐标属性 attribute vec2 texCoord0;// 纹理坐标 varying vec2 texCoordVarying;// 片段着色器输入变量,负责获取纹理坐标的值 uniform mat4 modelViewProjectionMatrix;//视图变换矩阵 void main (){ texCoordVarying = texCoord0; gl_Position = modelViewProjectionMatrix*position; }

b.片段着色器

  precision mediump float;//设置float精度
  varying  vec2 texCoordVarying;
  uniform sampler2D sam2DY;  // 纹理采样器Y
  uniform sampler2D sam2DUV;// 纹理采样器UV
  void main(){
  mediump vec3 yuv;
  lowp vec3 rob;
  // YUV  转RGB 的转换矩阵 
  mediump mat3 convert = mat3(1.164,  1.164, 1.164,
                            0.0, -0.213, 2.112,
                            1.793, -0.533,   0.0);
  yuv.x = texture2D(sam2DY,texCoordVarying).r - (16.0/255.0);
  yuv.yz = texture2D(sam2DUV,texCoordVarying).rg - vec2(0.5, 0.5);
  rgb = convert*yuv;
  gl_FragColor = vec4(rgb,1);
  }

如果想要了解更多关于着色器语言的知识，请猛戳我

创建着色器程序创建着色器程序的目的是编译刚才我们编写好的着色器源代码，以及将着色器的变量和我们的应用程序代码相关联 /** * 创建编译shader程序 * * @param vshName 顶点着色器文件名称 * @param fshName 片段着色器文件名称 */ -(void)createShaderProgramVertexShaderName: (NSString*)vshName FragmentShaderName: (NSString*)fshName{ self.shaderManager = [OSShaderManager allocinit]; // 编译连个shader 文件 GLuint vertexShader,fragmentShader; NSURL *vertexShaderPath = [NSBundle mainBundleURLForResource:vshName withExtension:@"vsh"]; NSURL *fragmentShaderPath = [NSBundle mainBundleURLForResource:fshName withExtension:@"fsh"]; if (undefined||! self.shaderManager compileShader:&fragmentShader type:GL_FRAGMENT_SHADER URL:fragmentShaderPath){ return ; } // 注意获取绑定属性要在连接程序之前 location 随便你写,如果你随便写请记住他，后面要用到 self.shaderManager bindAttribLocation:GLKVertexAttribPosition andAttribName:"position"; self.shaderManager bindAttribLocation:GLKVertexAttribTexCoord0 andAttribName:"texCoord0"; // 将编译好的两个对象和着色器程序进行连接 if(undefined){ self.shaderManager deleteShader:&vertexShader; self.shaderManager deleteShader:&fragmentShader; } _textureBufferY = self.shaderManager getUniformLocation:"sam2DY"; _textureBufferUV = self.shaderManager getUniformLocation:"sam2DUV"; _modelViewProjectionMatrixIndex = self.shaderManager getUniformLocation:"modelViewProjectionMatrix"; self.shaderManager detachAndDeleteShader:&vertexShader; self.shaderManager detachAndDeleteShader:&fragmentShader; // 启用着色器 self.shaderManager useProgram; }

// 上面的OSShaderManager 这个类，我把着色器程序编译链接的一些方法简单的封装了一下,具体的方向看下面

    /**
     *  编译shader程序
     *  @param shader shader名称
     *  @param type   shader 类型
     *  @param URL    shader 本地路径
     *  @return 是否编译成功
   */
    - (BOOL)compileShader:(GLuint *)shader type:(GLenum)type URL:(NSURL *)URL;
   /**
   *  连接程序
   *  @return 连接程序是否成功
   */
  - (BOOL)linkProgram;
  /**
   *  验证程序是否成功
   *  @param prog 程序标示
   *  @return 返回是否成功标志
   */
  - (BOOL)validateProgram;
  /**
   *  绑定着色器的属性
   *  @param index 属性在shader 程序的索引位置
   *  @param name  属性名称
   */
  - (void)bindAttribLocation:(GLuint)index andAttribName:  (GLchar*)name;
  /**
   *  删除shader
   */
  - (void)deleteShader:(GLuint*)shader;
  /**
   *  获取属性值索引位置
   *  @param name 属性名称
   *  @return 返回索引位置
   */
  - (GLint)getUniformLocation:(const GLchar*) name;
  /**
   * 释放, 删除shader
   *  @param shader 着色器名称
   */
  -(void)detachAndDeleteShader:(GLuint*)shader;
  /**
   *  使用程序
   */
  -(void)useProgram;

方法的具体实现请阅读工程文件

纹理采样器指向 glUniform1i(_textureBufferY, 0); // 0 代表GL_TEXTURE0 glUniform1i(_textureBufferUV, 1); // 1 代表GL_TEXTURE1

在这里我有必要提醒你，这两个方法，一定要放在着色器程序链接成功之后，不然你调用这个两个方法，没有效果。

如何将YUV 数据分离，并且加载到两个着色器中去, 这里我们又要用到之前我们使用过的框架了CoreVideo. 干涉么的呢，专门处理我们的像素数据的。我们从视频采集到的视频是CVPixelBufferRef 类型的下面我们先看一下我们像素数据的格式 <CVPixelBuffer 0x7fa27962c9c0 width=2048 height=1024 pixelFormat=420v iosurface=0x0 planes=2> <Plane 0 width=2048 height=1024 bytesPerRow=2048> <Plane 1 width=1024 height=512 bytesPerRow=2048> <attributes=<CFBasicHash 0x7fa279623910 0x10296ba40>{type = immutable dict, count = 4, entries => 1 : <CFString 0x102d183b8 0x10296ba40>{contents = "PixelFormatType"} = <CFArray 0x7fa27c414bc0 0x10296ba40>{type = mutable-small, count = 1, values = ( 0 : <CFNumber 0xb000000343230763 0x10296ba40>{value = +875704438, type = kCFNumberSInt64Type} )} 2 : <CFString 0x102d17e78 0x10296ba40>{contents = "Height"} = <CFNumber 0xb000000000004002 0x10296ba40>{value = +1024, type = kCFNumberSInt32Type} 5 : <CFString 0x102d17d38 0x10296ba40>{contents = "PropagatedAttachments"} = <CFBasicHash 0x7fa27c51c590 0x10296ba40>{type = mutable dict, count = 4, entries => 0 : <CFString 0x102d18058 0x10296ba40>{contents = "CVImageBufferYCbCrMatrix"} = <CFString 0x102d18098 0x10296ba40>{contents = "ITU_R_601_4"} 1 : <CFString 0x102d181b8 0x10296ba40>{contents = "CVImageBufferTransferFunction"} = <CFString 0x102d18078 0x10296ba40>{contents = "ITU_R_709_2"} 2 : <CFString 0x106eadc88 0x10296ba40>{contents = "ColorInfoGuessedBy"} = <CFString 0x106eadca8 0x10296ba40>{contents = "VideoToolbox"} 5 : <CFString 0x102d18138 0x10296ba40>{contents = "CVImageBufferColorPrimaries"} = <CFString 0x102d18178 0x10296ba40>{contents = "SMPTE_C"} } 6 : <CFString 0x102d17e58 0x10296ba40>{contents = "Width"} = <CFNumber 0xb000000000008002 0x10296ba40>{value = +2048, type = kCFNumberSInt32Type} } propagatedAttachments=<CFBasicHash 0x7fa27962caa0 0x10296ba40>{type = mutable dict, count = 10, entries => 0 : <CFString 0x106eadc88 0x10296ba40>{contents = "ColorInfoGuessedBy"} = <CFString 0x106eadca8 0x10296ba40>{contents = "VideoToolbox"} 1 : <CFString 0x102d18058 0x10296ba40>{contents = "CVImageBufferYCbCrMatrix"} = <CFString 0x102d18098 0x10296ba40>{contents = "ITU_R_601_4"} 2 : <CFString 0x102d17ed8 0x10296ba40>{contents = "CVFieldCount"} = <CFNumber 0xb000000000000012 0x10296ba40>{value = +1, type = kCFNumberSInt32Type} 3 : <CFString 0x102d17f98 0x10296ba40>{contents = "CVPixelAspectRatio"} = <CFBasicHash 0x7fa279728c10 0x10296ba40>{type = immutable dict, count = 2, entries => 1 : <CFString 0x102d17fb8 0x10296ba40>{contents = "HorizontalSpacing"} = <CFNumber 0xb000000000000012 0x10296ba40>{value = +1, type = kCFNumberSInt32Type} 2 : <CFString 0x102d17fd8 0x10296ba40>{contents = "VerticalSpacing"} = <CFNumber 0xb000000000000012 0x10296ba40>{value = +1, type = kCFNumberSInt32Type} } 4 : <CFString 0x102d17d78 0x10296ba40>{contents = "QTMovieTime"} = <CFBasicHash 0x7fa27c51db40 0x10296ba40>{type = immutable dict, count = 2, entries => 0 : <CFString 0x102d17d98 0x10296ba40>{contents = "TimeValue"} = <CFNumber 0xb000000000000003 0x10296ba40>{value = +0, type = kCFNumberSInt64Type} 1 : <CFString 0x102d17db8 0x10296ba40>{contents = "TimeScale"} = <CFNumber 0xb000000000075302 0x10296ba40>{value = +30000, type = kCFNumberSInt32Type} } 5 : <CFString 0x102d18138 0x10296ba40>{contents = "CVImageBufferColorPrimaries"} = <CFString 0x102d18178 0x10296ba40>{contents = "SMPTE_C"} 8 : <CFString 0x102d181b8 0x10296ba40>{contents = "CVImageBufferTransferFunction"} = <CFString 0x102d18078 0x10296ba40>{contents = "ITU_R_709_2"} 9 : <CFString 0x102d18318 0x10296ba40>{contents = "CVImageBufferChromaSubsampling"} = <CFString 0x102d18278 0x10296ba40>{contents = "TopLeft"} 10 : <CFString 0x102d18218 0x10296ba40>{contents = "CVImageBufferChromaLocationBottomField"} = <CFString 0x102d18338 0x10296ba40>{contents = "4:2:0"} 12 : <CFString 0x102d181f8 0x10296ba40>{contents = "CVImageBufferChromaLocationTopField"} = <CFString 0x102d18338 0x10296ba40>{contents = "4:2:0"} } nonPropagatedAttachments=<CFBasicHash 0x7fa27962ca60 0x10296ba40>{type = mutable dict, count = 0, entries => } >

我们从上面的日志输出找到了下面的东西

    <CVPixelBuffer 0x7fa27962c9c0 width=2048     height=1024 pixelFormat=420v iosurface=0x0 planes=2>
    <Plane 0 width=2048 height=1024 bytesPerRow=2048>
    <Plane 1 width=1024 height=512 bytesPerRow=2048>

我们能得到的信息是:

像素格式: 420v

数据通道: 2 个

通道1: width=2048 height=1024

通道2: width=1024 height=512

从上面信息可以得出我们数据的排列方式为YY....YY....UV.....UV,

2048_1024 个Y 数据，1024_512 从 bytesPerRow 可以看出每个Y、U、V 各占一个字节.

接下来就是如何将数据加载到我们的纹理缓冲区去了

CVReturn CVOpenGLESTextureCacheCreateTextureFromImage(
CFAllocatorRef CV_NULLABLE allocator,
CVOpenGLESTextureCacheRef CV_NONNULL textureCache,
CVImageBufferRef CV_NONNULL sourceImage,
CFDictionaryRef CV_NULLABLE textureAttributes,
GLenum target,
GLint internalFormat,
GLsizei width,
GLsizei height,
GLenum format,
GLenum type,
size_t planeIndex,
CV_RETURNS_RETAINED_PARAMETER CVOpenGLESTextureRef CV_NULLABLE * CV_NONNULL textureOut )

这个函数作用是: 通过CVImageBufferRef 创建一个纹理对象

allocator : 写默认值就可以了 kCFAllocatorDefault

textureCache：我们需要手动创建一个纹理缓冲对象,

sourceImage：传我们的CVImageBufferRef 数据

textureAttributes：纹理属性，可以为NULL

target:纹理的类型(GL_TEXTURE_2D 和GL_RENDERBUFFER)

internalFormat:数据格式,就是这个数据步伐的意思

width：纹理的高度

height : 纹理的长度

format: 像素数据的格式

type: 数据类型

planeIndex: 通道索引

接下来看我们的代码:

 // 启用纹理缓冲区0
glActiveTexture(GL_TEXTURE0);
err = CVOpenGLESTextureCacheCreateTextureFromImage(kCFAllocatorDefault,
                                                   _videoTextureCache,
                                                   pixelBuffer,
                                                   NULL,
                                                   GL_TEXTURE_2D,
                                                   GL_RED_EXT,
                                                   width,
                                                   height,
                                                   GL_RED_EXT,
                                                   GL_UNSIGNED_BYTE,
                                                   0,
                                                   &_lumaTexture);
if (err) {
    NSLog(@"Error at CVOpenGLESTextureCacheCreateTextureFromImage %d", err);
    
}

glBindTexture(CVOpenGLESTextureGetTarget(_lumaTexture), CVOpenGLESTextureGetName(_lumaTexture));
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);

// UV-plane.
// 启用纹理缓冲区1
glActiveTexture(GL_TEXTURE1);
err = CVOpenGLESTextureCacheCreateTextureFromImage(kCFAllocatorDefault,
                                                   _videoTextureCache,
                                                   pixelBuffer,
                                                   NULL,
                                                   GL_TEXTURE_2D,
                                                   GL_RG_EXT,
                                                   width /2,
                                                   height /2,
                                                   GL_RG_EXT,
                                                   GL_UNSIGNED_BYTE,
                                                   1,
                                                   &_chromaTexture);
if (err) {
    NSLog(@"Error at CVOpenGLESTextureCacheCreateTextureFromImage %d", err);
}


glBindTexture(CVOpenGLESTextureGetTarget(_chromaTexture), CVOpenGLESTextureGetName(_chromaTexture));
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameterf(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);

GL_RED_EXT 代表 1位数据 GL_RG_EXT 代表2位数据。UV 就是两位数据所以我们选择GL_RG_EXT。

刚才说了，参数中需要一个纹理缓冲TextureCacha,接下来我们就自己创建一个.

  CVReturn err = CVOpenGLESTextureCacheCreate(kCFAllocatorDefault, NULL, self.eagContext, NULL, &_videoTextureCache);

以上基本的工作都做完了，接下来，我们就只剩下显示了

渲染绘制 // 清除颜色缓冲区 glClearColor(0, 0, 0, 1); glClear(GL_COLOR_BUFFER_BIT); if (_isVR){ // 渲染双屏 glViewport(0, 0, self.view.bounds.size.width, self.view.bounds.size.height*2); glDrawElements(GL_TRIANGLES, _numIndices, GL_UNSIGNED_SHORT, 0); glViewport(self.view.bounds.size.width, 0, self.view.bounds.size.width, self.view.bounds.size.height*2); glDrawElements(GL_TRIANGLES, _numIndices, GL_UNSIGNED_SHORT, 0); }else{ // 渲染单屏 glViewport(0, 0, self.view.bounds.size.width*2, self.view.bounds.size.height*2); glDrawElements(GL_TRIANGLES, _numIndices, GL_UNSIGNED_SHORT, 0); }

到这里,视频已经可以显示了。

视图矩阵初始化 -(void)initModelViewProjectMatrix{ // 创建投影矩阵 float aspect = fabs(self.view.bounds.size.width / self.view.bounds.size.height); _projectionMatrix = GLKMatrix4MakePerspective(GLKMathDegreesToRadians(OSVIEW_CORNER), aspect, 0.1f, 400.0f); _projectionMatrix = GLKMatrix4Rotate(_projectionMatrix, ES_PI, 1.0f, 0.0f, 0.0f); // 创建模型矩阵 _modelViewMatrix = GLKMatrix4Identity; float scale = OSSphereScale; _modelViewMatrix = GLKMatrix4Scale(_modelViewMatrix, scale, scale, scale); // 最终传入到GLSL中去的矩阵 _modelViewProjectionMatrix = GLKMatrix4Multiply(_projectionMatrix, _modelViewMatrix); glUniformMatrix4fv(_modelViewProjectionMatrixIndex, 1, GL_FALSE, _modelViewProjectionMatrix.m); }
全景单屏模式手势操纵矩阵
(void)touchesMoved:(NSSet *)touches withEvent:(UIEvent *)event { if(self.isVR || self.vedioType == OSNormal ) return; UITouch *touch = touches anyObject; float distX = touch locationInView:touch.view.x - touch previousLocationInView:touch.view.x; float distY = touch locationInView:touch.view.y - touch previousLocationInView:touch.view.y; distX *= -0.005; distY *= -0.005; self.fingerRotationX += distY * OSVIEW_CORNER / 100; self.fingerRotationY -= distX * OSVIEW_CORNER / 100; _modelViewMatrix = GLKMatrix4Identity; float scale = OSSphereScale; _modelViewMatrix = GLKMatrix4Scale(_modelViewMatrix, scale, scale, scale); _modelViewMatrix = GLKMatrix4RotateX(_modelViewMatrix, self.fingerRotationX); _modelViewMatrix = GLKMatrix4RotateY(_modelViewMatrix, self.fingerRotationY); _modelViewProjectionMatrix = GLKMatrix4Multiply(_projectionMatrix, _modelViewMatrix); glUniformMatrix4fv(_modelViewProjectionMatrixIndex, 1, GL_FALSE, _modelViewProjectionMatrix.m); } - (void)touchesEnded:(NSSet *)touches withEvent:(UIEvent *)event { if (self.isVR || self.vedioType == OSNormal) return; for (UITouch *touch in touches) { self.currentTouches removeObject:touch; } } - (void)touchesCancelled:(NSSet *)touches withEvent:(UIEvent *)event { for (UITouch *touch in touches) { self.currentTouches removeObject:touch; } }
全景 VR模式使用角度传感器 -(void)startMotionManager{ self.motionManager = [CMMotionManager allocinit]; self.motionManager.deviceMotionUpdateInterval = 1.0 / 60.0; self.motionManager.gyroUpdateInterval = 1.0f / 60; self.motionManager.showsDeviceMovementDisplay = YES; self.motionManager startDeviceMotionUpdatesUsingReferenceFrame:CMAttitudeReferenceFrameXArbitraryCorrectedZVertical; self.referenceAttitude = nil; [self.motionManager startGyroUpdatesToQueue: [NSOperationQueue allocinit] withHandler:^(CMGyroData * _Nullable gyroData, NSError * _Nullable error) { if(self.isVR) { self calculateModelViewProjectMatrixWithDeviceMotion:self.motionManager.deviceMotion; } }]; self.referenceAttitude = self.motionManager.deviceMotion.attitude; } -(void)calculateModelViewProjectMatrixWithDeviceMotion:(CMDeviceMotion*)deviceMotion{ _modelViewMatrix = GLKMatrix4Identity; float scale = OSSphereScale; _modelViewMatrix = GLKMatrix4Scale(_modelViewMatrix, scale, scale, scale); if (deviceMotion != nil) { CMAttitude *attitude = deviceMotion.attitude; if (self.referenceAttitude != nil) { attitude multiplyByInverseOfAttitude:self.referenceAttitude; } else { self.referenceAttitude = deviceMotion.attitude; } float cRoll = attitude.roll; float cPitch = attitude.pitch; _modelViewMatrix = GLKMatrix4RotateX(_modelViewMatrix, -cRoll); _modelViewMatrix = GLKMatrix4RotateY(_modelViewMatrix, -cPitch*3); _modelViewProjectionMatrix = GLKMatrix4Multiply(_projectionMatrix, _modelViewMatrix); // 下边这个方法必须在主线程中完成. dispatch_async(dispatch_get_main_queue(), ^{ glUniformMatrix4fv(_modelViewProjectionMatrixIndex, 1, GL_FALSE, _modelViewProjectionMatrix.m); }); }}

操作矩阵这里，暂时不想讲,后面我会专门来讲矩阵变换和角度传感器的使用,因为这两个东西在游戏和VR,还是AR的世界,都太重要了。今天先说的这里，给几张展示图欣赏一下。