最近，我一直在使用 Assimp 库在 OpenGL 中加载和播放骨骼动画，遵循一些不同的教程/资源，例如OglDev和Ephenation OpenGL。我仍然不是 100% 确定我知道它是如何工作的，所以澄清一下：
- 关键帧从
aiNodeAnim's 的mChannels成员中加载aiAnimation - 这些关键帧用于生成每个骨骼矩阵，表示骨骼从其父骨骼的转换
- 每个骨骼的矩阵乘以它的父矩阵
- 然后将该矩阵乘以相应骨骼的偏移矩阵，从
aiMesh->mBones[BONE]->mOffsetMatrix。这样，从关键帧生成的骨骼矩阵可以替换该骨骼的绑定姿势的默认变换 - 绘制时在顶点着色器中使用这个最终矩阵（每个骨骼一个）
我的问题的第一部分：这是正确的吗？据我所知，假设我的理解是正确的，它应该在逻辑上起作用。我还没有看到任何反对这一点的来源/帖子/问题。
我已经编写了一个实现（在 C++ 中），但它并没有像我预期的那样工作。部分网格看起来像他们应该的样子，甚至处于预期的姿势，但大部分网格看起来都皱巴巴的。大部分是可区分的，我可以识别大部分网格，但似乎大多数骨骼矩阵都不正确。当我在搅拌机中打开我的模型时，所有关键帧都是正确的，并且模型动画完美。我会发布我的应用程序的屏幕截图，但我没有地方可以这样做（我没有 imageshack 帐户或任何东西。）
有什么理由为什么只有一些骨头会起作用而其他骨头不起作用？
以下是我的代码中最有问题的部分。我的jawSkeleton课程基于ModelLearnOpenGL.com 的课程。（忽略jaw前缀，这是我做的私人事情。）
类原型，在jawSkeleton.h：
class jawSkeleton
{
public:
/* Functions */
// Constructor, expects a filepath to a 3D model.
jawSkeleton(GLchar* path);
jawSkeleton();
// Draws the model, and thus all its meshes
void Draw();
/* Functions */
// Loads a model with supported ASSIMP extensions from file and stores the resulting meshes in the meshes vector.
void Load(std::string path, bool bSmoothNormals = false, UINT unFlags = NULL);
private:
bool Loaded;
/* Model Data */
std::vector<jawSkeletalMesh> meshes;
std::string directory;
// Stores all the textures loaded so far, optimization to make sure textures aren't loaded more than once.
std::vector<jawSub_Texture> textures_loaded;
jawSkeletalBone* RootBone;
std::vector<jawSkeletalBone> bones;
glm::mat4 Matrices[64];
glm::mat4 GlobalInverseTransform;
std::vector<jawSkeletalAnim> Animations;
// finds a bone in the model data with the specified name
jawSkeletalBone* findBone(std::string Name);
// finds a bone in the model data with the specified name (Must be 'int' so that we can use -1 to denote no bone)
int findBoneIndex(std::string Name);
// Processes a node in a recursive fashion.
// Processes the bones identified by the meshes
void processNodeMeshBones(aiNode* node, const aiScene* scene, std::vector<std::string> *BoneNames);
// Populates the bone heirarchy structure
void populateBoneHeirarchy(const aiScene* scene, std::vector<std::string> *BoneNames);
// Processes each individual mesh located at the node and repeats this process on its children nodes (if any).
void processNodeMeshes(aiNode* node, const aiScene* scene);
jawSkeletalMesh processMesh(aiMesh* mesh, const aiScene* scene);
// Checks all material textures of a given type and loads the textures if they're not loaded yet.
// The required info is returned as a jawSub_Texture struct.
std::vector<jawSub_Texture> loadMaterialTextures(aiMaterial* mat, aiTextureType type, std::string typeName);
// Load an animation for the model
void processAnim(aiAnimation* Anim);
// Updates the matrices for every bone
// DOES NOT pull keyframes from the animation
// Merely travels down the bone heirarchy and multiplies each child matrix by the parent matrix
void UpdateBoneMatrices();
// Updates the array of matrices that is sent to the shader
void UpdateBoneMatrixArray();
// Sets all the bone matrices to match a particular time
// Before drawing, call UpdateBoneMatrices and UpdateBoneMatrixArray after calling this function
void SetBoneMatrices(double Time);
};
加载关键帧数据，在jawSkeleton.cpp：
void jawSkeleton::processAnim(aiAnimation* Anim) {
this->Animations.push_back(jawSkeletalAnim());
jawSkeletalAnim &_Anim = this->Animations.back();
_Anim.Name = Anim->mName.C_Str();
double TicksPerSecond = (Anim->mTicksPerSecond == 0.0) ? 1.0 : Anim->mTicksPerSecond;
_Anim.Duration = Anim->mDuration / TicksPerSecond;
for (GLuint i = 0; i < Anim->mNumChannels; i++) {
aiNodeAnim* ThisAnim = Anim->mChannels[i];
jawSkeletalBoneAnim BoneAnim;
BoneAnim.Index = this->findBoneIndex(ThisAnim->mNodeName.C_Str());
if (BoneAnim.Index > this->bones.size() || BoneAnim.Index < 0)
continue;
// Translation
BoneAnim.TranslateKeys.reserve(ThisAnim->mNumPositionKeys);
for (GLuint j = 0; j < ThisAnim->mNumPositionKeys; j++) {
aiVector3D v = (ThisAnim->mPositionKeys + j)->mValue;
BoneAnim.TranslateKeys.push_back(jawTranslateKey{
glm::vec3(v.x, v.y, v.z),
ThisAnim->mPositionKeys[j].mTime / TicksPerSecond
});
}
// Rotation
BoneAnim.RotateKeys.reserve(ThisAnim->mNumRotationKeys);
for (GLuint j = 0; j < ThisAnim->mNumRotationKeys; j++) {
aiQuaternion v = (ThisAnim->mRotationKeys + j)->mValue;
BoneAnim.RotateKeys.push_back(jawRotateKey{
glm::quat(v.w, v.x, v.y, v.z),
ThisAnim->mRotationKeys[j].mTime / TicksPerSecond
});
}
// Scaling
BoneAnim.ScaleKeys.reserve(ThisAnim->mNumScalingKeys);
for (GLuint j = 0; j < ThisAnim->mNumScalingKeys; j++) {
aiVector3D v = (ThisAnim->mScalingKeys + j)->mValue;
BoneAnim.ScaleKeys.push_back(jawScaleKey{
glm::vec3(v.x, v.y, v.z),
ThisAnim->mScalingKeys[j].mTime / TicksPerSecond
});
}
// The BoneAnims member is an std::unordered_map
_Anim.BoneAnims.insert(std::pair<GLuint, jawSkeletalBoneAnim>(BoneAnim.Index, BoneAnim));
}
}
准备各个矩阵，在jawSkeletal.cpp：
void jawSkeleton::SetBoneMatrices(double Time) {
// This is the function to edit
jawSkeletalAnim& Anim = this->Animations[0];
for (GLuint i = 0; i < this->bones.size(); i++) {
auto Result = Anim.BoneAnims.find(i);
if (Result == Anim.BoneAnims.end())
continue;
jawSkeletalBoneAnim &BoneAnim = Result->second;
glm::mat4 T = glm::translate(glm::mat4(1), BoneAnim.TranslateKeys[70].Value);
glm::mat4 R = glm::toMat4(BoneAnim.RotateKeys[70].Value);
glm::mat4 S = glm::scale(glm::mat4(1), BoneAnim.ScaleKeys[70].Value);
this->bones[i].CurrentMatrix = T * R * S;
}
}
我做了一些其他未显示的事情，例如递归地将每个骨骼的矩阵乘以其父矩阵，然后将它们乘以它们的偏移矩阵。当我使用单位矩阵代替单个骨骼矩阵时，网格是以其正常绑定姿势绘制的，所以我几乎可以肯定这不是网格导入问题。正如我所说，部分网格看起来是正确的，所以我猜这也不是蒙皮问题。我已经研究这个问题大约一个星期了。有什么线索吗？任何帮助表示赞赏。