在看ｔｉ公司提供的ＳＣＤ运动目标检测中，说SCD做两项工作：
At a high level, SCD is a body of methods that work together to perform two primary operations:
a) monitor video for differences between video frames, and
b) monitor video for differences between a learned model of the scene and incoming frames as well as the presence of blockage in front of a camera.
第一种是比较两个视频帧之间的差异
第二个是一种学习模型场景和输入帧之间的差异
我的问题是a learned model of the scene是一个什么样的学习方法，是不是通过这个方法可以产生类似于输入帧一样的视频帧？tamper detection翻译成篡改检测？
