论文部分内容阅读
提出了一个新的两阶段二维多速度运动检测模型.该模型首先利用基于乘积侧抑制速度检测子MIVD的一维速度检测特性,构成了一维多速度运动感知层.检测出一维速度信息,然后利用胞神经网络(CNN)模拟皮层功能柱间的局部动态交互,完成多速度二维运动检测.实验证明,这种分层并行多频道模型是对初级视觉运动感知的良好建模.应用这种模型可以实现基于“运动信息驱动”的注意力集中机制,也可用于构成视觉高层反馈的“目的驱动”机制.
A new two-phase two-dimensional multi-velocity motion detection model is proposed. First, the model uses one-dimensional velocity detection based on product side suppression velocity detection sub-MIVD to form a one-dimensional multi-velocity motion perception layer. One-dimensional velocity information is detected, and then the CNN is used to simulate the local dynamic interaction between the functional cortical columns to complete multi-speed two-dimensional motion detection. Experiments show that this hierarchical parallel multi-channel model is a good model for perception of primary visual motion. The application of this model can be based on “motion information driven” attention concentration mechanism can also be used to form the vision of high-level feedback “purpose-driven” mechanism.