论文提出了基于阴影提取、车辆纹理与灰度对称性相结合的车辆检测方案。
This thesis proposes a vehicle detection proposal based on the extraction of shadow with a combination of vehicles' texture and symmetry of gray-scale.
然后，通过分形维数计算车辆的纹理特征，排除非车辆区域。
Second, the no-vehicle regions are filtered by texture feature of AOI based on DBC.
通过分析阴影的物理模型，得出局部纹理的光照不变性，利用基于比值判决的LBP纹理法来区分运动车辆和阴影，并应用亮度约束和几何启发式准则进一步改善阴影检测效果。
The distribution of improved LBP texture is discussed and a significant test is performed to classify each moving pixel into foreground object or moving shadow according to this theory.
通过分析阴影的物理模型，得出局部纹理的光照不变性，利用基于比值判决的LBP纹理法来区分运动车辆和阴影，并应用亮度约束和几何启发式准则进一步改善阴影检测效果。
The distribution of improved LBP texture is discussed and a significant test is performed to classify each moving pixel into foreground object or moving shadow according to this theory.
应用推荐