基于神经网络与形态学的一种区域标识方法

Region Labeling Method Based on Double-Pulse Coupled Neural Networks and Morphology

提出了一种双层脉冲耦合神经网络与数学形态学相结合的区域标识算法。首先利用一层脉冲耦合神经网络对图像进行分割,得到初步的结果;再利用另一层网络检测到的边缘点与形态学结合,对分割的区域进行相应的操作:区域定位,合并小区域,提取边界,对边界进行细化;最后对区域进行了标识。仿真结果证明,该算法对于后续的特征度量和模式识别具有重要的借鉴价值和实际意义。

This paper brings forward a new approach for region labeling by using double Pulse Coupled Neural Networks （PCNN） and morphology. Firstly, the preliminary result is got by using one PCNN and then further operation is performed on the result using the other PCNN and morphology： region orientation, uniting small region, edge extraction and thinning, and region labeling. The results of computer simulations show that the method is useful and meaningful to pattern recognition.