基于参数自适应脉冲耦合神经网络的黄瓜目标分割

Cucumber Image Segmentation Based on Weighted Connection Coefficient Pulse Coupled Neural Network

摘要:

对脉冲耦合神经网络的参数进行简化，并自适应确定各参数，将图像的空间信息和灰度信息耦合到加权耦合连接系数中，进行温室黄瓜图像分割，采用二维Tsallis熵选择最佳迭代结果。试验结果表明：用区域对比度（GC）和区域一致性（UC）评价方法评价，该方法的分割效果好于采用香农熵和最小交叉熵终止迭代的标准脉冲耦合神经网络分割效果。

Abstract:

Parameters of pulse coupled neural network(PCNN) were simplified and adaptive to determine. Spatial information and gray information of image were coupled to the weighted connection coefficient for greenhouse cucumber segmentation by using the 2-D Tsallis entropy to select the best results of iteration. Experimental results showed that, methods of contrast and regional consistency were employed to evaluate effect of different segmentation. Segmentation results of prospered method was better than using Shannon entropy and minimum cross entropy to terminate iteration of standard pulse coupled neural network segmentation.