摘要
为了获得较好的图像边缘信息,提出了一种将BP神经网络与形态学融合的边缘检测算法。在BP神经网络中常用Sigmoid函数作为激励函数,但传统的Sigmoid函数形式单一、缺少灵活性,因此给出具有可调节性的Sigmoid函数是至关重要的。首先,给出一类充分光滑的Sigmoid函数构造方法,将其作为BP神经网络中的激励函数对图像的边缘进行初步有效的检测。然后,采用多尺度多结构的思想,提出了一种改进的形态学边缘检测算法,应用该算法得到噪声小且连续的边缘图像。最后,利用小波分析将BP神经网络与改进的形态学算法进行融合,进而得到一种边缘检测融合算法。仿真结果表明,融合算法的评价指标优于单一的边缘检测算法且检测的图像边缘线条完整并清晰。
In order to obtain better image edge information,an edge detection algorithm combining BP(Back Propagation)neural network and morphology is proposed.The Sigmoid function is commonly used as the excitation function in BP neural networks,but the traditional Sigmoid function is single in form and lacks flexibility.Therefore,it is very important to provide an adjustable Sigmoid function.First,a fully smooth Sigmoid function construction method is given,which is used as the excitation function in the BP neural network to detect the edge of the image effectively.Then,using the idea of multi-scale and multi-structure,an improved morphological edge detection algorithm is proposed,and the edge image with small noise and continuous is obtained by applying this algorithm.Finally,the wavelet analysis is used to fuse the BP neural network and the improved morphological algorithm,and then an edge detection fusion algorithm is obtained.The simulation results show that the evaluation index of the fusion algorithm is better than a single edge detection algorithm and the detected image edge lines are complete and clear.
作者
岳欣华
邓彩霞
张兆茹
YUE Xin-hua;DENG Cai-xia;ZHANG Zhao-ru(School of Applied Sciences, Harbin University of Science and Technology, Harbin 150080, China)
出处
《哈尔滨理工大学学报》
CAS
北大核心
2021年第5期83-90,共8页
Journal of Harbin University of Science and Technology
基金
国家自然科学基金(11871181).