# 基于异形卷积核的卷积神经网络图像分类方法Based on the Heterogeneous Convolution Kernel Image Classification of Convolutional Neural Network

DOI: 10.12677/CSA.2020.1011207, PDF, HTML, XML, 下载: 89  浏览: 180  科研立项经费支持

Abstract: With a strong ability to learn image features, convolution neural network has made a breakthrough in machine learning. Directing at the shortcomings of the current convolution neural network methods, the concept of heterogeneous convolution kernel is creatively proposed by changing the shape of the convolution kernel, expanding its receptive field and improving the ability of the network to extract image features, without changing the number of convolution kernel parameters. Through the method of image shift, the feasibility of heterogeneous convolution kernel is solved. The convolution neural network which combines the heterogeneous convolution kernel and the rectangular convolution kernel is constructed in order that the heterogeneous convolution kernel can be effectively combined with the traditional convolution neural network. The experimental results show that, compared with the traditional convolution neural network, the convolution neural network combined with special-shaped convolution kernel has higher classification accuracy.

1. 引言

2. 卷积神经网络(CNN)

Figure 1. CNN network structure diagram

$\underset{w,b}{\mathrm{min}}C\left(\omega ,b\right)=\frac{1}{2n}{\underset{x\in X}{\sum }‖y\left(x\right)-S\left(w,b,x\right)‖}^{2}$(1)

3. 异形卷积核

(a) 左凸异形 (b) 右凸异形(c) 上凸异形 (d) 下凸异形(e) 叠加感受野

Figure 2. Shaped convolution kernel and its receptive field

4. 融合异形卷积核和矩形卷积核的卷积神经网络

4.1. 异形卷积核的实现

${\stackrel{¯}{c}}_{i,j}=\underset{s=1}{\overset{3}{\sum }}\underset{t=1}{\overset{3}{\sum }}\left({\alpha }_{s,t}×{\stackrel{¯}{b}}_{i+s-2,j+t-2}\right)\text{}$(3)

${\stackrel{¯}{b}}_{i,j}=\left\{\begin{array}{l}{b}_{i,j}\text{}i=2k+1,\text{}k=1,2,\cdots ,m-1,\text{}j=1,2,\cdots ,n\text{}\\ {b}_{i,j-1}\text{}i=2k,\text{}k=1,2,\cdots ,m,\text{}j=1,2,\cdots ,n\end{array}$ (4)

Figure 3. The realization of the special-shaped convolution kernel

4.2. 融合移位图片的卷积神经网络

Figure 4. Converged network structure

5. 实验

Figure 5. Network structure

Table 1. Experimental data

Table 2. Experimental results

Table 3. Comparison of accuracy rates of various methods

 [1] Lecun, Y. and Bottou, L. (1998) Gradient-Based Learning App-Lied to Document Recognition. Proceedings of the IEEE, 86, 2278-2324. https://doi.org/10.1109/5.726791 [2] Krizhevsky, A., Sutskever, I.I. and Hinton, G. (2012) Imagenet Classification with Deep Convolutional Neural Networks. Proceedings of the Advances in Neural Information Processing Systems, Lake Tahoe, 1097-1105. [3] Simonyan, K. and Zisserman, A. (2014) Very Deep Convolutional Networks for Large-Scale Image Recognition. ICLR 2015, San Diego, 7-9 May 2015, 4898-4906. [4] Szegedy, C., Liu, W., Jia, Y., et al. (2014) Going Deeper with Convolutions. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Boston, 7-12 June 2015, 1-9. https://doi.org/10.1109/CVPR.2015.7298594 [5] He, K., Zhang, X., Ren, S., et al. (2016) Deep Residual Learning for Image Recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, 27-30 June 2016, 770-778. https://doi.org/10.1109/CVPR.2016.90 [6] Huang, G., Liu, Z., Weinberger, K.Q., et al. (2016) Densely Connect-ed Convolutional Net Works. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, 21-26 July 2017, 2261-2269. https://doi.org/10.1109/CVPR.2017.243 [7] Sabour, S., Frosst, N. and Hinton, G.E. (2017) Dynamic Routing between Capsules. Advances in Neural Information Processing Systems, Long Beach, 3856-3866. [8] Gu, J., Wang, Z., Kuen, J., et al. (2015) Recent Advances in Convolutional Neural Networks. Computer Science. [9] Zeiler, M.D. and Fergus, R. (2014) Visualizing and Understanding Convolutional Networks. In: European Conference on Computer Vision, Springer, Berlin, 818-833. https://doi.org/10.1007/978-3-319-10590-1_53 [10] Cao, K. and Jain, A.K. (2015) Latent Orientation Field Estimation via Convolutional Neural Network. International Conference on Biometrics, Phuket, 19-22 May 2015, 349-356. https://doi.org/10.1109/ICB.2015.7139060 [11] 李章维, 胡安顺, 王晓飞. 基于视觉的目标检测方法综述[J]. 计算机工程与应用, 2020, 56(8): 1-9. [12] 陈超, 齐峰. 卷积神经网络的发展及其在计算机视觉领域中的应用综述[J]. 计算机科学, 2019, 46(3): 63-73. [13] 唐贤伦, 杜一铭, 刘雨微, 李佳歆, 马艺玮. 基于条件深度卷积生成对抗网络的图像识别方法[J]. 自动化学报, 2018, 44(5): 855-864. [14] Zhang, K., Guo, Y., Wang, X., et al. (2019) Multiple Feature Reweight DenseNet for Image Classification. IEEE Access, 7, 9872-9880. https://doi.org/10.1109/ACCESS.2018.2890127 [15] 张顺, 龚怡宏, 王进军. 深度卷积神经网络的发展及其在计算机视觉领域的应用[J]. 计算机学报, 2019, 42(3): 453-482.