Abstract:Land-cover information is significant for land-use planning, urban management, and environment monitoring. This paper presented a novel extended topology-preserving segmentation (ETPS)-based multi-scale and multi-feature method using the convolutional neural network (EMMCNN) for high spatial resolution (HSR) image land-cover classification. The EMMCNN first segmented the images into superpixels using the ETPS algorithm with false-color composition and enhancement and built parallel convolutional neural networks (CNNs) with dense connections for superpixel multi-scale deep feature learning. Then, the multi-resolution segmentation (MRS) object hand-delineated features were extracted and mapped to superpixels for complementary multi-segmentation and multi-type representation. Finally, a hybrid network was designed to consist of 1-dimension CNN and multi-layer perception (MLP) with channel-wise stacking and attention-based weighting for adaptive feature fusion and comprehensive classification. Experimental results on four real HSR GaoFen-2 datasets demonstrated the superiority of the proposed EMMCNN over several well-known classification methods in terms of accuracy and consistency, with overall accuracy averagely improved by 1.74% to 19.35% for testing images and 1.06% to 8.78% for validating images. It was found that the solution combining an appropriate number of larger scales and multi-type features is recommended for better performance. Efficient superpixel segmentation, networks with strong learning ability, optimized multi-scale and multi-feature solution, and adaptive attention-based feature fusion were key points for improving HSR image land-cover classification in this study.

An Image Segmentation Method Using Image Enhancement and Pcnn with Adaptive Parameters

Automatic Image Segmentation Based on PCNN with Adaptive Threshold Time Constant.

Improved Pulse Coupled Neural Network for Target Segmentation in Infrared Images

A New Algorithm of Automatic Image Segmentation Based on Pcnn

An adaptive image segmentation method based on a modified pulse coupled neural network

Image segmentation method based on an improved PCNN

Image Segmentation Based on Activity Degree with Pulse Coupled Neural Networks

A New Method of PCNN's Parameter's Optimization

Image Segmentation Based on Double-Level Parallelized Firing Pcnn in Complex Environments

A Segmentation Method of Color Images Based on Improved PCNN

Image Segmentation with PCNN Using Histogram and Contoured Product Mutual Information

Emmcnn: An Etps-Based Multi-Scale And Multi-Feature Method Using Cnn For High Spatial Resolution Image Land-Cover Classification

Medical Images Segmentation Based on Improved Three-Dimensional Pulse Coupled Neural Network

Image Segmentation Based On Pulse Coupled Neural Network

Automated Image Segmentation Using the ULPCNN Model with Ultra-Fuzzy Entropy

PCNN-based Image Segmentation with Contoured Product Mutual Information Criterion

A Modified PCNN Model for Images Processing

A New Texture Image Segmentation Method Based On Gabor Wavelet And Improved Pcn

An improvement on competitive neural networks applied to image segmentation

An Image Segmentation Approach Based on Fuzzy-Neural-network Hybrid System

An Improved MBNN for Image Segmentation