Multi-source collaborative enhanced for remote sensing images semantic segmentation

MFCA-Net: a deep learning method for semantic segmentation of remote sensing images

Semantic segmentation of remote sensing images (RSI) is an important research direction in remote sensing technology. This paper proposes a multi-feature fusion and channel attention network, MFCA-Net, aiming to improve the segmentation accuracy of remote sensing images and the recognition performance of small target objects. The architecture is built on an encoding-decoding structure. The encoding structure includes the improved MobileNet V2 (IMV2) and multi-feature dense fusion (MFDF). In IMV2, the attention mechanism is introduced twice to enhance the feature extraction capability, and the design of MFDF can obtain more dense feature sampling points and larger receptive fields. In the decoding section, three branches of shallow features of the backbone network are fused with deep features, and upsampling is performed to achieve the pixel-level classification. Comparative experimental results of the six most advanced methods effectively prove that the segmentation accuracy of the proposed network has been significantly improved. Furthermore, the recognition degree of small target objects is higher. For example, the proposed MFCA-Net achieves about 3.65-23.55% MIoU improvement on the dataset Vaihingen.

Mask R-CNN-Oriented Pottery Display and Identification System

Traditional pottery identification methods are time consuming and costly. In order to cater to more pottery industry needs, we propose a Mask R-CNN-based pottery identification method to build an automatic pottery identification system. We first improve the loss function of Mask R-CNN by using generalized intersection over union loss function, through the pattern of migration learning, to compensate for the disadvantage of a small collective amount of pottery data. For different types of pottery, we use the mask algorithm to enhance the features of the outer contour of the pottery. In addition, we use the minimum external matrix algorithm to accurately extract the outer contour bit pose features of pottery to improve the model's accuracy in recognizing the outer contour of pottery. To meet the testing conditions of pottery, with the support of potters and archaeologists, we make our pottery data set according to pottery categories. The experimental results prove that our method performs best in the comprehensive recognition accuracy of pottery, with the recognition accuracy above 90%. The recognition accuracy is also the best in pottery color decoration and grain decoration, and the grain recognition accuracy stays above 87%, which is better than other pottery recognition methods.