Research of Infrared Image Segmentation of Composite Insulator Based on Feature Fusion

With the advantage of the encoding-decoding structure of U-net, this paper proposes a composite insulator infrared image segmentation method, which focuses on the feature fusing of the extracted image and temperature value features. Firstly, the infrared images of composite insulators are transformed to derive the temperature data matrix with the same size. Secondly, the feature extraction network, which is used as the encoder of U-net network, is adopted for the feature extraction of the infrared image and temperature data matrix respectively. Then, in the decoding process, the feature fusing is realized that the outputs of feature extraction networks are fused and the fused feature map is gradually restored to the original input size through up sampling. Finally, the experimental results show that the average intersection-over-union of proposed model is 74.09 ^% and the precision is 84.33%, which are both better than the results of traditional U-net with one type of input.