Multi-band Image Fusion With Infrared Broad Spectrum For Low And Slow Small Target Recognition

Abstract

While the widespread use of low and slow UAVs brings convenience to all areas of society, it also poses a serious threat to the safety of the low-altitude domain. In the current field, radar detection and identification technology and infrared image recognition technology are widely used in target detection and identification. The neural network designed in this paper adopts fully connected neural network and convolutional neural network to extract global feature information and local feature information from the infrared broad spectrum data of low and slow small targets respectively, and the extracted feature information is fed into the target detection networks of different time periods for recognition training to obtain image recognition models and spectral recognition models of different time periods, and finally, the image recognition and spectral recognition Finally, the recognition rates of image recognition and spectral recognition are fused to obtain the final recognition rate. By combining the strengths of infrared hyperspectral images, making up for the deficiencies of multi-band images for target hours which are not easy to recognize, and fusion processing at multiple levels, the multi-band images break through the limitations of airborne target recognition, improve the anti-interference ability of recognition network, and also improve the accuracy rate of airborne target recognition.