Multispectral target detection blind guiding system

Abstract

The invention provides a multispectral target detection blind guiding system, which comprises a data input module used for obtaining a visible light image and an infrared thermal image; a deformable feature extractor module used for respectively extracting a visible light image feature map and an infrared thermal image feature map; a candidate box extraction network used for extracting a visible light image candidate box and an infrared thermal image candidate box; a candidate box complementation module used for adding the part, which is not covered by the infrared thermal image candidate box,in the visible light image candidate box into the infrared thermal image candidate box, and adding the part, which is not covered by the visible light image candidate box, in the infrared thermal image candidate box into the visible light image candidate box; and obtaining a visible light image region feature map and an infrared thermal image region feature map; a cross-modal attention fusion module used for fusing the visible light image region feature map into the infrared thermal image region feature map according to the similarity relationship among the region features to obtain enhancedthermal image features; and a classification and regression module used for obtaining a target detection result.

Description

technical field [0001] The invention relates to the field of computers, in particular to a multi-spectral target detection and blinding system. Background technique [0002] In recent years, the tremendous development of computer vision has brought new opportunities and possibilities to the blind guide system. Deep learning models based on convolutional neural networks (CNN) have reached or even surpassed human performance in image classification (ImageNet dataset) and object detection (COCO dataset). The visual perception system (especially the object detection system) based on deep learning technology has achieved good results in applications such as unmanned driving. Therefore, using this technology to assist the blind to perceive the environment has formed a new trend. However, the previous target detection models were generally constructed based on visible light color images, and the applicable scenes were limited by the lighting conditions, so they could not be appli...

AI Technical Summary

Problems solved by technology

However, previous target detection models are generally based on visible light color images, and the applicable scenes are limited by lighting conditions, so they cannot be applied at night or in places with strong light

Similarly, the guide system using this technology also has this problem, and cannot assist the blind to perceive the environment around the clock

Images