Multi-target detection method and system adaptive to multi-band image

Abstract

The invention discloses a multi-target detection method and system adaptive to multi-band images, and the method comprises the steps: firstly carrying out the preprocessing of inputted multi-band image data, and obtaining a multi-band image sequence which is uniform in size and matched in features; generating a feature fusion graph for retaining the significant features of the multi-band image byusing a self-encoding neural network method; constructing a multiband image joint data set by applying a labeling tool and a shared labeling mode; training and verifying a multiband detection model Multiband-net built by a convolutional neural network to obtain a model file containing weight parameters; and calling the model file to achieve cross-band flexible detection of visible light, infraredlight and a fused image of the visible light and the infrared light, and outputting a detection result image and a video. The method can be widely applied to the deep learning training process, the precision of the target detection model and the generalization performance of cross-band detection are improved, the average precision of the system classification model in the visible light band and the average precision of the system classification model in the infrared band reach 84.89% and 87.95% respectively, and better detection performance is obtained on different adaptive bands.

Description

technical field [0001] This application relates to the field of deep learning and image processing technology, in particular to a multi-band image joint data set production method based on image registration and feature fusion technology, and multi-band image joint data set training based on this method to obtain multi-band targets A method and system for testing a neural network model. Background technique [0002] Multi-object detection is a basic problem in the field of computer vision, and it is the basis for solving higher-level visual tasks such as object recognition, object tracking, semantic segmentation, image description, scene understanding, and event detection, and it is widely used in many fields of artificial intelligence and information technology. There are a wide range of applications, such as robot vision, autonomous driving, content-based image retrieval, intelligent video surveillance, etc. The multi-target detection technology based on multi-band featur...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem of increased redundancy of detection models under the requirements of multi-band detection, and improve the overall detection accuracy of the detection system and the generalization of cross-band detection, the present invention provides a multi-target detection method and system adapted to multi-band images, specifically It is a multi-target detection neural network model construction method and detection system that can adapt to visible light and infrared multi-band images at the same time, including a multi-band image joint data set production method based on image registration and feature fusion technology, and a A method for constructing a multi-band target detection neural network model, through training, testing and calling the model to achieve high-precision detection of multi-band targets

Images