Single-emission multi-frame infrared target detection method

Abstract

The invention relates to a single-transmission multi-frame infrared target detection method, which comprises the following steps of: starting from a feature pyramid network, describing the inequalityof contribution of each feature layer to fusion output on the basis of learnable weights, achieving bidirectional multi-scale feature weighted fusion between a low-resolution and strong-semantic feature layer and a high-resolution and weak-semantic feature layer, and constructing an auxiliary network; based on the cross-parallel ratio, considering the influence of an overlapping region and a non-overlapping region on a target function at the same time, constructing detector positioning loss keeping invariance to target scale change, constructing a target detector, and improving the sensitivityof a detection model to small target positioning errors. A VGG16 convolutional neural network is used as a feature extraction network and is integrated with an auxiliary network and a target detectorto form a single-shot multi-frame infrared target detection model fusing multi-scale feature weighted fusion and scale invariance positioning loss. The method has autonomous learning ability and highdetection rate, and is an effective way for solving the infrared imaging guidance target detection problem in a complex environment.

Description

technical field [0001] The invention belongs to the technical field of infrared target detection, and in particular relates to a single-shot multi-frame infrared target detection method. Background technique [0002] At present, target detection is the basis for the automatic target recognition system of infrared imaging guidance to complete subsequent tasks such as recognition and tracking. The existing system does not have the ability to learn target characteristics independently, and once the task environment exceeds the pre-planned conditions, it will be powerless. The single-stage target detection based on deep learning has self-learning ability and high computational efficiency, which is an effective way to solve the problem of infrared imaging guidance target detection in complex environments. Single Shot MultiBox Detector (SSD) is a classic single-stage detection model. The SSD target detection model can be decomposed into two modules: feature extractor and target d...

AI Technical Summary

Problems solved by technology

[0004] In view of this, in order to solve the above-mentioned deficiencies in the prior art, the object of the present invention is to provide a single-shot multi-frame infrared target detection method, which integrates multi-scale feature weighted fusion and scale-invariant positioning loss, and has the ability of self-learning and detection It is an effective way to solve the problem of infrared imaging guidance target detection in complex environments.

Images