Target detection method and moving target tracking method using same

Abstract

The invention provides a target detection method. The method comprises the following steps: inputting a to-be-detected image into a backbone network to extract an initial feature map; inputting the initial feature map into enhancement layers of a neck network to extract at least one enhanced feature map, wherein each enhancement layer of the neck network is provided with an FPN structure, a PAN structure and an attention module specially referred to by the target feature which are connected in sequence; the attention module of the target feature especially comprises a space attention branch and a feature especially branch, the output of the space attention branch and the output of the feature especially branch are subjected to element level multiplication and normalization, and then element level multiplication and element level addition are sequentially carried out on the output of the space attention branch and the output of the feature especially branch and the corresponding fusion feature map to obtain a corresponding enhanced feature map; and each enhanced feature map into a head prediction network is input to obtain a prediction result. According to the method, the attention module specially referred to by the target features can effectively sense the space areas occupied by different types of objects through the combined use of the two branches, and the accuracy and efficiency of target detection are improved.

Description

technical field [0001] The present application relates to the technical field of target detection, in particular to a target detection method and a moving target tracking method using the same. Background technique [0002] Computer vision has gradually become more and more important in people's lives. It is widely used in robot navigation, intelligent video surveillance, industrial inspection, aerospace and many other fields. It has important practical significance to reduce the consumption of human capital through computer vision; Detection and target tracking is a popular direction of computer vision and digital image processing. It is an important branch of computer vision and image processing, and it is also the core part of the intelligent monitoring system; due to the further development of the combination of artificial intelligence and cameras, the intelligent monitoring system And slowly began to come into everyone's sight. [0003] Although some progress has been ...

AI Technical Summary

Problems solved by technology

[0003] Although some progress has been made in the field of object detection, there are still many problems in the actual design and use of object detection methods and object detection models.

On the one hand, in the attention module in the target detection model, the convolutional neural network will accumulate many convolutional layers and pooling layers to obtain contextual semantic information, which increases the complexity of time and space, and the existing spatial attention module Or channel attention will ignore the underlying features in the image to be detected. In particular, common spatial attention modules and channel attention modules cannot effectively perceive the spatial regions occupied by different types of objects.

On the other hand, in single-stage object detection, it is often achieved by optimizing object classification and object localization, but conventional object classification and object localization are parallel, which may lead to a certain degree of spatial misalignment.

Images