Small target detection network model based on darknet53 network and detection method

Abstract

The invention provides a small target detection network model based on a darknet53 network and a detection method. The detection network model comprises a feature extraction module based on a darknet53 network, a multi-size feature fusion module, a feature enhancement module and a boundary regression module. The detection method comprises the following specific steps: to-be-detected image into thefeature extraction module is input to obtain multi-size deep features of the image; a multi-size feature fusion module performs grouping fusion on the multi-size deep features; the feature expressiveforce of the fused features is further enhanced through a feature enhancement module; and acting a boundary regression module on the target object in the enhanced feature positioning graph. The position information of more small targets can be obtained and effectively combined with the semantic information of the small targets, so that the detection rate of the small targets is greatly increased.

Description

technical field [0001] The invention belongs to the field of computer vision, in particular to a small target detection network model and detection method based on darknet53 network. Background technique [0002] Object detection is a popular research direction in the field of computer vision, which is widely used in various fields such as face recognition, intelligent driving and intelligent monitoring. The task of target detection is to judge the existence or non-existence of the target object from the image and locate it. Due to the rapid development of artificial intelligence technology, object detection methods based on deep neural networks are superior to traditional object detection methods in terms of detection efficiency and accuracy. [0003] The target detection network model based on the deep neural network can be divided into two types: single-stage and double-stage. The single-stage network has become the first choice for industrial applications because of its...

AI Technical Summary

Problems solved by technology

[0003] The target detection network model based on the deep neural network can be divided into two types: single-stage and double-stage. The single-stage network has become the first choice for industrial applications because of its light weight, high efficiency and detection accuracy not weaker than the two-stage network. But it can only detect fixed-size images, that is, there is a large limit to the size of the input image

[0004] Compared with large targets, small targets occupy fewer pixels in the image, and the texture and edge features are not obvious. Due to the limitation of the single-order network on the image size, the characteristics of small targets are further amplified, resulting in a single-order network for small target detection. The detection rate is low

Images