Camouflage target detection and identification method based on deep neural network

Abstract

The invention provides an end-to-end camouflage target detection and identification method based on a deep neural network. According to the method, a 'segmentation-recognition' double-path neural network is designed. According to the segmentation path, overall image information of a target is utilized, and a Receptive Field Block is added by taking an anti-attention convolution module as a main body so as to ensure that the network obtains a larger receptive field; a U-net network structure is adopted in the overall design, and depth visual features such as colors and textures of a camouflage target can be captured more effectively and accurately. The recognition path adopts a double-branch Resnet structure, and target pixel information is converted into semantic information so as to identify a specific target type. Experimental results show that the method can effectively solve the problem that a conventional target detection and identification method is difficult to detect and completely segment the camouflage target, and can effectively separate and identify multiple camouflage targets such as animals and plants, human bodies and military facilities in multiple complex environments such as ocean, jungle, snowfield and desert.

Description

technical field [0001] The invention relates to the technical field of target detection and computer vision, in particular to a method for detecting and identifying a camouflaged target based on a deep neural network. Background technique [0002] Camouflage refers to the way or means that things use various methods to hide the truth and show the falsehood in order to achieve a certain purpose. Camouflage technology widely exists in nature and human society, such as various animals and plants adopt camouflage to capture food or avoid natural enemies; modern military widely uses camouflage technology as a very important anti-reconnaissance method to provide false information and confuse opponents the goal of. According to the types of camouflage characteristics, camouflage types can be divided into: similar color and texture camouflage, false target camouflage, smoke camouflage, cover camouflage, etc. At present, the mainstream detection and processing methods of camouflage...

AI Technical Summary

Problems solved by technology

[0005] Constrained by the image background of the camouflaged target and the complexity of the target information, the existing neural network structure is still difficult to meet the high-precision segmentation requirements, and there will be problems such as incomplete segmentation or the segmentation area does not match the actual area.

In addition, most of the existing camouflage target detection models focus on the segmentation of camouflage target images, and there are relatively few models for camouflage target category recognition.

Images