Remote sensing image vehicle target recognition model and method based on deep neural network

Abstract

The invention provides a remote sensing image vehicle target recognition model and method based on a deep neural network. According to the model, convolution blocks in a PANet network in a Yolov4 network structure are replaced by inverted residual modules, specifically, two layers of convolution blocks sequentially connected from bottom to top in an up-sampling unit in the PANet network are replaced by the inverted residual modules, and the two layers of convolution blocks are defined as a first inverted residual module and a second inverted residual module respectively; in the PANet network, two layers of convolution blocks which are sequentially connected from top to bottom by a down-sampling unit are replaced by inverted residual modules and are respectively defined as a third inverted residual module and a fourth inverted residual module; an inverted residual error module is newly added between the second inverted residual error module and the third inverted residual error module and is defined as a fifth inverted residual error module, and the fifth inverted residual error module is used for down-sampling the input image. The method is more accurate in vehicle target recognition in complex environments such as shadow shielding.

Description

technical field [0001] The invention relates to the technical field of remote sensing image target recognition, in particular to a vehicle detection and recognition method in remote sensing images of street scenes in complex environments, and in particular to a remote sensing image vehicle target recognition model and method based on a deep neural network. Background technique [0002] In aerospace remote sensing image recognition tasks, vehicle recognition, as an important means of transportation, has always been a difficult and hot research point. Vehicle target detection based on remote sensing images plays a very important role in both military applications and civilian applications. However, in remote sensing images, problems such as small vehicle target pixels, complex surrounding environment, poor semantic information, buildings or shadow occlusions lead to unsatisfactory recognition effects of existing frameworks on cars. [0003] At present, there are many research...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that the traditional target recognition method has a poor recognition effect in the face of remote sensing image recognition of small targets such as vehicles in a complex environment, the present invention provides a remote sensing image vehicle target recognition model and method based on a deep neural network

Images