Hierarchical feature fusion method for multi-target detection of mobile robot

Abstract

The invention relates to the field of environmental perception of mobile robots, in particular to a hierarchical feature fusion method for multi-target detection of a mobile robot, and aims to improvethe detection capability of a target detection algorithm for targets of different scales so as to improve the environmental perception capability of an intelligent robot. The method includes: inputting the images in the data set into a pre-trained improved VGG-16, and preliminarily obtaining a feature map; respectively inputting the preliminarily acquired feature maps into a cavity convolution pyramid structure which comprises three cavity convolution integral branches with different expansion rates and is used for matching targets with different scales acquired by a visual sensor when the robot moves; fusing the feature maps acquired by different branches in a layered superposition mode provided by the invention, so that all channels in the feature maps contain feature information of different scales; gradually convolution is carried out on the fused feature maps to obtain feature maps of different sizes; and finally, obtaining the category and the bounding box of the to-be-detectedobject.

Description

technical field [0001] The invention relates to the field of environment perception of mobile robots, in particular to a layered feature fusion method for multi-target detection of mobile robots. Background technique [0002] With the continuous expansion of the application range of intelligent robots in the home environment, people have put forward higher and higher requirements for the robot's environmental perception ability. In the process of the robot searching for objects, since there are often objects of different scales in the robot's visual sensor, and the existing target detection algorithms cannot detect these objects well, it is necessary to improve the target detection algorithm's ability to detect objects of different scales. Detection ability, thereby improving the environment perception ability of intelligent robots. [0003] In order to enhance the detection effect of the network on targets of different scales, many scholars have improved the two-stage targ...

AI Technical Summary

Problems solved by technology

[0004] Aiming at the problem that the existing technology only uses a single-scale convolution kernel to extract features that are not rich, and the detection ability of objects of different scales in the same scene is low, the present invention uses dilated convolutions with different expansion rates to simulate receptive fields of different sizes, thereby Features of different scales are extracted, and at the same time, a hierarchical feature fusion method is proposed to fuse features of different scales. Compared with the channel splicing methods of other methods, the hierarchical feature fusion method of the present invention makes different channels contain different The feature information of the scale can effectively improve the detection ability of the target detection algorithm for objects of different scales in the same scene, thereby improving the efficiency of intelligent robots to search for objects

Images