Machine vision indoor positioning method based on improved convolutional neural network structure

Abstract

The invention discloses a machine vision indoor positioning method based on an improved convolutional neural network structure and mainly provides an improved convolutional neural network structure and a neural network model training method for the structure. And finally, the input video images are classified through the trained convolutional neural network to obtain indoor position of a mobile robot equipped with a RGB camera. The convolutional neural network function comprises the steps of extracting position features of the semantic segmentation image and the RGB image, and determining thereal-time indoor position of the mobile robot by utilizing the two types of position features. The improved convolutional neural network structure is a product of combination of U-Net, front 13 layersof two VGG16Net and rear 3 layers of one VGG16Net, and the convolutional neural network is composed of five parts of U1, VGG2, VGG3, VGG4 and an ArcFace classifier. According to the invention, real-time positioning of the indoor position of the mobile robot can be accurately realized.

Description

technical field [0001] The present invention relates to the technical field of artificial intelligence, deep learning algorithm research and image processing algorithm research, in particular to a machine vision indoor positioning method based on an improved convolutional neural network structure. Background technique [0002] With the increasing development of artificial intelligence technology, various types of robots have been widely used in all walks of life. In the application process of mobile robots, real-time detection and monitoring of the robot's position is a prerequisite for better service to humans. Therefore, mobile Robot wireless positioning technology has gradually become a research hotspot. In the outdoor environment, the global positioning system based on the mobile signal, the Beidou navigation system and the cellular positioning technology can meet most of the positioning requirements, but these methods are not suitable for the positioning of the indoor e...

AI Technical Summary

Problems solved by technology

Traditional visual positioning methods usually use image matching methods. However, this type of method is susceptible to shallow features such as shooting angles, lighting changes, and content changes in non-fixed building outlines in the scene.

With the popularization of deep learning technology, many scholars use deep convolutional neural network to extract deep position features of images. The deeper the convolutional neural network, the better the feature extraction effect. The number of training samples also increases geometrically, and in many application scenarios, it is extremely difficult to label training samples

Images