Image positioning method based on data modal missing and Embedded Net

Abstract

The invention discloses an image positioning method based on data modal missing and Embedded Net, which is composed of two stages of off-line modeling and on-line positioning, and comprises the following steps: S1, collecting training images, preprocessing the training images, carrying out category division, and taking the training images as a training data set; S2, extracting features of the training images and fusing with the features; performing offline classification learning to obtain a feature extraction network, a feature fusion network and a position classification model of optimal parameters; S3, collecting a target image, and performing image preprocessing on the target image; and S4, performing feature extraction and feature fusion on each target image, and then obtaining a position estimation value of the target. The multi-image measurement technology and the machine learning technology are combined, target positioning for the indoor complex environment is achieved, follow-up positioning work is conducted through the images collected by the camera, the problem of signal attenuation is solved, and the implementation cost of the scheme is reduced.

Description

technical field [0001] The invention relates to an indoor target positioning method, in particular to an image positioning method based on data mode loss and EmbraceNet, and belongs to the field of positioning navigation technology and machine learning applications. Background technique [0002] In recent years, with the development and popularization of computer communication technology and mobile smart terminals, people are increasingly relying on their electronic devices to obtain their own physical location information, thereby satisfying various service requirements based on location information. [0003] At this stage, the development of positioning technology for outdoor environments is very mature, and various satellite navigation systems are widely used, such as the Beidou satellite navigation system independently developed by my country, GPS in the United States, GLONASS in Russia, and GALILEO in Europe. Although the above-mentioned satellite navigation systems gen...

AI Technical Summary

Problems solved by technology

For the lack of multimodal data that is common in most current deep learning network models, the common solution is to reuse the obtained values, use default values ​​or use interpolation methods to fill in missing data, but such When the processing method is applied to indoor positioning, it is difficult to ensure the robustness of the positioning performance of the solution

Images