Indoor typical scene matching and positioning method based on neural network

Abstract

The invention provides an indoor typical scene matching and positioning method based on neural network. The method comprises the steps of step 1, establishing a standard typical scene positioning image library at a server side; step 2, adopting a Siamese deep neural network model, through training of a large amount of data, enabling the neural network to learn from the data to judge similarity measurement; step 3, outputting a feature vector by the deep neural network, calculating the similarity with a standard typical scene image library by utilizing the feature vector, judging the matching degree of the typical scene according to the similarity, and evaluating the quality of the model; and step 4, loading the trained model into a server, obtaining video data, sending the video data intothe trained deep neural network in the server, calculating similarity, and judging the current position. The method has the advantages of being high in training efficiency, high in convergence, high in modeling precision, good in matching effect, capable of meeting complex environments and the like, and matching and positioning of equipment in online indoor typical scenes can be accurately and efficiently achieved.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a neural network-based indoor typical scene matching and positioning method. Background technique [0002] With the advancement of science and technology and the improvement of people's economic strength, people pay more and more attention to location positioning services. At present, the outdoor positioning system is very mature. Outdoor positioning systems such as GPS systems cannot effectively locate. Existing indoor positioning methods based on the iBeacon Bluetooth module and methods using Wi-Fi positioning technology are easily affected by factors such as low positioning accuracy of the positioning method itself, building occlusion and other factors, and cannot accurately locate the user's current location. [0003] The vision-based positioning technology emerging today has attracted widespread attention because of its simple equipment and small impact factors. Since the ca...

AI Technical Summary

Problems solved by technology

[0004] In the vision-based indoor positioning method, image matching technology is the most important technical link. Traditional image matching technology (such as histogram, SIFT algorithm) has been difficult to meet the current requirements of large data volume and complex environment.

Images