Posture recognition method and system of parallel fusion network based on urban smart sports

Abstract

According to the posture recognition method and device of the parallel fusion network based on the urban smart sports provided by the embodiment of the invention, the brand new Ousvalidan network is adopted to carry out feature extraction on the input preprocessed image, and feature extraction is carried out on the images with different resolutions from high to low; in the whole process, prediction is carried out after high-resolution and low-resolution features are fused. According to the method, multiple times of multi-scale fusion are carried out, so that each representation from high resolution to low resolution repeatedly receives information from other parallel representations, and abundant high-resolution representations are obtained. Therefore, the predicted key points are more accurate and more accurate in space.

Description

technical field [0001] The invention relates to the technical fields of big data and artificial intelligence, in particular to a gesture recognition method and system based on a parallel fusion network of urban smart sports. Background technique [0002] OpenPose is an open source library based on convolutional neural network and supervised learning and written in the framework of caffe. It can track people's facial expressions, torso, limbs and even fingers. It is not only suitable for single people but also for multiple people. robustness. It can be called the world's first real-time multi-person two-dimensional pose estimation based on deep learning. It is a milestone in human-computer interaction and provides a high-quality information dimension for machines to understand people. [0003] The traditional attitude estimation system is mainly based on the OpenPose algorithm to extract bone information for video monitoring, or acceleration sensor monitoring. However, the ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the problem that the existing detection schemes are not accurate enough to detect the similarity of sports behaviors and improve the matching between user actions and standard actions, the present invention provides a city-based The attitude recognition method and system of the parallel fusion network of smart sports, the method uses a new bone extraction network-Oswald network to enhance the recognition accuracy of small targets at high resolution, and the network can maintain in the whole process High-resolution and low-resolution features are fused for prediction, and multiple multi-scale fusions are performed, so that each high-resolution to low-resolution representation repeatedly receives information from other parallel representations, thereby obtaining rich high-resolution representations

Images