Human body posture detection method and system, storage medium, equipment and terminal

Abstract

The invention belongs to the technical field of human body posture detection, and discloses a human body posture detection method and system, a storage medium, equipment and a terminal.The human body posture detection method comprises the steps: constructing a data set of a human body skeleton model based on a depth image and a data set of human body posture classification; carrying out transfer learning on an openpose neural network model, taking the depth image as network input, and taking the obtained heatmap (articulation point coordinates) and paf (articulation point vector field) as labels for training to obtain a human skeleton model of the depth image; and performing posture feature included angle calculation on the labeled data through an posture analysis module, and inputting the included angle and a label into a random forest for training to obtain an posture recognition model. According to the method and system, the depth image is skeletonized on the basis of openpose and can be converted into a 3d skeleton to be observed by medical staff, and the accuracy is high.

Description

technical field [0001] The invention belongs to the technical field of human body posture detection, and in particular relates to a human body posture detection method, system, storage medium, equipment and terminal. Background technique [0002] At present, with the rapid development of my country's modern economic society and the advancement of scientific production technology, the quality of life of the people in our country is getting better and better compared with the past, and the level of hygiene and medical conditions have been greatly improved and improved. , the average life expectancy of my country's elderly population has been significantly extended, and the age structure of the population is gradually showing the development characteristics of population aging. Since my country officially entered the aging society in 1999, the number of elderly groups is growing rapidly and the number is large, making our country enter an aging society before realizing social mo...

AI Technical Summary

Problems solved by technology

[0013] (1) Existing technology costs too much, the detection of wearable devices is not accurate enough, and it is not comfortable enough to wear. At the same time, receiving it at home also affects the lives of children

[0014] (2) Most of the existing human posture detection technologies based on computer vision processing are implemented based on RGB images. Installing rgb cameras in the elderly’s homes to collect their daily actions is a bit of an invasion of privacy

[0015] (3) The traditional DTW algorithm has the problem of pathological mapping; the traditional image and video-based gesture recognition algorithm has problems such as high computational cost, many camera blind spots, and easy privacy leakage.

[0016] The difficulty in solving the above problems and defects is as follows: the current attitude detection models are based on experimental data, there is no real life data for model training, and the false detection rate is high; the recognition degree of depth images is lower than that of color images, and the migration of models difficult to learn

Images