CN112257580A

CN112257580A - Human body key point positioning detection method based on deep learning

Info

Publication number: CN112257580A
Application number: CN202011134250.6A
Authority: CN
Inventors: 冯志珍; 张卫山; 于强
Original assignee: China University of Petroleum East China
Current assignee: China University of Petroleum East China
Priority date: 2020-10-21
Filing date: 2020-10-21
Publication date: 2021-01-22

Abstract

The invention provides a human body key point positioning detection method based on deep learning, which is based on a key point detection algorithm in the deep learning, provides a new quadratic regression detection method and accurately determines the position of a key point again by utilizing the structural characteristics of a human body. Based on the convolutional neural network and the cyclic neural network technology in deep learning, the human body key points in the picture are accurately positioned by constructing a key point detection neural network. When the key point positioning algorithm of the human body is designed and selected, the key characteristics of the human body contacted by the safety belt wearing specification are fully considered, the method mainly improves the precision of position detection of key points of 7 important bones at the shoulder, the chest, the waist, the thigh and the like, and adopts a mode of fusing a key point positioning detection module and a safety belt detection module to judge whether the safety belt wearing of a person meets the specification or not with high precision.

Description

Human body key point positioning detection method based on deep learning

Technical Field

The invention relates to machine learning, deep learning computer vision, graphic image processing and key point detection, in particular to a human body key point positioning detection method based on deep learning.

Background

The key points of the human skeleton are important for describing the human posture and predicting the human behavior. Therefore, human skeletal key point positioning detection is the basis of many computer vision tasks, such as motion classification, abnormal behavior detection, and automatic driving. Many human key points directly acquire three-dimensional information of the human key points through a Convolutional Neural Network (CNN), and the deep neural network can acquire three-dimensional human key point prediction values with reasonable precision from a single view. In recent years, with the development of deep learning technology, the detection effect of key points of human bones is continuously improved, and the method has started to be widely applied to the related field of computer vision. The invention combines the human body key point positioning technology to carry out the standard detection of the wearing of the human body safety belt.

The detection method of the human body key points is roughly divided into Bottom-up (Bottom-up) and Top-down (Top-down). The positions of all key points in the graph are firstly predicted from top to top, then all key points are associated with human instances, such as VGG-19 is used as a main network, image features are output, and then a heat map of the human key points and groups of the key points are generated through a multi-level network. From top to bottom, the method benefits from the breakthrough of an object detection algorithm, the picture can be divided into a plurality of sub-pictures only containing one person, and then the positions of all key points are detected.

Disclosure of Invention

In order to solve the defects and shortcomings in the prior art, the invention provides a human body key point positioning detection method based on deep learning, which is an efficient method for performing secondary accurate regression on key point positioning on the basis of the original computer vision human body key point detection and performing high-accuracy judgment on whether human body safety belt wearing meets the standard or not by adopting a mode of fusing a key point positioning detection module and a safety belt detection module.

The technical scheme of the invention is as follows:

a human body key point positioning detection method based on deep learning is characterized by comprising the steps of image preprocessing, generation of a confrontation network model, human body key point detection and safety belt detection. The method comprises the following steps:

step (1), intercepting a personnel safety belt data set from a video frame, and screening pictures meeting conditions;

step (2), separating three color channels of r, g and b by adopting data enhancement, adding noise, expanding a data set by using a method for generating a confrontation network model, and preprocessing the data set in the step (1);

step (3), marking and training a personnel key point position detection module by using the data in the step (2), and extracting 7 key point information of shoulders, chests, waists, thighs and the like;

step (4), marking the data of the human body safety belt by using a marking tool to obtain a corresponding target JSON file, extracting features by using a mask rcnn algorithm, and training a corresponding human body safety belt detection model;

and (5) detecting pictures, detecting key point information by using the key point detection module (3), connecting 7 key points, dividing the safety belt target by using the human safety belt detection module (4), comparing whether the part connected with the key points is positioned in the safety belt dividing target, and if the condition is met, determining that the part is the safety belt wearing standard, and returning a detection result.

The invention has the beneficial effects that:

(1) when the human body key point positioning algorithm is designed and selected, the human body key characteristics contacted by the safety belt wearing specification are fully considered, and the precision of position detection of 7 important skeleton key points such as shoulders, breasts, waists, thighs and the like can be improved;

(2) the invention carries out secondary accurate regression for key point positioning detection, and accurately determines the position of the key point by effectively utilizing the structural characteristics of the human body. The method can adapt to complex scenes of the power field, and the algorithm has more robustness.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.

FIG. 1 is a general flowchart of a human key point localization detection method based on deep learning according to the present invention;

fig. 2 is a schematic diagram of human body key points of the deep learning-based human body key point positioning detection method of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

As shown in FIG. 1, the method for detecting the localization of human key points based on deep learning of the present invention comprises the following basic steps: generating an confrontation network model, image preprocessing, a human body key point detection model, a safety belt detection model and a multi-model fusion judgment algorithm.

The following describes in detail the human body key point localization detection method based on deep learning with reference to fig. 1 and fig. 2:

The invention discloses a human body key point positioning detection method based on deep learning, which is based on a key point detection algorithm in the deep learning, provides a new quadratic regression detection method, and accurately positions human body key points in a picture by constructing a key point detection neural network based on a convolutional neural network and a cyclic neural network technology in the deep learning. The key point positioning technology is adopted to be fused with the detection of the safety belt, and whether the safety belt is worn normally is judged through whether the safety belt exists at the key point positions of 7 important bones of the shoulder, the chest, the waist, the thigh and the like. The key characteristics of the human body contacted by the safety belt wearing specification are fully considered in the design and selection of the key point positioning algorithm of the human body, so that an efficient algorithm for standard positioning of the key points of the human body is designed.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims

1. A human body key point positioning detection method based on deep learning is characterized by comprising the steps of image preprocessing, generation of a confrontation network model, human body key point detection and safety belt detection. The method comprises the following steps: