Human body key point detection method, device and system and computer readable medium

Abstract

The invention provides a human body key point detection method, device and system and a computer readable medium, human body image data are preprocessed through polar coordinate transformation, a detection model used for detecting human body key points of a polar coordinate image is obtained through training on the basis of a training set formed by human body key point coordinates of the polar coordinate image, and in actual detection, the human body key point detection accuracy is improved. According to the human body key point detection method, a human face detector is used for detecting a human face detection frame, polar coordinate conversion is carried out on an image, human body key points of a polar coordinate image are obtained through a detection model, and finally the human body key points are converted into a rectangular coordinate system, so that the precision of a human body key point detection model is improved. The human body image data is preprocessed by adopting polar coordinate transformation, so that the training difficulty is reduced; and on the other hand, as the number of face key points is much larger than that of body key points, the proportion of the face in the image can be increased through transformation, the face key points are predicted more accurately, and the overall performance of human body key point detection is improved.

Description

technical field [0001] The present invention relates to the technical field of image processing, in particular to face detection and recognition, and in particular to a method, device, system and computer-readable medium for detecting key points of a human body based on polar coordinate transformation. Background technique [0002] Human target recognition is an important application based on computer vision technology. Through the images collected by the front-end camera, target recognition processing is performed on the edge device or background server, so as to perform strategic processing such as people flow counting and control, or based on target recognition. Follow-up applications such as action prediction and animation production. [0003] The process of human object recognition processing includes object detection and object localization. Target detection refers to extracting the approximate position of the human body from the collected image, and cropping the huma...

AI Technical Summary

Problems solved by technology

However, in the human body image, the face accounts for a very small proportion of the whole human body image, which leads to inaccurate prediction results of the model directly trained according to the key points of the face and body key points to predict the key points of the face, while the face The number of key points of the body is often large, and the number of body key points is small, so it will affect the overall accuracy of human body key point detection

Another method is to detect the position of the human body and face separately, intercept the images of the human body and face, and then detect the key points of the limbs and face respectively. Although this method has high accuracy, it requires multiple model predictions, and the calculation is time-consuming.

Images