Human face forehead area detection and positioning method and system of low-resolution thermodynamic diagram

Abstract

The invention discloses a face forehead area detection and positioning method for a low-resolution thermodynamic diagram, and the method comprises the following steps: removing no-face and blurred images in thermodynamic images collected by an infrared camera, obtaining an effective thermodynamic image set, dividing the thermodynamic image set into a training set and a test set, and marking corresponding labels; performing data enhancement processing on the training set with the label and the test set with the label; combining a bidirectional multi-scale feature and a rapid normalization fusion method to form a multi-scale feature fusion network TwFPN; solving a DEfficientNet human face forehead detection model based on the multi-scale feature fusion network TwFPN with the weight and a joint scaling algorithm; inputting the training set with the label into a DEfficientNet human face forehead detection model, extracting an optimal forehead region feature, and obtaining an optimal humanface forehead detection model; inputting the test set with the label into an optimal human face forehead detection model to obtain a forehead area of the human face, and adding a detection box to theforehead area.

Description

Technical field [0001] The invention relates to the technical field of face target detection, in particular to a method and system for detecting and locating the forehead region of a human face with low resolution heat map. Background technique [0002] With the progress and development of society, face detection technology plays a very critical role in the field of public safety. For example, face detection algorithms can detect people with high fever in videos; in addition, face recognition technology can be used in video surveillance Real-time detection of suspicious criminals to help relevant personnel to report to the police and early warning in time. Therefore, many scientific researchers spend a lot of time and cost to develop a real-time and efficient face detection algorithm to meet different functional needs. At present, in actual scenes, face detection mainly has the following difficulties which lead to a low recognition rate of face detection: [0003] First, the came...

AI Technical Summary

Problems solved by technology

At present, in actual scenarios, face detection mainly has the following difficulties, which lead to a low recognition rate of face detection:

[0003] First, the position of the camera is generally relatively high, and there is a certain angle between the pedestrian and the pedestrian, so it is difficult to collect the front face image;

[0004] Second, as the number of people increases, mutual occlusion or obstruction between pedestrians cannot obtain clear and complete images of facial features (eyes, nose, mouth, etc.);

[0005] Third, the resolution of pedestrians is related to the position of the camera, the farther away from the camera the resolution of pedestrians is smaller;

[0006] Fourth, visible light has a greater impact on illumination changes, visibility at night, and other harsh environments. The parallax between face pictures and real pictures is large, so it is impossible to obtain information in real scenes, especially in low-resolution situations. Harder to tell apart images of faces

This process is too complicated, which increases the complexity of the model and makes the algorithm inefficient;

[0010] Third, because the resolution of different images has different effects on the detection results of face key

Images