Pedestrian feature mask generation method based on local key points

• Main Authors: HUANG ZENGXI, QIN YUSONG
• Format: Patent
• Language: Chinese; English
• Published: 18.08.2023
• Subjects: CALCULATING; COMPUTING; COUNTING; PHYSICS

The invention provides a pedestrian feature mask generation method based on local key points, and relates to the technical field of image processing, and the method comprises the steps: detecting key points of a target pedestrian, obtaining a position probability thermodynamic diagram, coordinates and confidence, and estimating attitude information; key point feature mask areas are set, and feature mask areas between the key points are set according to the human physiological topological structure and the position relation between the key points; initializing a mask value of the feature mask region according to the confidence of the key points, and initializing a mask value of the feature mask region between the key points according to the confidence of the key points related to the topology; if the key points of other pedestrians exist near the feature mask region, reducing the mask value of the feature mask region in combination with the position probability thermodynamic diagrams, coordinates and confidenc