Statistical method based on target detection and foreground detection heat map

Abstract

The present invention provides a thermal map statistics method based on target detection and foreground detection, which specifically includes: using a Gaussian mixture model to model the monitoring environment and detecting the moving foreground; fusing the targets detected by deep learning and the Gaussian mixture model , and then accumulate to form a spatial heat map; use the target tracking algorithm to track the point targets in the heat map, and count the number of targets in a unit time to form a time heat map and other steps. Based on the target detection and foreground detection thermal map statistics method, the present invention combines the motion foreground with deep learning detection to avoid unnecessary motion interference on the heat map, and also uses the motion foreground information to remove the objects that have been detected by deep learning. Detects immovable targets, greatly improving the accuracy of moving target information.

Description

technical field [0001] The invention relates to a thermal map statistical method, in particular to a thermal map statistical method based on target detection and foreground detection. Background technique [0002] Heatmap uses different colors and color shades to represent data distribution information. It has good visualization effects and is widely used in statistical analysis of surveillance video scenes. In monitoring video scenes, heat maps are usually used to indicate the frequency of motion within the monitoring range. Most of the existing technologies use foreground extraction algorithms of video frames, such as background difference or Gaussian mixture model to extract moving objects, and then foreground motion Points are accumulated to form a heat map. However, in the existing technology, most of the heat map statistics are based on the moving foreground. Since the moving foreground is based on the direct difference between the foreground and the background to jud...

AI Technical Summary

Problems solved by technology

However, in the existing technology, most of the heat map statistics are based on the moving foreground. Since the moving foreground is based on the direct difference between the foreground and the background to judge whether it is a moving target, when the observed heat map is facing an object that moves frequently, Or in scenes with abnormal weather such as rain, object motion and weather motion will be counted as motion heat maps, which will interfere with important targets such as pedestrians who are actually more relevant, and more serious ones may cover up important moving target information

Images