Moving object detection method based on long-term video sequence background modeling framework

Abstract

The present invention provides a method for detecting a moving object based on a long-time video sequence background modeling framework, which includes the following steps: (a) obtaining a long-time panoramic video through a fixed panoramic camera; (b) performing a long-time panoramic video on the long-time panoramic video Decoding processing, and reasonably describe the processed long-time panoramic video through a stacked self-encoding neural network to obtain multiple background descriptors; (c) organize the background descriptors by clustering to obtain typical background descriptors, and Organize the typical background descriptors in the form of a hierarchical tree to obtain the background word bag; (d) use the ViBe background modeling method to fuse the background word bag to obtain the background modeling framework, and then obtain the background model, and then use the background The model detects moving objects.

Description

technical field [0001] The invention relates to computer vision technology and machine learning technology, in particular to a moving target detection method based on a long-time video sequence background modeling framework. Background technique [0002] In the field of video surveillance, cameras are more and more widely used. An important issue in video surveillance is how to obtain moving objects through background modeling. [0003] Existing background models are mainly divided into models based on time domain information and models based on fusion of time and space domain information. Models based on time domain information usually use the statistical characteristics of pixels in a short period of time in the past to predict the short-term future state of the pixel; models based on time domain information fusion also pay attention to the distribution characteristics of pixels in the spatial domain while using time domain information . [0004] The background differen...

AI Technical Summary

Problems solved by technology

However, since the scene is changing, the background model will be inaccurate, resulting in the inability to correctly obtain the moving target.

Among them, the sudden change of scene light is a difficult problem. If it is not handled well, the accuracy of moving target detection will be greatly reduced, which will bring difficulties to subsequent video processing algorithms.

Images