Video moving object detection system, method and terminal based on deep fusion network

Abstract

The invention provides a video moving object detection system based on a deep fusion network, and the system comprises a video feature extraction module which receives video sequence input, carries out the feature extraction of video contents, obtains the feature expression of scene information in a video, i.e., the feature expression of a video scene, and sends the feature expression of the videoscene to a deep fusion module; a basic result detection module used for receiving video sequence input, detecting a moving object by utilizing a basic detector to obtain a corresponding basic detection result, and sending the corresponding basic detection result to the depth fusion module; and a deep fusion module used for receiving the video scene feature expression and the basic detection result, carrying out optimal fusion by utilizing a deep neural network and outputting a final detection result. The invention also provides a video moving object detection method and a terminal. Accordingto the invention, a high-accuracy detection result can be obtained.

Description

technical field [0001] The present invention relates to the technical field of video moving object detection, in particular to a video moving object detection system, method and terminal based on a deep fusion network. Background technique [0002] Video moving object detection can be used as the first link in video image processing and video content analysis, providing preliminary analysis results for subsequent operations, and helping to improve the performance of the entire video processing and analysis system, so video moving object detection is a crucial important technology. [0003] For the problem of video moving object detection, researchers have proposed a large number of methods. However, most of these research results are aimed at a certain or a certain type of specific scene, based on feature engineering, and using the method of manually designing operators for method design. These traditional methods are divided into statistical model-based, cluster-based, sp...

AI Technical Summary

Problems solved by technology

This method can achieve high-accuracy detection results, but requires manual intervention and cannot be completed automatically

[0005] The biggest difficulty in using deep learning to obtain a detection model is the lack of training data. Without enough labeled data, the neural network cannot be effectively trained.

At present, there is no description or report of the similar technology of the present invention, and no similar data at home and abroad have been collected yet.

Images