A multi-instance multi-label learning method for safe city video surveillance applications

Abstract

The invention discloses a multi-instance and multi-label learning method oriented to safe city video monitoring applications. The invention obtains multi-instance and multi-label data sets of safe city video surveillance, and excavates the internal relationship between these multi-instance data and multi-label data. It is possible to predict new video surveillance, thereby judging the potential public security and traffic conditions hidden in the area where the new video surveillance is located; the present invention mainly makes two contributions. There are many types of problems, and the goal of retaining the integrity of multi-labels without losing the associated information between labels is achieved. The second is the first time that the convolutional neural network is introduced into the safe city video surveillance network, taking advantage of the advantages of the convolutional neural network. The correlation between examples is sufficiently deep learned, and the information between examples is fully mined.

Description

technical field [0001] The invention belongs to the technical field of computer science and multi-instance multi-label learning, and relates to a multi-instance multi-label learning method oriented to safe city video surveillance applications. Background technique [0002] Building a safe city is the primary goal of building a harmonious society. The improvement of urban traffic and public security management is the top priority of building a safe city. Currently, there are still many problems in building a safe city, and there are still many areas that can be improved, such as video surveillance network . Today's urban video surveillance network has become an important tool for urban management. However, many video data are unlabeled, scattered and chaotic. Managers cannot know from these data which parts of the city need traffic relief and which places need to be rectified. The information obtained by data mining cannot be obtained. Managers cannot obtain the areas that ...

AI Technical Summary

Problems solved by technology

[0007] Aiming at the problems that the traditional multi-instance and multi-label learning algorithm is difficult to learn the internal relationship between high-level features and the amount of calculation is too large, the present invention provides a multi-instance and multi-label learning method for video surveillance applications in safe cities

Images