Potential sudden criminal person identification method based on assumed evaluation and Bayesian learning

Abstract

The invention discloses a potential sudden criminal person identification method based on assumed evaluation and Bayesian learning, and particularly relates to the technical field of image identification. Comprising the following steps: redefining an original problem, setting a key assumption, initializing a likelihood probability and a prior probability required in a Bayesian learning model, constructing and training the Bayesian learning model, and synchronizing related information to other platforms or components in the system when a posterior probability is greater than a set threshold value. Or synchronizing the data as intelligence information to an internal network, using the data to train, verify and test a supervised learning model, using the constructed supervised learning model as an auxiliary study and judgment model, and using the model constructed according to key assumption evaluation and Bayesian learning at the same time. According to the method, the key assumption evaluation and Bayesian learning are combined to identify potential people intending to suddenly crime, and the method is further combined with a supervised learning model to solve the problems of early warning and research and judgment of such emergencies to a certain extent.

Description

technical field [0001] The present invention relates to the technical field of image recognition, and more particularly, the present invention relates to a method for identifying potential sudden criminals based on hypothesis evaluation and Bayesian learning. Background technique [0002] Sensor devices represented by cameras are widely used in real-time security monitoring scenarios. Thanks to the rapid development of artificial intelligence technology, many researchers in academia and industry have used them in smart cities and smart security scenarios. Cloud computing can extract many features in sensor stream data (such as camera video stream) in real time, and at the same time synchronize alarm information, captured images and extracted features to the background data center according to the business direction. At the same time, it is difficult to identify potential persons who intend to commit crimes in public service scenarios such as gas stations, schools, power plan...

AI Technical Summary

Problems solved by technology

At the same time, it is difficult to identify potential personnel who intend to commit sudden crimes in public service scenarios such as gas stations, schools, power plants, and water plants, especially for those who have no criminal record or do not have a sensitive intranet blacklist in the landing scenario Under the circumstances, it is impossible to use the current technical means to prevent in advance

[0003] The existing technology has the following deficiencies: existing intelligent security products output structured feature data extracted from video stream data, captured images, alarm information, etc. into the data, such as Kafka, ElasticSearch, MySQL, etc. source of data, but this part of high-value data is usually ignored, without further analysis and modeling based on professional business direction

Images