Research And Application Of Pattern Recognition For Data Aggregation

With the popularity of the mobile Internet and social media,compared to text data,expression via image has become more and more popular among the public.In the composition of network security incidents,the proportion of issues related to network information content security and public sentiment is also increasing year by year.This puts forward new requirements for data aggregation technology,which is the basis of network situational awareness.Image data has high-level semantic features,which makes it a big gap in understanding with traditional log-type data.This research uses pattern recognition technology to solve the problem of image content perception and sentiment classification in image data aggregation,and relies on the recognition results to achieve data aggregation of image data.The main work of this article is as follows:1.Aiming at the problem of image content perception,we use deep learning-based object detection technology.Based on the SSD object detection algorithm,we propose an improved object detection algorithm ACSD from three aspects:hyperparameter setting,network structure,and training data.The ACSD algorithm improves the detection accuracy while maintaining a high algorithm performance,and to a certain extent,improves the generalization ability of the model and suppresses overfitting.2.Aiming at the problem of image sentiment perception,an image sentiment classification scheme based on sentiment regions is adopted.By combining with the above-mentioned object detection algorithm,a sentiment region proposal network is proposed,and feature map and ROI pooling are used to share features with the sentiment classification network for the full image.While ensuring recognition accuracy,it reduces repeated calculations,implements end-to-end inference,and improves algorithm performance and execution efficiency.3.Based on the first two technologies,an internet image data aggregation system is designed and implemented.Contains four modules:data collection,content awareness,emotion perception,and system management.Realized a series of functions from the Internet image data collection to aggregation,and completed the system test.This paper proposes a solution that uses pattern recognition technology to solve the problem of image data understanding and classification aggregation in data aggregation,improves the object detection and image sentiment classification algorithms involved,and designs and implements an image data aggregation prototype system.It provides a data foundation for subsequent data fusion and situational awareness decisions.