A view landmark retrieval method based on end-to-end deep learning

Abstract

The invention discloses a view landmark retrieval method based on end-to-end deep learning, and the method comprises the following steps: S1, collecting a key landmark image, carrying out the preprocessing operation, and enabling the key landmark image to serve as training data; S2, embedding a local aggregation descriptor feature vector method into the CNN to form an end-to-end CNN model; S3, inputting the collected training data into an end-to-end CNN model, extracting image local invariant features, training the CNN model through an error function, and learning an optimal aggregation cluster center point; S4, performing key frame picture extraction operation on the to-be-identified video stream, and performing down-sampling operation after the to-be-identified video stream and the to-be-identified picture stream are subjected to the down-sampling operation to generate a to-be-identified landmark data set Q; S5, inputting Q into the trained CNN model, performing local invariant feature vector extraction, and outputting a calculation result of each landmark category through a full connection layer and a data output layer; And S6, according to a key landmark category threshold value set by training, judging whether each piece of data in Q has a key landmark category or not, and if yes, outputting a picture source name and landmark prompt.

Description

technical field [0001] The invention relates to Internet new media view recognition technology, in particular to a view landmark retrieval method based on end-to-end deep learning. Background technique [0002] According to the China Internet Network Information Center (CNNIC) "China Internet Development Report (2018)", domestic Internet information flow, especially hot or sensitive events, is gradually moving away from text-based media. transfer. Especially on all kinds of emerging social media (mainly Weiwei and Yiji), it is difficult to quickly query, identify, and locate the location only by the text of the message, the text in the video or the picture. It is difficult for law enforcement agencies and government agencies at all levels to quickly locate the location of hot events through pictures and videos to analyze and control public opinion. If it can solve the needs of agencies at all levels for fast, high-precision and automatic video image landmark retrieval, and...

AI Technical Summary

Problems solved by technology

This process is equivalent to extracting local feature vectors through a black box, resulting in uncontrollable influence of parameter optimization on landmark instance recognition in the process of training the CNN network.

However, when the classical image feature extraction method is applied to the landmark instance search task, its algorithm performance after extracting local invariant features does not have the powerful nonlinear representation ability of deep learning, and cannot fully understand the deeper information of the image. Can not achieve good precision effect

Images