Large-scale picture infringement detection method and system based on feature selective aggregation

Abstract

The invention provides a large-scale picture infringement detection method and system based on feature selective aggregation, and belongs to the field of picture infringement detection. According to the method, a deep residual convolutional neural network pre-trained on a large-scale picture data set is used for extracting high-level semantic feature maps of pictures, a selective feature aggregation algorithm SCDA is used for selectively aggregating the feature maps, and a principal component analysis PCA is used for compressing features obtained through aggregation into a low-dimensional dense feature vector for representation. And an index is constructed for the feature vectors in a product quantization mode so as to realize large-scale picture infringement detection. The method is high in efficiency, and millisecond-level searching in million-level vectors can be achieved.

Description

technical field [0001] The invention relates to the field of image infringement detection, in particular to a large-scale image infringement detection method and system based on feature selective aggregation. Background technique [0002] With the rapid development of information technology, the amount of information on the Internet is increasing day by day. As one of the main information carriers, pictures are easily tampered with and stolen by others after being spread on the Internet, resulting in infringement. Due to the huge scale of pictures on the Internet, it is not feasible to rely on manual identification of infringing pictures. It is necessary to actively detect infringements on the Internet through automatic picture infringement detection methods. Image infringement detection relies on the method of Near-duplicate Image Detection (NDID). For homologous pictures, there are two definitions of categories. In a narrow sense, that is, the picture obtained by the sam...

AI Technical Summary

Problems solved by technology

[0008] Aiming at the shortcomings of traditional image infringement detection methods, such as low detection accuracy and slow detection speed, the present invention proposes a large-scale image infringement retrieval method based on feature selective aggregation, which utilizes the depth The residual convolutional neural network (Residual neural network, ResNet) extracts the high-level semantic feature map of the picture, uses the selective feature aggregation algorithm SCDA (Selective Convolutional Descriptor Aggregation) to selectively aggregate the feature map, and uses the principal component analysis method PCA ( Principal components analysis) compresses the aggregated features into a low-dimensional dense feature vector representation, and uses product quantization (Product quantization) to construct an index for the feature vector to achieve large-scale image infringement detection

Images