34results about How to "Purify the network environment" patented technology

The invention discloses an image retrieval system and an image retrieval method based on multi-feature and sparse representation. The system comprises a feature extraction module, a feature dictionary building module, a similarity measurement module, an information storage module and an inquiry interaction module, wherein the feature extraction module adopts shape and color combining image features, color enhanced Gaussian Laplace features (CLOG features) and SURF features; the feature dictionary building module compresses original features into overcomplete overcomplete features through an on-line dictionary learning algorithm, and overcomes the defect that the original features are too dense; the similarity measurement module introduces a sparse representation theory, compares the residual error size generated by original dictionary and relevant dictionary representation, and judges the similarity of the two images, and the problem of higher feature dependence of the traditional similarity measurement method is solved. The system and the method provided by the invention have the advantages that rotating, noise and illumination change images can be effectively retrieved, and the image retrieval robustness is obviously improved.

The embodiment of the invention provides a label user identification method and related equipment, which are used for quickly identifying a label user in a live broadcast platform, purifying a networkenvironment and providing user experience in the platform. The method comprises the following steps: constructing a target directed graph, wherein the target directed graph indicates an information interaction relationship between any two users in a live broadcast platform; calculating the weight of an edge, connected with a vertex corresponding to a target user, in the target directed graph, wherein the target user is a user with a to-be-identified tag; determining an initial label score of a user corresponding to each vertex in the target directed graph; calculating the label score of the target user according to the initial label score and the weight iteration of the edge connected with the vertex corresponding to the target user; judging whether the label score of the target user is greater than a preset threshold value or not when a preset iteration termination condition is met; and if yes, determining that the target user is a label user.

The invention discloses a method for intelligently examining self-media audio and video contents. The method comprises the following steps: step S10, detecting a format and a code used by a self-media file uploaded by a user; s20, identifying the audio stream, and converting the audio stream into a subtitle text; step S30: according to a preset sensitive word library, carrying out sensitive word detection on the subtitle text; step S40, carrying out real-time detection on the audio stream, and detecting whether an audio time point of abnormal sound exists or not; step S50, intervening the audio stream at the target position by using a preset silencing mode, and generating an audio file at the same time; step S60, performing voice recognition on the temporary audio file and converting the temporary audio file into text information; s70, checking whether sensitive vocabularies exist in the character information or not; step S80, synchronizing the audio frame and the video frame after examination intervention; and S90, returning a detection result. According to the method, the sensitive words in the target live broadcast data stream are intelligently identified, so that the sensitive words are extracted and replaced, and the network environment is favorably purified.