Image Feature Query Method Combining Gravity Sensor and Image Feature Point Angle

Abstract

The invention discloses an image feature query method combining a gravity sensor and an image feature point angle. A camera carrying a gravity sensor is used to collect a target image to obtain an image sample with gravity information. The query image and the image sample are both processed as follows , to form an overall compact vector representation of the image sample: extract the feature points of the image that contain gravity information; increase the angle attribute value in all the extracted feature points by an angle calculated by the formula for rotation, and all the points in the rotated image Carry out bag-of-words model training on the image feature points of the rotated image to obtain multiple cluster centers; perform feature encoding on all image feature points in the rotated image, map the feature points to each cluster center, and partition according to the angle of the feature points , feature aggregation is performed on each partition to form an overall compact vector representation. Use the nearest neighbor search method to find the image samples that best match the query image.

Description

technical field [0001] The invention belongs to the technical field of computers and relates to a local feature aggregation method combining gravity sensors and image feature point angles. Background technique [0002] As an important research content of current mass image retrieval, local feature aggregation is a technology for aggregating and compressing encoded image feature points. for compact representation. The current mainstream mass image retrieval algorithms all use the bag-of-words model and its variant algorithms, that is, the image feature points are mapped into visual words through feature coding to represent them. The visual words are usually obtained by K-means clustering and other methods, while the overall representation of the image The local feature aggregation technology is used to accumulate and aggregate the feature points mapped into visual words. The quality of the local feature aggregation method directly affects the image retrieval accuracy. [0...

AI Technical Summary

Problems solved by technology

On the other hand, it is impossible to rely on the exhaustive algorithm to accurately predict the actual rotation of the image.

In addition, the method of taking the minimum distance will reduce the penalty for dissimilar images, which will affect the retrieval accuracy

Images