Precise image searching method based on region convolutional neural network

Abstract

The invention provides a precise image searching method based on a region convolutional neural network. By use of the convolutional neural network, contents of an image are analyzed, and regions and the number of objects included by the image are calibrated. By inputting searching conditions, the image including the objects can be precisely output by traversing the whole image database. The region convolutional neural network is capable of analyzing all objects in the image and the convolutional neural network is quite high in accuracy. Compared with the traditional method, the situation of omission or mistaken searching is avoided to the great extent and an objective of precise searching is achieved.

Description

technical field [0001] The invention relates to the fields of image processing and machine learning, in particular to an accurate image retrieval method based on a regional convolutional neural network. Background technique [0002] In the field of image retrieval, there are generally two ways of retrieval through image features and image tags. [0003] Through the image feature retrieval method, extract the features of the input image and the database image, compare the image similarity, such as color similarity, texture similarity, scale similarity, filling similarity, etc., to find the image with the highest similarity, but subject to Due to the complexity of the image itself and the influence of the background pattern of the image, these features will not achieve very accurate results. [0004] Through the method of image tag retrieval, directly traverse the image tags in the database, compare the keywords, and find out the images whose keywords meet the retrieval condi...

AI Technical Summary

Problems solved by technology

[0003] Through the image feature retrieval method, extract the features of the input image and the database image, compare the image similarity, such as color similarity, texture similarity, scale similarity, filling similarity, etc., to find the image with the highest similarity, but subject to Due to the complexity of the image itself and the influence of the background pattern of the image, these features will not achieve very accurate results

[0004] Through the method of image tag retrieval, directly traverse the image tags in the database, compare the keywords, and find out the images whose keywords meet the retrieval conditions, but this method must manually add tags to the entire database, the workload is very large, and for new If you cannot add tags in time to the images that come in, this part of the images will not be retrieved, which is not conducive to the update of the system

[0005] The accuracy of existing image retrieval methods is low. One of the cases is that an image contains many different objects. The classification algorithm cannot clearly judge the category of the image. Even if the category of the image is judged, the number and location of the objects cannot be pointed out. , so the existing image retrieval methods cannot achieve the effect of accurate retrieval

Images