Image retrieval method based on robot vision platform

Abstract

The invention discloses an image retrieval method based on a robot vision platform. The image retrieval method based on the robot vision platform includes: using a binocular camera to collect an image sample, identifying content of the image sample, decomposing an image after being identified according to an image decomposition rule, and obtaining image fragments; sending the image fragments in a tabular form to a retrieval user, and enabling the retrieval user to mark the image fragments; classifying the image fragments after being marked according to a support vector machine model, finding out an image fragment set boundary, removing a part of the image fragments, close to the image fragment set boundary, according to an indexing rule, performing similarity retrieval on the residual image fragments after the image fragments close to the image fragment set boundary are removed, and obtaining an image material highest in similarity. The image retrieval method based on the robot vision platform uses binocular distance measurement to obtain distance features of a space target, improves image decomposition accuracy and objectivity, and improves efficiency and accuracy of the similarity retrieval of the image fragments by establishing the image decomposition rule and the indexing rule.

Description

technical field [0001] The invention relates to the field of image recognition, in particular to an image retrieval method based on a robot vision platform. Background technique [0002] The robot vision platform, RobotEye for short, is a robot vision retrieval system that collects visual data from robots or smart devices and feeds back valuable knowledge information to robots or smart devices. The current image recognition methods, such as relying on keyword search, comparing graphic feature parameters, and so on. Among them, searching by keywords and identifying by human eyes are subjective and cannot accurately reflect image information. The method of comparing graphic feature parameters also has problems such as excessive calculation, too many parameters, and cumbersome implementation. However, a certain characteristic of a single contrasting figure cannot fully and concretely describe the image information. Usually, the existing image retrieval methods do not have a ...

AI Technical Summary

Problems solved by technology

Among them, the use of keywords to search and the recognition of human eyes are subjective and cannot accurately reflect the image information.

The method of comparing graphic feature parameters has problems such as excessive calculation, too many parameters, and cumbersome implementation.

And a certain feature of a single comparison graphic cannot fully describe the image information specifically.

Usually, the existing image retrieval methods do not have a strong ability to decompose image content, intelligent knowledge association ability, and agile self-learning ability, and when collecting images, use ordinary cameras to obtain single-frame image samples. The ability to capture continuous images and multiple parallax images is not strong, which greatly reduces the measurement accuracy of the target

To sum up, none of the current image recognition methods can solve the retrieval problems of accuracy, simplicity, decomposition, association, and self-learning of images.

Images