Transformer-based glass identification method

Abstract

The invention discloses a transformer-based glass identification method, and belongs to the technical field of image identification. The method comprises the steps: collecting a glass image, and making a training set and a verification set according to the glass image; then, constructing a deep learning network model according to a Transform network; training the deep learning network model by using the training set, and verifying the trained deep learning network model by using the verification set to obtain a verification value; calculating an average accuracy value MAP according to the verification value, and determining whether a deep learning network model is stored or not according to the MAP; and if the deep learning network model is judged to be stored, identifying the glass by utilizing the stored deep learning network model. The invention aims to overcome the defect that in the prior art, the robot collides with barrier glass due to low glass recognition precision of the robot. According to the invention, the position area of the glass can be effectively identified, and the detection precision is improved; and therefore, the robot can be prevented from colliding with barrier glass, and high practicability is achieved.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and more specifically relates to a Transformer-based glass recognition method. Background technique [0002] Robots need to rely on sensors to perceive the surrounding environment in complex scenes. Common ones include laser sensors, single-eye and binocular cameras, and various depth cameras. However, these sensors can allow robots to easily perceive some objects with clear textures, but they can be seen everywhere. In the case of glass products, the content inside the glass area is often similar to the content behind the glass, making it difficult to identify it. If our robot's sensors can't accurately identify the glass, it will cause the robot to make a big mistake in the initial SLAM process. [0003] Researchers initially used laser range finders and designed algorithms based on the reflection characteristics of laser beams to meet the robot's detection of glass walls, but they c...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to overcome the low precision of robot identification glass in the prior art, which will cause the robot to bump into the obstacle glass, and propose a glass identification method based on Transformer, which can effectively identify the position area of ​​the glass, and improve The detection accuracy is improved, and the robot can be prevented from hitting the obstacle glass, which has high practicability

Images