Service robot grabbing detection method based on double-channel convolutional neural network

Abstract

The invention belongs to the technical field of service robots, particularly relates to a service robot grabbing detection method, system and device based on a double-channel convolutional neural network, and aims to solve the problem that a robot grabbing detection method is difficult to have both real-time performance and accuracy. The method comprises the following steps: acquiring an originalcolor image and an original depth image of a surrounding environment, and converting the original color image into an original grayscale image; obtaining a target object bounding box, and obtaining afirst depth image region and a first gray level image region; adjusting the image area to a set size; encoding the image regions with the adjusted sizes, and adding the encoded image regions; decodingthe added feature maps to obtain a captured quality feature map, a captured width feature map, a captured height feature map, a captured first angle feature map and a captured second angle feature map; and obtaining an optimal grabbing rectangle and an optimal grabbing detection frame corresponding to the optimal grabbing rectangle in the original color image, thereby realizing grabbing detectionof the target object. According to the invention, the real-time performance and accuracy of grabbing are ensured.

Description

technical field [0001] The invention belongs to the technical field of service robots, and in particular relates to a method, system and device for grasping and detecting a service robot based on a dual-channel convolutional neural network. Background technique [0002] In recent years, with the rapid development of artificial intelligence, sensing technology, computing processing technology, etc., service robots have gradually entered people's daily life. Among them, service robots equipped with mechanical arms are popular because they can provide object grabbing services. focus on. In order to grasp the target object, the robot needs to first use deep learning-based object detection methods (such as Faster R-CNN, YOLO, SSD, etc.) to detect the target object, and then determine the best capture of the target object in the image The detection frame, and then guide the robotic arm to carry out the grasping operation, in which determining the best grasping detection frame of ...

AI Technical Summary

Problems solved by technology

[0004] In order to solve the above problems in the prior art, that is, in order to solve the problem that the existing robot grasping detection method is difficult to have good real-time performance and good accuracy at the same time, the present invention proposes a method based on dual-channel convolution A neural network service robot grasping detection method, the method comprising:

Images