Method for identifying flexible pressure sensing array graph based on deep learning

Abstract

The invention relates to a method for identifying a flexible pressure sensor array diagram based on deep learning, which comprises the following steps of: pasting a sensor on a manipulator to grab an object, and classifying the identified object; generating a flexible pressure sensing array graph sample by using a deep convolutional generative adversarial network technology; training a flexible pressure sensing array graph sample by using an improved neural network model and introducing an attention mechanism to obtain an optimal model; testing the test set in the sample according to the optimal model obtained by training; the trained optimal model is applied to the manipulator, an object is dynamically grabbed, and the type of the object is recognized; the method is high in recognition accuracy, high in recognition efficiency and strong in generalization ability, and can meet the requirements of a large number of application scenes.

Description

technical field [0001] The invention relates to the technical field of image recognition, in particular to a deep learning-based flexible pressure sensing array image recognition method. Background technique [0002] The force generated by the contact between objects and objects plays a very important role in many industrial fields. Different from the general type of sensor, the flexible pressure sensing array can detect the force of flexible-flexible and flexible-rigid contact. Flexible pressure sensing arrays not only have the characteristics of ordinary sensors, but also have the characteristics of thin, soft, and bendable. The flexible pressure sensing array can detect the magnitude and distribution information of the contact force in the bending state, especially in the contact force detection of some curved surfaces or non-planar shapes, it has a good effect, and has been widely used in many fields. [0003] At present, most of the existing researches on the collecti...

AI Technical Summary

Problems solved by technology

[0006] (1) When identifying objects during training, if the differences between samples are small, the accuracy of object recognition will decrease and the generalization ability will be poor;

[0007] (2) Different methods of grabbing objects by the manipulator and different positions of the grabbed objects will affect the recognition results of the objects, which cannot meet the needs of most scenarios;

[0008] (3) When the manipulator grabs an object, the recognition efficiency of the grabbed object will affect the scene where it is used, thus affecting its practicability

Images