Deep learning based recognizing and classifying robot for bottle and can type garbage

Abstract

The invention belongs to the technical field of robot design, and relates to a deep learning based recognizing and classifying robot for bottle and can type garbage. The robot is characterized in thata robot system comprises three parts, namely, a garbage image acquiring system, a digital image handling and recognizing training system and an industrial robot sorting system, wherein the image acquiring system is used for acquiring an image through an industrial CCD camera; the handling and recognizing training system is used for extracting properties (outline and OCR character property) of a garbage sample training set based on convolutional neural network and creating a classifier and training the classifier to realize positioning recognizing and classifying; and the robot sorting systemis sued for transferring the obtained recognizing classifying information to a mechanical arm in an instruction form so as to realize garbage sorting. According to national related policies, environmental protection industry and profitability analysis, most of recoverable bottle and can type garbage cannot be easily decomposed and screened and classified in a normal manner, so that classifying andrecovering the abovementioned garbage is of huge value and significance to environment and resource recycling.

Description

technical field [0001] The invention belongs to the field of garbage disposal, and in particular relates to an intelligent garbage identification and classification robot for bottle and can garbage based on deep learning. Background technique [0002] As urban life becomes more and more automated and intelligent, the sorting and processing of urban waste is gradually becoming intelligent. It is of great significance to recycle and reuse garbage and turn garbage into resources. According to the environmental protection industry and relevant national policies, and through profit analysis, it is concluded that most recyclable bottle and can waste, such as: glass (beer bottles), plastic bottles and most cans (beverage bottles), are not easy to decompose and cannot be used conventionally. It has the characteristics of magnetic attraction and vibration screening and classification, and has great value for the environment and resource reuse. [0003] Object recognition is an impo...

AI Technical Summary

Problems solved by technology

In the existing patent and technical research, for example, the visual positioning system based on image recognition technology is poorly applied to a wide variety of garbage. In the complex garbage processing process, the complexity of feature matching is high and the robustness is poor. It is difficult to meet the requirements of identification and classification, such as China Patent No. CN201610056834.3

In the Chinese patent CN201810939861.4, a garbage sorting system based on visual recognition and convolutional neural network was designed. The recognition system based on convolutional neural network greatly improved the accuracy of recognition and classification and designed a The execution structure of the identification and sorting system introduces the sorting operation process of each part in detail. However, for a wide variety of garbage types and garbage dumps with complex environments, it is obviously impossible to achieve accurate classification. In the identification and classification system It is very difficult to design different convolutional neural network structures and algorithms for various types of garbage, and the workload is too large, so it is difficult to achieve the optimal design of the positioning and identification system. There is no clear definition of various types of garbage in this patent. Algorithm parameter design

Among them, the structural design of the garbage sorting system is mainly introduced in detail, but the stability of its structure and the coordination and optimization of the combined identification and control system are poor, and it is difficult to realize the task of identification and sorting in a complex environment

Images