Computer Vision Based Automated Sorting System using Machine Learning

Project Title

Computer Vision Based Automated Sorting System using Machine Learning

Project Area of Specialization Artificial IntelligenceProject Summary

Computer Vision Based Automated Sorting System Using Machine Learning :an Industrial Model

Introduction: The industries  are at the verge of the fourth  technological revolution formally termed as  Industry 4.0 that would  implement industrial scale mechanisms via  smart methodologies such as IoT,IoS,machine Learning,Computer Vision etc in a more efficient ,reliable and optimum manner.The overwhelming number of computers: envisaged from Moore’s law and the smart methodologies mentioned above which are top notch technologies uptill now which  date back to the “Summer Vision Project” of MIT elevates the potential of the whole industrial automation process ,therefore, the the tasks that were done manually and required time and human effort are now  accomplished at an accelerated rate ,with little or no  human assistance ,thus giving rise to a smarter environment.

 

The Project Idea: Computer Vision based automated sorting system using machine learning :an Industrial Model is an industry 4.0 based project that moves a way forward from  the current methods  deployed at the industries for objects sorting which are either solely human assisted or less optimum and accurate if there is some automation mechanism involved. We are taking the vision based machine learning initiative to tackle the problem under consideration,automated feeding and sorting of objects reliably and accurately is the main goal of this project ,hence, paving the  way for industrial automation and  leading further towards  industry 4.0.

The project idea is to replicate the industrial environment; therefore creating a prototype which will automatically feed via a feeder different objects (Lego bricks for the sake of simplicity) at one end ,the objects will traverse through the first conveyor belt ,the object will be analyzed on the basis of some predefined  parameters ,characteristics  and attributes the system will separate different lego bricks ;than  a single  lego brick will traverse  through the second conveyor where it will be analyzed to which brick type it belongs  ,the objects will get to their respective sorted bins via a sorter connected at the other end.

'figure 1 below illustrates the working machenism of the model'

Expected Results: The expected results for such models range from 80%-95% ,if the model is trained in a correct manner .

Conclusion: The proposed project caters the need of the forthcoming industrial automation era that would rule out the current methodologies which are either  manual or semi automatic in nature demanding time and human effort ,following the industry 4.0 trend such techniques will subsequently  become redundant and eventually later will get obsolete .

Project Objectives

The project aims to devolop a model that is streamlined with industry 4.0,it will be compatible with the the forthcoming industrial automation era.

The project will use Computer Vision coupled with Machine Learning :the former provide the means to detect,sepearte and subsequently sort the objects and latter eases the learning process ,as a result we dont have to spoon feed every instruction to the machine ,but instead we relay on Neural networks that are easier to implement ,better performing and are not prone to inaccuracies and shortcomings.

The proposed model will be automated to the full extent ,it will interface Rasberry pi with PLC via RS 485 ,hence facilitating the the communication between hardware model and software system and the result will be increased manuverabilty of the model.

The model also aims to provide a sorting system that will be easier to interface with other industrial opreations and consequently will be able to synchronize itself with rest of the industry.

Project Implementation Method

The implementation of the project is divided into three phases stated below:

1.Devolping Hardware Model:the first stage of the project is crafting a hardware model that automatically feeds 15-20 industrial objects (lego bricks are considered as standard industrial objects in developmental phase) via feeder from one end of the model ,the objects traverse through conveyor belt 1 ,they are separated at conveyer belt 2 ,conveyer belt 2 takes a single object and the raspberry pi/PLC will perform sorting mechanism and the object will be subsequently sorted at the other end.

'figure 2 illustrates the hardware model'

The RS 485 interfaces  PLC and Rasberry pi : the former is responsible for controlling  Servo motor of the feeder,DC gear motor of conveyer belt 1 and converyer belt 2 and latter captures the image of the lego brick, the image is processed by the neural network to identify whether the image contains a lego brick or not ,once it is identified as a lego brick,the image is transferred to a complex neural network stored in the linux server to process the image and compute to which brick type it belongs and after this computation the Raspberry Pi/PLC catogerizes the brick to its respective sorted bin .

2.The Training Phase:once the hardware model of the prototype is ready the second of the project begins; as stated earlier the project uses Machine Learning which require an adequate amount of data to train itself for accurate predicitons ,so we will provide enough datasets to the model that it can asses lego brick from different positions,orientations,angles etc.therefore the model will no longer be susceptible to the shortcomings that were posed to the previous models and thwarted the whole automation process.

The training phase can be further divided into two substages mentioned below:

2.1 Identifying Lego brick:The first and foremost part is to identify the lego brick with the help of Computer Vision techniques that is:capturing the image,processing it on some Neural network and yielding some output which triggers the further mechanism stated in 2.2.

2.2 Categorizing the lego brick:once the lego brick is detected the image of that brick will be transferred to the neural network(stored in google cloud based server)that after processing accurately sort the object to its respective bin.

The model in the training phase will be experimented with different neural networks such as CNN,RCNN,etc or a combination of different neural networks to yield the best results.

3.The Testing Phase:after the training we will test our model that will be based on industrial metrics such as separating accuracy,sorting accuracy etc;moreover that results will suggest if the model needs to be trained further or not.

 

Benefits of the Project

The project will provide the means to interface the sorting mechanism with other industrial operations and synchronize it with the rest of the industry.

The project proposed is automated to the full extent ,is streamlined with the industry 4.0 and leverages the full potential of the AI based solutions.

The project is not susceptible to any inaccuracies,shortcomings etc which degenerated the previous models to a semiautomatic nature and rendered the whole model useless.

the project does not need any human assistance of any kind , which was a big challenge in the previous models as they required frequent human interventions.

 

Technical Details of Final Deliverable

The final deliverables of the prototype that we are devolp[ing would conist the following :

1.Rasbperry pi 3B

2.Camera

3.PLC

4.Conveyer belts

5.Arduiono uno

6.DC gear motor

7.Servo motor

8.IR beam sensor

9.Raspberry pi adapter

10.Miscellneous minor components

Final Deliverable of the Project HW/SW integrated systemCore Industry ManufacturingOther Industries IT Core Technology Artificial Intelligence(AI)Other Technologies RoboticsSustainable Development Goals Industry, Innovation and InfrastructureRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	70000
Raspberry pi 3B	Equipment	1	7500	7500
Raspberry pi Camera	Equipment	1	6000	6000
PLC	Equipment	2	10000	20000
Servo Motor	Equipment	2	6800	13600
DC gear motor	Equipment	1	5600	5600
other Compmonents	Miscellaneous	1	2500	2500
raspberry pi case	Equipment	1	1000	1000
raspberry pi adapter	Equipment	1	1200	1200
IR sensor	Equipment	1	1500	1500
arduino microcontroller	Equipment	1	1300	1300
PLC power supply	Equipment	1	9800	9800