Machine vision platform for industrial applications
Our vision is to provide innovative machine vision solutions aiming to improve the safety and environment of the industrial processes and products, working accordingly with the worldwide Sustainable Development Goals(SDGs). This project takes on the challenge of reducing the hazards associat
2025-06-28 16:34:04 - Adil Khan
Machine vision platform for industrial applications
Project Area of Specialization Artificial IntelligenceProject SummaryOur vision is to provide innovative machine vision solutions aiming to improve the safety and environment of the industrial processes and products, working accordingly with the worldwide Sustainable Development Goals(SDGs).
This project takes on the challenge of reducing the hazards associated with the products and processes employed by world’s most cancer-causing industry; the Tobacco Industry. Around the world, 1 in every 8 people willingly smoke tobacco, but also unknowingly inhale foreign impurities or Non-Tobacco Related Material (NTRM), which comes in various forms and may be even more toxic to health.
Our Real-time Optical Detection and Visual aid system, currently employed in the Green Leaf Threshing(GLT) plant of Pakistan’s largest tobacco producer; Pakistan Tobacco Company(PTC), increases the efficiency of the impurity removal process and makes the manual worker’s job easier(by providing visual aid), in an attempt to semi-automate the company's inefficient manual impurity removal process.
Images containing the NTRM were captured in a non-destructive manner using a Global shutter camera, placed in a purpose designed black box, meant to reduce external and ambient lighting effects. Several processing techniques were applied for feature extraction from NTRM, from basic image processing to more complex machine & deep learning algorithms. The segmentation of impurities was achieved by training deep learning models for over 5000 samples. The trained models are processed by a Jetson Nano to provide real-time detection and co-ordinates of the impurities are then mapped out and sent to an actuation system, employing a projector, to provide visual aid for the manual searchers in the removal of hazardous impurities.
Project ObjectivesOur project aims to provide innovative machine vision solutions which increase product quality and safety, improve working environment for labor, and provide monetary benefit while beng in accordance to the Sustainable Development Goals (SDGs).
Our Optical Detection System, employed in Pakistan Tobacco Company's tobacco treatment facility aims to
- Decrease the enanched health risks associated with hazardous foreign material in tobacco leaves.
- Decrease mental drain on the manual searchers working in the difficult environment of the tobacco treament facilities.
- Enhance impurity removal process's efficiency to by atleast 30%.
- Provide cost effective process innovation
- Provide environment friendly solution
- Provide easy to adopt and maintain solution
For the implementation, we have come up with a few simple steps that the entire team is going to adhere to and follow at every step of the way till the product has been deployed and successfully implemented.
These comprise of:
• Product awareness: includes B2B marketing that will primarily be handled by the marketing team.
• Demonstarting a prototype: includes showing a working model of our system (both detection and tracking mechanism) to prospective clients on their own factory setu.
• Holding Discussions: includes arranging meeting with the client and reviewing the final dealings.
• Deployment: deployment of the hardware at the factory.
• Testing: includes testing and maintenance of the final setup and the training of the factory staff to the new equipment.
In terms of the timeline, so far we have been successful at demonstrating a working prototype to our client – the first of which happens to be Pakistan Tobacco Company.
Right now we are holding discussions to finalize the cost and logistics after which we’ll be deploying the product - we anticipate this will take around 2-4 weeks and then finally we will move on to the testing and maintenance phase.
Around the world the tobacco trade amounts to around 80 billion USD and unfortunate for Pakistan, it has only been able to penetrate 0.03 percent of this vast market with international exports amounting to no more than a mere 25 million dollar paycheck for the entire tobacco industry of Pakistan. It is important to realize that, with the recent developments in this sector – specifically in the manufacturing processes with the increase in number of machinery, this industry has the full capability to export large quantities but has so far been unsuccessful to achieve this potential.
This is highlighted in the following excerpt for a news article:
“The report stated that local tobacco firms had received an import order earlier but were unable to meet quality standards so the entire stock has to be discarded”.
The same problem is also faced by the biscuit producers of Pakistan where they have been barred from exporting their product due to low quality standards. Another huge problem they face is the waste of the useful products along with the rejected pieces due to lack of the required technology or the existing technology being too expensive these products cost between 1.4 and 1.7 Million dollars. With our novel algorithm, we have brought the price down to a mere 1500 dollars.
Continuing with the comparison, our custom designed mountings have been fabricated to take very little space on the conveyors and since there are no physically moving parts, the maintenance cost is at a bare minimum. By far the biggest advantage we possess in comparison to the other competitors in the market is that we have even addressed a problem that concerns every industry out there, and that is; for implementation, there are no changes required in the existing setup so our product’s ease of adoption is incomparable.
Software:
Programming language used: Python 3.6
Libraries used: Gxipyand OpenCV
The aim is to develop a program that can help collect data either manually or switch on a live stream recording to be used as a test video.
For user purpose, Key press ‘Q’ was recorded against a data sample being captured and collected.
Data Annotation: Data annotators help categorize content. They can work with things like videos, advertisements, photographs and other types of material. They assess the content and then attach tags to the content.
Data Labelling: Data labeling is used to enable the system’s AI to tell the difference between objects sky by labeling their key.
Interconnectivity between Detection algorithm and Tracking mechanism: The algorithm running over Jetson Nano grabs the coordinate of where the impurity is and sends it over Wi-Fi via socket communication using python to Raspberry pi that controls the display of the screen projector.
Beam-box projection generation using screen projector: The coordinates received using socket communication can be used to generate a red box in specified location using OpenCV on python (over a black background). The projected box is moved at a speed equivalent to the conveyor.
Hardware used:
• NVdia Jetson Nano
• Rasberrt pi
• Epson Projector
• Daheng Machine vision camera
• Projector Mounting
• Camera Mounting
• Black box
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 80000 | |||
| Global Shutter | Equipment | 1 | 40000 | 40000 |
| Microcontroller | Equipment | 2 | 5000 | 10000 |
| Frame | Equipment | 1 | 5000 | 5000 |
| Projector | Equipment | 1 | 15000 | 15000 |
| Logistics | Miscellaneous | 1 | 10000 | 10000 |