16 Apr
Companies necessitate to continuously upgrade their existing infrastructure and methods/rules to be more efficient and provide safety for employees. As the country begins to slowly reopen or restart their work from the COVID-19 lockdown, several people are going back to work. While this is excellent for the nation’s supply chains and overall economy, a developing concern exists to improve the risk of virus transmission. Every industry’s first and foremost concern and priority should be to give safety to their employees from the COVID-19 and also from their heavy industrial machinery.
After the modern/manufacturing revolution, humanity or understanding has made tremendous improvements in manufacturing and other industries. With time changing, we have observed more and more ordinary manual work being replaced by automation. Working in an industrial atmosphere has always been dangerous/critical, and factory managers attempt to find the best solutions to reduce fatalities. The purpose is to decrease the hazard, manage risks, and prevent accident situations.
We assume that recent progress in AI (or Deep Learning and other related technologies) will help stimulate this trend towards automation more engagingly. This is because AI attaches one very significant component that the companies/firms have been avoiding or missing until today — “The strength of machines to understand and recognize.”
Let’s discuss more information about worker safety in industries and How AI enhances this process, and so on.
Table of content:-
- What is the objective of workers’ safety detection?
- Why do we need workers’ safety detection?
- How does an AI-based solution work in workers’ safety detection?
- Complete process
- Data extraction
- Data annotation
- Data processing
- Model development
- Model inference & deployment
What is the objective of workers’ safety things detection?
The main goal or objective of workers’ safety thing detection is to identify the workers in the image/video stream wearing industrial safety things such as goggles, helmets, etc., or not. If workers wear safety things, then determine wearing properly or not with the help of Artificial Intelligence techniques like computer vision and deep learning algorithms.
Why do we need a worker’s safety things detection?
The manufacturing or construction industries are the beating heart of any strong economy. To maintain that heart beating, organizations need to manage their workers safely. Safe workers are more comfortable, healthier, and more productive.
These enterprises need workers to engage in high-risk actions such as soldering, welding, metal cutting, raw material assembling, and heavy lifting and rigging. Furthermore, magnetic fields, compressed gases, and harmful radiation can negatively affect a worker’s health. Workplace hazards lead to approximately 150 deaths per day in the U.S. Naturally, these industries’ processes involve many hazards.
Companies can manage a team that is always watching workers’ activities and warning if anyone is not wearing safety. But this maintenance process is complicated if organizations have more workers or workers working in hiding places. In this process, we may face many human errors; these errors may lead to some employees’ death or damage to any industrial materials.
This has clearly brought into further focus the need to improve industrial place safety. organizations are trying to improve safety rules and training techniques using the latest technologies like Artificial intelligence techniques and methods. These days, every industry focuses more on providing safety as another prime goal, not simply because it’s the correct action to do, but because it’s also good business.
How does an AI-based solution work in workers’ safety things detection?
The most essential and required task in every manufacturing and construction industry is safety management. In these industriesWhole departments are dedicated to worker and facility safety. In most plants, workers are required to wear Personal Protection Equipment (PPE) kits. PPE consists of everything from hard hats or helmets and goggles to gloves, face masks, earplugs, etc.
The recent trend is to uncover hidden insights from all kinds of data across industries: artificial intelligence technology. The main question naturally arises as to whether AI-powered data can be utilized to enhance the workers’ health and safety at the industry level. The answer is yes.
Artificial Intelligence (AI) is a robust technology that can be applied to manufacturing businesses to automate operations and maintain safety measures while reducing overhead costs. The most serious impact of AI may be the technology’s ability to predict the future. Luckily, AI is coming to the construction/manufacturing industry is one of the organizations that need more benefiting—and protecting lives. However, the value AI brings to each industry is enormous.
The AI system utilizes algorithms that analyze existing industrial site workers’ images with PPE or safety things or without and combined with real-time monitoring and identity and alert workers who are removing their safety things by, fortunately, or unfortunately.
Complete process
Data extraction
We collect several images of people who are not wearing masks, wearing masks, and not properly wearing a mask. Then we extract images of Industrial workers for detecting safety things such as goggles, PPE, helmets, Etc. We extract all these kinds of images by extracting them from different sources, such as social media platforms and machines (cameras) of different resolutions.
Data annotation
We annotated the extracted raw images with correct label kinds such as masks, no mask, and not wearing them correctly, goggles, PPE kit, helmets, etc., for this problem statement. This information will help the model to recognize all these things from a video stream or images.
Our annotator team was annotating all these objects with an annotation type of bounding box within the images.
Only workers’ safety things detection models will detect all these things if annotating is appropriately done. Otherwise, this system should not identify target output without high quality annotated dataset. Just think about how vital the annotation process is to the model or system.
For this, we should provide better-annotated images to the model while building. Without an annotation, the model can’t know where it will take patterns and then learn. So automatically, AI-based models shouldn’t discover any patterns to recognize our target outputs.
Data processing
Extracted images needed preprocessing before moving to the subsequent steps. Data preprocessing is the collection of methods used to apply on the raw extracted input images to convert them into clean versions, which could be fed to Artificial Intelligence models or neural network machine learning models.
The first step of image data preprocessing is the conversion of images into grayscale images. Because the RGB (Red Green Blue) color image carries so much repetitive or irrelevant information that is not necessary for face mask detection. Then, we reshaped the images to keep the uniformity of the input images to the architecture. Then, the images are normalized, and after normalization, the value of a pixel stays in the range from 0 to 1. Normalization increased the learning algorithm to learn faster and obtained the necessary features from the images.
Model development
Artificial intelligence gives us a straightforward solution for these workers’ safety detection. By enforcing guidelines around the wearing of PPE, ensuring social distancing, and thermal imaging capabilities, AI can reduce transmission and ensure staff members’ health and safety.
Identifying multiple objects like masked and unmasked faces in images can be solved by a traditional object detection model. The process of object detection primarily includes localizing the objects in images and classifying them.
This workers safety things detection model follows two main stages. This model’s first step involves a Face Detector, which identifies multiple faces in images of different sizes and detects faces even in overlapping situations. The detected faces are extracted from the image/video stream then transferred to the second stage of object detection model architecture, a computer-vision-based deep learning neural network object classifier. The second stage results are decoded. The final result is the image with all the detected faces in the image and classified as any available target labels, either mask, goggles, helmet, PPE kut, etc.
First stage: Face recognition
The preprocessed images are passed as the input to the face detector model—the face detector extracts and displays all the human faces detected in the input image with their bounding box coordinates. Building a highly accurate or performance value face detector requires a lot of labeled datasets, time, and compute resources. The face detection process is challenging for the model used in this first stage of the model architecture. It demands the detection of human faces that could also be covered with masks, the helmet’s shadow, etc.
Second stage: multi-objects detection
The second stage of the present detection model is multi-object detection (identify objects and then classify them). This stage takes input as the processed or detected face-related features from the previous stage’s output.
Model inference & deployment
We can use our trained model in different ways. Below are a few ways to utilize trained AI-based workers’ safety models or systems at the industrial level work.
Imagine: as workers report at their firms, workers place an industrial-grade wearable gadget on their wrists, put on their PPE (Personal Protective Equipment), and get to work. Across the workday, as employees become utilized in the task at hand, they casually step closer than 6 feet to their co-workers or remove their PPE. They will hurt their face or co-workers’ face and fail to put it back on properly. The AI, using high-definition cameras, notifies this action and conveys a gentle warning to the employee via their wearable to stay six feet away or substitute their PPE while sending another notification to the manager, which allows them to recognize staff members who may require additional coaching.
We can install AI-powered cameras to monitor employee PPE. This process can be two steps. First, placing cameras in the entrance checkpoints of a department. This reduces the requirement for manual checks by humans and alternatively only requires someone to be informed if a PPE violation is recognized. The next following step is to continue observing the facility to make sure employees continue to wear all required PPEs or not and are in the factory’s regulated areas.