Computer vision (CV) at Edge
Computer vision (CV) is an approach to use camera to capture and analyze video, apply AI/ML techniques to make predictions and trigger business insights. Analyzing digital images from cameras and videos with deep learning data models can accurately identify, classify objects and respond based on what they “see” through the eyes of computer vision.
Some of the critical business problems and opportunities of newer ways of working, need the combined power of the cloud and ultra low latency. This needs an architecture which has the compute capabilities locally(at the edge) and the scale on cloud, to deliver optimal performance real time.
LTI has developed applications that leverage edge compute capabilities, including extreme low latency and local compute, storage and integrated to the cloud for data modeling at scale. These applications help customer significantly prevent quality anomalies and drive 100% safety compliance. They leverage two new AWS services.
- Amazon Lookout for Vision
- AWS Panorama
Amazon Lookout for Vision
Amazon Lookout for Vision can be used to identify anomalies such as defects or missing components in an industrial product, irregularities in manufacturing lines, visible wear and tear in machine parts, defects in silicon wafers, and other similar tasks.
Example use cases are air bubbles in transparent medicinal capsules, scratches/discoloration/dents/ porosity on automobile parts, paint defects, damages/missing components in printed circuit boards etc.
LTI has built a solution for manufacturing companies to leverage Amazon Lookout for Vision in their in-line production systems.
A Japanese multinational conglomerate corporation known for manufacturing automobiles, motorcycles and power equipment.
United States
One of the marquee clients of LTI, this Japanese automaker is committed to strengthening customer trust by offering products that are of the highest quality and adhere to the most stringent safety standards. The company created a quality cycle that continuously enhanced quality at every stage of their process, encompassing design, development, production, sales, and after-sales service. The client chose LTI to execute a Computer Vision proof of concept (PoC) based on Amazon Lookout for Vision to identify visual defects in their production line. The solution had to provide in-line inspection of aluminum-machined parts on the headline for surface quality anomalies like porosity, dents, scratches, misruns, or burrs on five different surfaces. The LTI solution, powered by AWS, scans all surfaces, identifies the defects, and tags them for repair within 20 seconds with a high degree of accuracy, thereby improving quality and reducing cost.
AWS Panorama
AWS Panorama, a new Amazon managed service and edge appliance which enable customers to build and manage computer vision applications at the edge to analyze video feeds from onsite cameras, perform predictions locally, and facilitate real-time operational decision making. AWS Panorama makes it easy for customers to use computer vision at the edge with low latency, across all their onsite cameras, and with a high degree of customization to address a wide range of specific use cases such as quality control, workplace safety, and retail analytics. The AWS Panorama Appliance is an edge device that connects to the AWS Panorama service and has enough processing power to analyze video feeds from multiple cameras and generate highly accurate inference within milliseconds. Applications running on the AWS Panorama Appliance can act locally on the data generated by onsite network cameras, enabling customers to use computer vision in environments with constrained internet bandwidth and intermittent connectivity. AWS Panorama is ideal for organizations with data governance rules that require on premises processing and storage of video.
LTI has built solutions for companies to leverage AWS Panorama in their factories, facilities and sites for safety compliance and facility security.
Portland General Electric
United States
Portland General Electric (PGE), US-based utilities company worked with LTI for their mission-critical infrastructure projects, securing substations with real-time awareness for preventing theft, vandalism, accidents and other activities that could disrupt the stability of the power grid. LTI conducted a successful PoC over a 6-week period to capture, verify, and act on a “crime in progress”. If an intrusion occurs at a substation any time of the day or night, the camera captures a 10 second video, removes false alarms, and confirms whether it is a human or vehicle creating the alert. The solution immediately determines if the alert requires action and sends the appropriate notifications. The LTI solution, powered by AWS is helping PGE reduce material losses and prevent incidents. PGE is one of the first clients to implement LTI’s AI-powered solution based on AWS Panorama service for Computer vision at the edge.
L&T Construction
India
L&T Construction (a sister company of LTI) is India's largest construction organization and is ranked among the world's top 30 contractors. In 2020, L&T put an aggressive goal in place: Achieving zero incidents at all construction sites. With a large and growing workforce, L&T needed to ensure that all new employees and site visitors complied with requirements regarding proper Personal Protective Equipment or PPE. L&T turned to LTI to develop a computer vision solution at the edge to improve worker safety & compliance with wearing PPE such as helmets, glasses, shoes, jackets, and harnesses. LTI developed a PPE detection solution using AWS Panorama to identity and flag personnel who violate safety guidelines in an authorized work zone. With very low latency, the solution detects violations and generates alerts & notifications to supervisors along with a video clip of the non-compliance. This solution helps work sites achieve a 100% safety compliance.
