Video Analytics in Manufacturing Industry
Dive into the dynamic world of Video Analytics in Manufacturing Industry with our exclusive episode on Video Analytics in Manufacturing Industry. In this episode, we delve into the intricacies of deploying and optimizing Video Analytics as a Service, exploring its potential in revolutionizing the manufacturing landscape. Our experts, Dr. Ray Chi, Business Development Manager of video AI/VSMD group, Advantech, and Ajay S Kabadi, Founder & CEO, DocketRun Tech Pvt Ltd shed light on the significance of providing AI-based Video Analytics as a Service, unlocking unparalleled insights and efficiencies for manufacturing processes.
Furthermore, we uncover typical use cases for Video Analytics, showcasing how this technology can enhance security, streamline operations, and optimize quality control within the manufacturing ecosystem. Join us as we unravel the transformative power of Video Analytics in reshaping the future of manufacturing.
While talking about Advantech’s Edge AI Solutions with NVIDIA platforms, Dr. Ray Chi, Business Development Manager of video AI/VSMD group, Advantech, said: Our product portfolio is based on Nvidia Jetson platforms, with a focus on video analytics. According to Nvidia statistics, customers commonly use cameras with different interfaces, such as USB, MIPI, and Ethernet. The choice of interface depends on the specific application.
USB cameras are suitable for short-distance applications like AI, where USB 3.0 industrial cameras can be leveraged. In the case of Autonomous Mobile Robots (AMR), integration with 3D stereo cameras is preferred, making USB 3.0 cameras popular for AMR applications.
MIPI cameras are typically used in close proximity, but our approach involves connecting MIPI cameras with GMSL cameras. This allows us to build a GMSL module to interface with the BP interface. The communication distance with GMSL cameras can extend up to 100 meters, providing greater versatility, especially in applications like vehicles that require high speed, frame rate, and resolution.
Ethernet cameras, similar to IP cameras, are widely used for their convenience in streaming video (e.g., RTSP or rh). This makes them suitable for AI inference in both public and private areas, addressing needs such as human behavior analysis, human counting, safety, and security. The Ethernet camera is a reliable choice for capturing video in these scenarios.
We can integrate various sensors, including light and I/O, into our systems. If you require additional sensors, we can assist with integration. Comparing our DevKit and solution kits to NVIDIA’s, noteworthy features include more USB ports in our kits and support for a wider operating temperature range. This means our solution is suitable for deployment in critical areas or at intersections without any issues. Our product is robust and can handle different installation conditions. We also support industrial power adapters, which are crucial for job sites with unstable power supplies. Our product can withstand accidents, providing a level of durability that may not be present in commercial or consumer DevKits from NVIDIA.
Advantech offers development kits ranging from nano to larger sizes, providing full integration flexibility with MIC-33s support for your development. Additionally, we have pre-existing chassis, certifications, and a one-year warranty for C-label products to support your project.
Explaining to viewers how video analytics in the manufacturing industry works, Ajay S Kabadi, Founder & CEO, DocketRun Tech Pvt Ltd said: In the industry, there’s a widespread use of cameras, including CCTV-based ones, leading to productivity issues and safety concerns. Manual standard operating procedures (SOPs) exist in every industry, specifying how equipment should be operated and what panels and mechanical operations are required.
While there are tools to evaluate scenarios, there’s a lack of standard digitization. Video Analytics, particularly through Advantech’s online solution or on-premises server, addresses this gap. The software connects to any brand of CCTV cameras, including a mix of brands, and analyzes data for safety compliance.
The system checks various personal protective equipment (PPE) standards, such as fluorescent jackets, t-shirts for fires, steam suits, and electrical shock resistance jackets. It ensures adherence to PPE standards in respective industries and departments.
For safety, the system uses Geo-sensing to create zones around machines. Using a GUI tool, it stops the machine in real-time if an unauthorized person enters a restricted area. In case of immediate action, hardware analyzes video analytics, triggering responses like stopping the machine if someone is not wearing the required safety gear, such as a helmet.
The trigger system, located at the camera site, can activate a hooter when someone removes it, providing real-time feedback. Secondly, for immediate machine stoppage, we offer a relay-based output, a real-time PLC-based output that connects to your PLC or directly to the machinery. This potential-free output enhances the ability to stop the machine in real time.
The third aspect involves voice announcements for violations in preferred languages, such as Odia in Odisha or Marathi in Maharashtra. These announcements are based on observed violations. This entire process takes place on-site, ensuring the privacy of your video data, with real-time processing and simultaneous delivery of data in real-time. This is particularly relevant from a safety perspective.
For a more in-depth analysis, consider examining various mechanical scenarios, such as hydraulic operations, gas pipelines, pneumatics, conveyor belts, and electrical panels within an industry. Each of these operations follows a set of standard operating procedures (SOPs).
To assess these scenarios from a camera perspective, we evaluate all SOPs related to machinery operation. This involves scrutinizing each aspect of how employees work with the equipment. If any deviations from the established SOPs are detected, an alert is generated. This alert can then be integrated into the relevant data for further action.