April 2020

56 n MACHINE VISION April 2020 www.drivesncontrols.com Chips, standards and software drive next-gen vision technologies B y 2025, global shipments of machine vision sensors and cameras will reach 16.9 million, creating an installed base of 94 million industrial machine vision systems, according to a recent report* from ABI Research, (published before the Coronavirus became a pandemic). Of that installed base, it predicted that 11% will be based on deep learning (DL) technologies. Machine vision systems are already well- established on production lines for applications such as barcode reading, quality control and inventory management. Vision underpins many emerging technologies, including robotics, AI (artificial intelligence) and smart glasses. “These solutions often have long replacement cycles and are less prone to disruption,” says ABI’s principal analyst, Lian Jye Su. “Due to the increasing demands for automation, machine vision is finding its way into new applications. Robotics, for example, is a new growth area for machine vision: collaborative robots rely on machine vision for guidance and object classification, while mobile robots rely on machine vision for Slam and safety.” Deep learning-basedmachine vision is an emerging data-driven technology that uses a statistical approach, allowingmachine vision models to improve as more data is gathered for training and testing. Some vision vendors are already starting to realise the potential of DL and are expanding to encompass the new technology. Cognex, for example, has acquired Sualab, a Korean developer of DL vision software for industrial applications, while Zebra Technologies acquired Cortexica Vision Systems, a London-based developer of AI-based business-to-business computer vision systems. At the same time, chip vendors are launching new chipsets and software stacks to facilitate the implementation of DL-based vision systems. The chip-maker Xilinx has worked with the image sensor manufacturer Sony and camera vendors such as Framos and IDS Imaging to apply its Versal Acap system-on-chip technology. Intel is offering its OpenVino technology for developers to deploy pre-trained DL-based machine vision models through a common API. Another chip-maker, Lattice Semiconductor, is focusing on low-powered AI for embedded vision through its senseAI stack, which includes hardware accelerators, software tools and reference designs. These stacks aim to ease development and deployment challenges and to create platform“stickiness”. On the standards front, machine vision suppliers are now bringing 10GigE (Gigabit Ethernet) and 25GigE cameras into factories. Improvements in video capturing and compression technologies are also generating better image and video quality for DL-based machine vision models, helping to future-proof machine vision systems. ABI has ranked the leading machine vision vendors on various criteria and concluded that the leader is the German supplier Basler, which scored highest in innovation and topped seven of the eight criteria that it assessed. Basler was followed closely in the rankings by Cognex and Flir Systems. “The machine vision landscape has gone through multiple iterations, and the top three players have demonstrated their strength in keeping up with the technology evolution,” says Su. “Basler and Cognex have managed to offer end-to-end machine vision capabilities to their clients through the acquisition of key companies such as mycable and Silicon Software by Basler, and ViDi Systems and Sualabs by Cognex. Flir Systems, on the other hand, develops a family of user-friendly and powerful DL- based hardware and software, which allows for more flexible and customisable implementation depending on user needs.” Below the top three, there are various niche Machine vision is a mature technology with established incumbents. But advances in chips, software and standards are bringing innovation into the sector, especially in the area of deep learning. A recent study has analysed the vision market an identified the sectors leaders and technological trends.