Paul Danini, a technology and market analyst at Yole Développement, says, in a maturing machine vision market, companies will have to adapt to new areas and reduce costs to survive
Industrial machine vision subsystems demand will recover from last year’s slowdown to see relatively stable, approximately 8 per cent, compound average annual growth over the next five years, but the maturing market means low costs, higher volumes, and diversification will become key strategies for success.
The $1.2 billion industrial machine vision market saw only 0.7 per cent growth in 2012, as the slowing electronics industry faced overcapacity for everything from silicon chips to solar cells and displays, and reduced its capital expenditures, particularly hurting suppliers of high-end products for this sector. Suppliers of lower cost, entry-level performance cameras for other industrial markets did somewhat better, as declining prices expanded the market for more applications.
While the semiconductor and electronics industry demand will recover in the next couple of years fuelled by the establishment of a new 18-inch wafer infrastructure at Samsung, Intel and TSMC, other recent trends will continue. The growing labour cost in Asia will result in a trend towards automation of factories, which will present an opportunity for both traditional machine vision players and new local players that may profit from this on-going automation in Asia to compete on price. Meanwhile, markets for new technologies such as 3D imaging and hyperspectral imaging will likely remain only niches.
As a consequence, prices will continue to fall and expand demand to more applications, transforming the sector from a small volume niche into a medium volume market. And with limited room for disruptive technical innovation, machine vision will become more of a commodity, competing primarily on price.
To face that commoditisation three strategies are open to machine vision companies: firstly, diversification into new applications such as surveillance and number plate recognition; or into new products such as sale of camera cores or even entire systems, similar to Basler’s move into the surveillance sector in 2008. Secondly, drastic cost optimisation by horizontal fusion/acquisition and lean manufacturing to gain market share while keeping sustainable profitability, and thirdly, technical innovation to penetrate very niche applications generating high margins, such as with image sensor backside illumination (BSI) techniques, hyperspectral imaging, etc.
Opportunities in surveillance and number plate recognition
With the traditional machine vision market for inspection for automated factory lines maturing, suppliers will have to look outside the manufacturing sector for future growth. We see the primary growth opportunities going forward will be for high-end IP surveillance systems aimed at site security, number plate recognition systems, and automotive driver assistance systems. These markets will all see ongoing growth rates of more than 15 per cent.
All these markets, however, will require traditional machine vision suppliers to listen very carefully to customer needs for a highly fragmented collection of specific applications with different requirements. And these customers will be a wide-ranging group of individuals who are not factory managers – from banks and casinos to government border control agents for security systems, and transportation departments collecting bridge tolls, cities controlling automobile access, and police departments tracking criminals for number plate recognition systems.
Hardware specialist suppliers of industrial cameras may have the edge for surveillance applications, which don’t need the high speeds or differentiated technologies of fast moving production lines. On the other hand, the more software-oriented suppliers of compact vision subsystems may have the edge for number plate recognition systems, with their demanding requirements for distinguishing characters on all kinds on backgrounds in all kinds of weather conditions.
Concerning the advanced driver safety systems, the automotive industry’s unique demands for automotive-qualified production mean that existing tier 1 automotive suppliers and their partners will likely control that very large potential market and exclude traditional industrial machine vision players from that market.
All these non-factory growth applications, and the maturing core factory inspection market as well, will increasingly be driven by cost, with limited room for companies to differentiate themselves by their technological innovations. Most successful companies will probably be those that can achieve sufficient volumes, both to drive down their manufacturing costs, and to reach a critical size to assure customers of reliable delivery of high volume orders.
Also key will be diversification across multiple markets to achieve those volumes, as well as for more stable revenues not whiplashed by manufacturing industries capital investment cycles. We expect ambitious machine vision companies will make acquisitions to increase their capacity and diversify their market reach, across broadened products portfolio, more applications and more geographical regions.
Sector moving to CMOS image sensors
From the image sensors standpoint, the machine vision market has traditionally benefitted from the heavy investments realised for the consumer markets, because the investment capacity in machine vision is limited by its relatively small size – around $120 million in 2012. For that reason machine vision image sensor suppliers have added incremental innovations to ‘consumer’ technologies in order to adapt them to subsystems manufacturers’ requirements.
Following the trend in consumer markets, a major technological shift has been on-going in machine vision for the last five years: technology is rapidly converting from CCD to CMOS image sensors. Though currently only about 35 per cent of the market, sales of the faster, higher resolution, lower power CMOS image sensors are seeing approximately 30 per cent compound average annual growth, and we expect them to command 50 per cent of the market as soon as 2015. The CMOS devices have largely solved their earlier problems with distortion from rolling shutters that imaged a scene only row by row. Most of the CMOS image sensors now have global shutter technology that enables them to match CCDs in imaging the whole scene at once.
The use of the widely available CMOS technology has also opened the market to a crowd of close to a dozen fabless chip design firms to now compete with the two long established CCD suppliers, Sony and Truesense Imaging (formerly Kodak). While CMOS technology is moving towards higher resolution and higher frame rates, those capabilities are of key interest primarily for inspecting electronics lines, so the CMOS side will remain highly dependent on demand and investment in the electronics manufacturing sector.
Speed and resolution alone are also no longer enough to differentiate competing suppliers, so designers are increasingly focusing on adding new capabilities, such as increasing sensitivity to low light, and to infrared and ultraviolet ranges. In the long term, more and more processing power will be efficiently integrated with the CMOS sensor, until eventually the CMOS sensor essentially becomes the camera. CCD sensors will continuing to maintain share, however, where they continue to offer better performance for some applications requiring infrared sensitivity or line scanning.
