Greg Blackman reports from the European Machine Vision Association's business conference in Prague, where embedded vision and the threat from Chinese manufacturers were both discussed
Jochem Herrmann, chief scientist at Adimec and EMVA president, has issued a wakeup call to the machine vision community about the opportunities and threats posed by embedded vision.
Speaking to EMVA members at the association’s business conference in Prague, Czech Republic on 23 June, he said that, in his assessment, PCs will eventually be replaced by embedded solutions, and that ‘things will change, that’s for sure’.
Frank Grube, CEO of Allied Vision, echoed Herrmann’s statements, commenting during a panel discussion at the conference that ‘we all have to adjust’, while Dr Albert Schmidt, CEO of Baumer Optronic, called embedded vision ‘a new lion entering the arena’.
Embedded processors like those from ARM or Xilinx are powerful computers; they are small, cheap with low power consumption. These kinds of processor mean ‘cameras will be redefined’, Herrmann said, adding that all customers will need is a sensor and an interface, and the processing will take place elsewhere.
A few machine vision camera suppliers – Allied Vision and Basler are two – are starting to adapt to the embedded market, Allied Vision through its new 1 camera product line that contains an ASIC processor and retails at €99, and Basler with board-level cameras and through the recent acquisition of an embedded computing consultancy firm.
These early moves into the embedded vision space – a market that originates from consumer electronics and autonomous driving, both sectors far larger than machine vision – are primarily to tap into areas outside of traditional factory automation, markets like retail or imaging on-board drones. This is where the opportunities offered by embedded vision lie and, while the price of products sold into these markets must be low, the volumes are higher.
The threat from embedded systems is competition from tech companies outside of the machine vision sphere. Herrmann warned: ‘If we as the machine vision community say: “Embedded vision is different and we will keep to our own experience in PCs”; if we don’t [enter this market], some other guys will. They will come from Silicon Valley, they will run over your market and you’re out. It’s better that we take the lead ourselves in this transition.’
He added that the price of the vision products will go down, but that the cost of components is also decreasing and the volumes going up. ‘In the end, it’s a big advantage for everybody.’
‘Once we crack the nut and make it more affordable and easier to use machine vision, it will be everywhere,’ Herrmann remarked.
Herrmann urged EMVA members to attend the embedded vision conference the association is organising with Messe Stuttgart, to be held in Stuttgart from 12 to 13 October 2017. The conference aims to give attendees a better idea of how embedded systems will affect their business, and what the risks and opportunities are. Discussion is certainly still required on the topic of embedded vision – both Grube and Schmidt agreed they don’t have a clear idea of how the term should be defined. Allied Vision launched its 1 camera product line at the Embedded Vision Summit in Santa Clara in early May, and from Grube’s experience at the conference, engineers from Silicon Valley use different terminology when talking about embedded systems. ‘We talk with our machine vision syntax, which means standards and how we define cameras – they don’t understand us,’ Grube said. ‘We’re teaching 360 [Allied Vision] employees our definition of embedded vision and to understand the ecosystem fully [in order] to be prepared to enter that market; that will be really fascinating and I look forward to it.’
Standards for embedded
Herrmann’s presentation at the EMVA business conference was about future standard initiatives, and the G3 Future Standards Forum has formed a study group investigating embedded vision. One of the conclusions from this work was that traditional standards like CoaXPress are too expensive and bulky for an embedded system. An embedded processor will already have all the I/O needed to connect to a sensor. So, to keep complexity and cost down, the group is considering the interfaces currently on the processors and sensors, MIPI being one candidate, and SLVS-EC being another, a new standard from Sony on-board its latest generation of sensors. The question concerning the SLVS-EC interface is: is Sony willing to make it an open standard? Herrmann said that, from initial discussions, there are some signs that Sony is willing to do this, although it is still early in the talks.
‘The goal is to come to a relatively simple interface,’ Herrmann commented, adding that it will probably be modular, with multiple cables that can run in parallel, for example. ‘We are really looking at design with cost in mind,’ he said. ‘Don’t raise the cost too much because otherwise you kill the benefit.’
The working group is still in the scoping phase, and nothing has been finalised yet. However, the group does have plans for GenICam 4.0 – the standard is now in version 3.0 – to support both classic and embedded systems. This means the software controlling the camera stays the same. ‘It’s a big enabler to be able to use embedded cameras and embedded hosts,’ Herrmann said.
The other big drive in terms of standardisation is the OPC UA companion specification for machine vision, being led by the VDMA. This began in March 2017, and is designed to make it easier for vision systems to communicate with the factory floor, preparing the way for Industry 4.0 and the more networked factories of the future.
Threat from China
Embedded vision wasn’t the only topic of conversation at the conference. Schmidt remarked during the panel discussion that ‘the largest threat at the moment is the high-volume, low-price entrance from China’.
Chinese manufacturers like Hikvision and Dahua are entering the machine vision market from the security sector, with incentives from the Chinese government to do so, according to Grube – both exhibited for the first time at the Vision show in Stuttgart last year. Grube said that it was important to offer a full product package, including pre- and post-sales support, so that it’s not just about price. He added that western companies still have a high reputation in China. Roughly 20 per cent of Allied Vision’s revenue comes from China, according to Grube.
Grube commented that Allied Vision was actively looking to acquire companies that fit with the firm’s growth strategy, although he added that growth would also be organic.
Dr Jost van Kuijk, CEO of Adimec, also commented during the panel discussion that the biggest threat for machine vision suppliers is self-complacency. ‘You have to reinvent yourself every year, inserting as much fear and uncertainty and doubt into the guys that want to enter, especially from China,’ he said. He observed that one threat is that image sensor suppliers will start to sell directly to end-users, bypassing camera manufacturers. ‘We’re trying to show that we’re an extension to our customer’s innovations, but also to the innovations of the image sensor companies.’
Europe saw 8.5 per cent growth in 2016 compared to 2015, reaching more than €3 billion in revenue, according to the latest EMVA market statistics presented during the conference. In addition, the first quarter of 2017 was the strongest quarter on record, with growth of 20.5 per cent compared to the same period last year. The EMVA expects this to slow over the year, predicting 8-10 per cent growth in 2017.
The threat from Chinese camera manufacturers taking market share in Europe could potentially be offset by the opportunities offered by the embedded vision sector. What is certain though is that machine vision suppliers will have to adapt to a changing vision market. In the final keynote of the conference, Adam Kingl from the London Business School warned EMVA members that their biggest competitor will be ‘someone outside this room’. ‘It will come out of leftfield,’ he said.
Related article:
Vision on a different scale - Greg Blackman investigates embedded vision
