Feature Story

What’s Hot in 2015: Smart Vision Systems Poised for Takeoff

Next generation computer vision technologies are paving the way for new applications across a wide range of industries, from automotive to consumer electronics and retail. GPUs in application processors that are often developed primarily for the mobile market are accelerating applications in platforms ranging from cars, to mobile phones and tablets, to drones and robots.

In particular, we will begin to see a major shift in many cameras becoming vision systems, providing such capabilities as enhanced face detection, smile shutter trigger and even analytics. Moving into 2015, the rapid evolution of these "smart" cameras will drive advanced features in video and stills, including face beautification for video conferencing.

Such capabilities are also providing the basis for products addressing surveillance, home security and driver safety. One example is Mobileye, whose proprietary software algorithms and chip are able to "interpret" a scene in real-time and provide drivers with an immediate evaluation based on its analysis. Automakers have already begun to adopt this camera-based technology into their rapidly expanding safety feature applications known as Advanced Driver Assistance Systems (ADAS).

In fact, Mobileye has now shipped more than 2.5 million EyeQ2 and EyeQ3 SoCs (Systems-on-Chips) that are based on Imagination's MIPS processor architecture. These vision processors, together with Mobileye's broad range of algorithms for mono-camera driver assistance systems, target vehicle active safety applications such as lane departure warning, vehicle detection, pedestrian detection, intelligent headlight control and traffic sign recognition. Mobileye has announced that it is leveraging Imagination's MIPS Aptiv and Series5 Warrior CPU cores in its next-generation design.

There's no doubt the classical role of the image sensor is changing rapidly, and will continue to do so in 2015 and beyond. In today's vision applications, the job of the image signal processor (ISP) is evolving away from its traditional role as a separate chip responsible only for producing the best possible picture and moving onto an SoC, where it's able to take advantage of other system resources such as the graphics processor (GPU).

The emergence of computational photography in mobile applications, as well as advanced computer vision algorithms using multi-camera arrays and higher pixel depths in every application area, have also led to the creation of a new class of smarter image signal processor cores. And emerging standards like OpenVX will help to drive the standard functions into hardware on the ISP, reducing power consumption and freeing up the GPU for even more advanced tasks.

Collaborative, heterogeneous computing is needed to address these changes and growing requirements. It's all about doing the most in the lowest power profile and leveraging all the computing resources on the SoC. Imagination Technologies' PowerVR Raptor imaging processor, for instance, does just that by providing scalable and highly-configurable solutions which join other PowerVR multimedia cores to form a complete, integrated vision platform that saves power and bandwidth for today's camera applications and other smart sensors.

PowerVR Raptor cores are low-power, highly-configurable ISP cores designed for SoC integration. The cores can be configured to meet the needs of a wide variety of markets, including mobile, automotive, surveillance, and industrial.

By implementing optimized data paths between the ISP and other on-chip multimedia processors (graphics, video encoders and decoder, and display), system designers can minimize memory bandwidth and latency when implementing algorithms such as those used for computational photography.

Highly intelligent and integrated smart camera technologies present a potentially enormous opportunity both for businesses and consumers. In fact, BCC Research reports that the global machine vision systems market is expected to reach $17.1 billion in 2015 and about $26.9 billion in 2020 (opsy.st/BCCResearchReport). And, according to Yole Développement, the CMOS image sensor market will reach $13 billion by 2018 (opsy.st/YoleDeveloppementReport). Companies that adopt and fuel these innovations early will become the leaders in their sector.

Return to: 2015 Feature Stories