To achieve intelligence, solving embedded vision system challenges is an urgent task!

Embedded vision systems are used in many fields, such as industrial automation, drones, traffic monitoring, mobile devices, automobiles, etc., with their powerful processing performance and diverse functions to replace traditional labor to improve production efficiency. With the development of technology and the growing demand for more services, embedded vision systems are also ushered in more challenges, such as power consumption, complex algorithms, processor performance, higher image resolution, etc. It is a more intelligent system, and the embedded vision system is an important part of implementing an intelligent system.

Figure 1: The composition of the embedded vision system

As the input of the system, CMOS and CCD are the two leading technologies used in image acquisition. CCD can provide higher image quality, but after the development of CMOS in the past 10 years, the gap between them is getting smaller and smaller, in power consumption and cost. And power consumption is much higher than the CCD trend. In addition, many applications require efficient parallel processing systems, so dedicated hardware processors such as GPUs, DSPs, FPGAs, and multi-core SoCs are required, but this will undoubtedly increase system cost, power consumption, and PCB size. Therefore, a cost-effective processor is also required by the industry. Of course, in practical applications, we should choose the appropriate processor based on the real-time performance, power consumption, image accuracy and algorithm complexity of the system.

To help users build their own embedded vision platforms and products, Xilinx alliance partner Avnet has launched a range of vision application solutions, such as the PicoZed embedded vision development kit , where the PicoZed SoM integrates the Xilinx Zynq-7030 All Progammable. The SoC also includes the PicoZed expansion board V2.0 , HDMI FMC expansion board (integrated camera interface) and a Python-1300-C SXGA (1280x1024) camera module.

Figure 2: PicoZed Embedded Vision Development Kit from Avnet

This PicoZed Vision Development Kit is ideal for developing machine vision applications. In addition to hardware, software tools and rich licensed IP core resources, it also supports the reVISION Stack technology stack. The reVISON Stack includes a wealth of design resources such as algorithms and hardware-accelerated OpenCV libraries. And the current popular neural network training data set. The embedded vision system is still evolving. Under the efforts of major manufacturers and engineers, it will break through various bottlenecks and gain more applications in the fields of machine vision, artificial intelligence, Internet of Things and industrial automation.