Smart cameras are popular with manufacturers who need low-cost, easy-to deploy automated inspection systems. By integrating lenses, sensors, processors, and software, smart cameras are considerably simpler and more compact than traditional vision systems. They can also offer greater reliability than PC-based systems thank to their rugged designs and minimized components.
However, smart cameras also have disadvantages. Their small size limits thermal dissipation and processing power. Thus, smart cameras are often restricted to one simple task like counting or barcode scanning. In addition, their optical systems are typically insufficient for high-speed, high-precision processing. Flexibility is also an issue: many smart cameras offer minimal programmability and I/O. These factors have significantly constrained the applications for smart cameras.
For more demanding applications, manufacturers have turned to embedded vision systems. These systems feature a high-performance industrial PC running a standard operating system, enabling complex image processing and an open programming model. Their versatile I/O allows for multiple high-performance cameras and flexible connection with other factory devices.
However, these features are overkill for many applications. This has created a frustrating gap between the low-cost but limited capabilities of smart cameras and the more advanced but complex and costly embedded vision systems. Hence the obvious question: Is possible to create a solution that combines the best aspects of smart cameras and embedded vision systems?
The new ADLINK NEON-1040/1020 (Figure 1) answers this question with a definitive “yes.” These smart cameras combines a 1.9 GHz quad core Intel® Atom™ processor, a 1” global shutter sensor – 4 MP/60 fps in the NEON-1040 and 2 MP/120 fps in the NEON-1020 – and rich software and I/O support. These features give the NEON-1040/1020 capabilities far beyond those of a typical smart camera while maintaining the simple, compact design that makes these cameras so popular. ADLINK has written an excellent white paper explaining how the cameras work; I'll highlight a few key details here.
Figure 1. The ADLINK NEON-1040/1020 brings new levels of performance to smart cameras.
The key innovation in the NEON-1040/1020 is their use of the Intel® Atom™ processor E3845. This quad-core processor offers far more performance than the single-core processors found in most smart cameras, while maintaining a low power consumption of 10 watts. This leap in processing power enables sophisticated image processing as well as robust multi-tasking.
The Intel® processor also allows the NEON-1040/1020 to offer full PC functionality, with support for 64-bit Microsoft* Windows* 7 embedded and Ubuntu Linux*. This open development environment facilities software porting from PC-based vision platforms, and simplifies integration with motion controllers and other factory devices. In addition, the cameras support a wide range of imaging software, including GeniCAM and GenTL support for image acquisition, STEMMER Common Vision Blox and MVTec HALCON middleware, and Open CV and Open CL programming.
The high-performance processor is matched by a high-performance optical system. Specifically, the cameras use 1” monochrome sensors (4 MP/60 fps or 2 MP/120 fps) with a global shutter. As illustrated in Figure 2, the global shutter is critical for high-speed imaging. Conventional smart cameras often use a rolling shutter, which collects sensor data line by line. This approach can distort fast-moving objects, and thereby limit inspection speeds. In contrast, a global shutter collects all pixel data from the sensor at once, enabling accurate high-speed imaging.
Figure 2. Rolling shutters distort fast-moving objects; global shutters do not.
The main drawback to global shutters is that they require considerably more processing power than rolling shutters. In the past, smart cameras simply did not have enough performance to handle this workload. Today, the high-performance Intel Atom processors used in the NEON-1040/1020, allow these small form factor smart cameras to handle a global shutter.
To add even more performance, the processor is supported by an FPGA coprocessor that accelerates LUT (look up table), ROI (region of interest) and shading correction. This coprocessor significantly reduces CPU loading and enables complex image acquisition at high speeds.
Looking beyond the core camera features, the NEON-1040/1020 integrates a great deal of functionality that would have previously required a full embedded vision system. Most notably, the NEON-1040/1020 supports:
- A gigabit Ethernet (GigE) Vision 1 slave camera for two-camera applications
- Built-in PWM lighting control
- VGA output for easy connection with HMI or monitors
- 4x isolated inputs, 4x isolated outputs, USB 2.0 port, and RS-232 ports for connection with other factory systems such as motion controllers
As shown in Figure 3, these features enable the NEON-1040/1020 to take on many responsibilities that previously required separate hardware, significantly shrinking the system footprint and bill of materials (BOM). The reduced component count also simplifies maintenance and enhances reliability.
Figure 3. ADLINK’s solution cuts system footprint and bill of materials (BOM)
All in all, the ADLINK NEON-1040/1020 represents an important new option for developers creating mid-range vision systems. The cameras nicely fill the gap between smart cameras and embedded vision systems. They minimize system size and complexity while offering high performance and programmability. In my opinion, these cameras will be particularly appealing to developers who need extra features like a second camera or a built-in HMI. For these design scenarios, the NEON-1040/1020 may be able to eliminate a costly vision system, and are worth a closer look.
Contact featured member:
Solutions in this blog:
ADLINK is a Premier member of the Intel® IoT Solutions Alliance.
Roving Reporter (Intel Contractor), Intel® Internet of Things Solutions Alliance
Editor-In-Chief, Embedded Innovator magazine
Follow me on Twitter: ad@kentonwilliston