The distinction between "embedded vision" and standard machine vision systems is sometimes not easy to discern. One approach to classification is to break them down into three parts.
Part 1: Includes classic vision systems with cameras and standalone PCs.
Part Two: Systems based on board-level cameras and application-specific hardware (such as small PCs).
Part Three: You will find a highly integrated system with a high degree of miniaturization and few or no standardized components. For example, in Parts One and Two, you would typically find camera systems using GigE, USB, shielded cables, etc., while in Part Three, you are more likely to find low-level interfaces such as LVDS with ribbon cables.
In other words, along the path from the first stage to the third stage, cameras become smaller and the number of standardized parts decreases. Cost savings are a result of embedded vision.
Compared to traditional PC setups, embedded methods not only save space and energy, but they can also be implemented at a significantly lower cost. A major factor contributing to the cost reduction of embedded systems is the software. For example, the Linux operating system and the OpenCV image processing library are open-source and freely available, so there are no licensing fees when using this combination.
Furthermore, the ARM-based processor family is constantly being upgraded. It is well-known for its affordable performance and availability across a wide range of performance levels, including those with multi-core architectures. ARM-based processes currently dominate the embedded systems market and are more prevalent than x86-based processors.
For SDKs running on ARM and x86-based architectures, the program code can usually be ported without spending a lot of time and effort. The reusability of already developed code can significantly save costs.
Hardware aspects of embedded vision include: System-on-a-Chip (SoC), System-on-a-Module (SOM), and Computer-on-a-Module (COM).
Processing boards used in embedded systems are typically platforms with x86 or ARM processors. The processors used here often integrate the graphics unit, bus system, and interfaces (USB, GigE, PCIe, etc.) into a so-called system-on-a-chip (SoC).
The next step in hardware integration is the use of a computer-on-module or system-on-module (COM or SOM, which can be used synonymously). The SOC, RAM, power management, and any other peripherals are combined on a circuit board into a module with plug connectors.
Reduce hardware development costs through system modules
Within the scope of hardware development for embedded applications, developers simply develop what is known as a carrier or substrate, and then use it to mount the SOM (Surface Mount Equipment) via suitable plug connectors. In short, this is the embedded processing board.
The advantage of this approach is that the most complex part of hardware development has already been done through the SOM. Connecting the SOM to the baseboard for external interfaces (USB, GigE, HDMI, etc.) is far more complex and less costly than a completely custom design of all the components required for development, for example, on a single circuit board.
Various SOMs with different SOCs (including x86 and ARM) are also available for industrial applications. Manufacturers typically design their SOMs to be compatible without requiring substrate adjustments, allowing lower-performance SOMs to be easily replaced with higher-performance ones.
Several manufacturer-independent standards have also been established, such as COMExpress, Qseven, and SMARC. However, in this case, SOM compatibility across different manufacturers' products typically only covers a portion of the SOM's functionality.
SOM (System-on-Oriented Design) makes developing embedded vision systems attractive, even in small unit volumes. While this fully custom design using the SOM approach is unlikely to reduce production costs, it still offers significant cost-effectiveness compared to traditional standard PC setups.
Disclaimer: This article is a reprint. If it involves copyright issues, please contact us promptly for deletion (QQ: 2737591964). We apologize for any inconvenience.
