Low power flagship products for embedded vision and AI applications
Shanghai, China, 3march2021 * * kangjiate, a leading embedded and edge computing technology supplier in the world, has launched a new low-power smarc2.1 computer module. The product appeared in the online exhibition of embedded world in Nuremberg, Germany, using NXPi.MX8MPlus Processor for industrial edge analysis, embedded vision and artificial intelligence (AI) applications. With machine learning and deep learning capabilities, this ultra-low power conga smx8 plus module enables industrial embedded systems to observe and analyze the surrounding environment, realize scene perception, visual detection, identification, monitoring, tracking, as well as gesture based contactless machine operation and augmented reality (AR) functions.
This platform with armcortex-a53 quad core processor has a number of technical highlights, including an integrated neural processing unit (NPU) for improving AI computing power, and an image signal processor (ISP) for parallel real-time processing of high-definition images and video streams generated through two integrated mipi-csi camera interfaces. This new smarc module has complete supporting facilities, such as prefabricated 3.5-inch carrier board, Basler camera, and AI software stack support, which provides verification for the concept of rapid product launch. This kind of low-power video and AI function module is only the size of credit card, and its application fields can be seen everywhere, including intelligent agriculture, industrial manufacturing, retail, transportation, intelligent city, intelligent building and so on.
“Engineers can make use of the rich and efficient functions of the new smarc module and our perfect product ecosystem, and realize the application-oriented functions through pciegen3, two USB3.0 and two SDIO interfaces, so as to build a 2-6w low-power platform for visual and AI applications, and ensure excellent reliability and stability,” said Martin danzer, product management director of kangjiate. According to different models, the new module can adapt to a wide temperature range of – 40 ℃ to 85 ℃
The new smarc module based on i.mx8mplus processor has a variety of special processing units, which can achieve amazing embedded visual and AI computing response speed with extremely low functions. Its advantages include:
·In addition to four high-performance multi-purpose armcortex-a53 processor cores, the neural processing unit (NPU) adds 2.3tops of special AI computing power.
·The integrated image signal processor (ISP) can process three high-definition video streams of 60 frames per second to improve the video quality.
·High quality digital signal processing (DSP) can realize local speech recognition without connecting to the cloud.
·Cortex-m7 processor provides real-time control function and an Ethernet port with time synchronization network function. It can also be used as a fault insurance unit.
·In addition to the encryption module (CAAM) for hardware accelerated ECC and RSA encryption, the armtrustzone scheme also integrates a resource domain controller (RDC) for independent execution of key software, and a secure HAB (high assurance boot) startup mode to prevent unauthorized software from being executed during startup.
The new smarc2.1 module for vision and AI applications has four different 4 cores NXPi.MX8MPlus All processors are based on armcortex-a53 architecture, which can adapt to industrial environment (0 ℃ to + 60 ℃) and even wider temperature range (- 40 ℃ to + 85 ℃). Meanwhile, they integrate ECC function and support up to 6GB of lpddr4 memory. The module can drive up to three displays at the same time, and has hardware accelerated video decoding and encoding functions (including h.265). It can directly receive HD camera video streams from two integrated mipi-csi interfaces. In the aspect of data storage, it can configure up to 128e MMC on board and run in the secure pslc mode. Peripheral interfaces include 1xpiegen3, 2xusb3.0, 3xusb2.0, 4xuart, 2xcanfd and 14xgpio. In the aspect of real-time network, the module has 1xgb interface and supports TSN technology and traditional GB Ethernet. The module can be equipped with onboard m.2 wireless network card and Bluetooth le for wireless connection. In terms of sound, there are also 2xi2s interfaces. The operating systems supported by this module include Linux, yocto2.0 and Android.