iCatch Technology: Ushering the Era of AI-Powered Image Processing

Several important landmarks have helped to shape the current automobile industry. One of it was the idea of installing a camera in an automobile, which was first displayed in the concept car Buick Centurion back in 1956 GM Motorama. Fast-forward it to today; cameras are a norm as almost every vehicle manufacturer provides it. However, with the advancement in technologies like deep learning, machine vision, and image signal processing, cameras are no longer just about providing reverse parking assistance or surround view or just straightforwardly capturing the images/videos via dash-cams. Instead, they need to act as a sensor that has excellent image processing capabilities to assist in providing intelligence for manufacturers working towards achieving full-autonomous vehicles, as well as things like collision, object, and driver drowsiness detection. Catering to the automotive imaging market demands since 2009 is iCatch Technology, a fabless IC design house. The company, with its expertise in image processing technology, combined with advanced video codec and system integration technology, has focused on the research and development of advanced digital imaging System-on-Chip (SoC) solutions. Besides, their team consists of talented and well-experienced imaging enthusiast, working since 1996 in the field, having excellent expertise in image signal processing, machine vision, IC design, and system integration. “We provide a complete high-performance image processing SoC solution to our customers,” says Charng Lee, Director, Strategic Planning at iCatch Technology.

Providing better intelligence through a camera requires a capable machine vision. However, machine vision still being at nascent stages is a significant roadblock for many manufacturers. Lee elaborates on three critical challenges under machine vision. First, there are no unified standards to transfer the captured image data from the camera to the electric control unit (ECU) of the vehicle, leading to the use of different standards to transfer. Second, the lack of edge intelligence in the camera modules that would enable pre-screening the information before the final image goes to ECU for processing, saving crucial time in case of collision detection where every nanosecond counts. Third, image processing working in conjunction with other sensors like LiDAR and ultrasonic, and the data generated from sensors becoming unified for better response time. iCatch Technology tackling challenges one at a time has successfully developed a series of automotive grade chip solutions, including CI1 for image signal processing (ISP) in camera module and CR1 for dashboard video recording. Both CI1 and CR1 have a flexible interface to handle different standards as well as if the client wants to port their algorithm, iCatch can assist them with their customization need as well. To tackle the edge intelligence problem, Lee mentions, “Our products leverage convolution neural network engine to provide intelligence at the edge and improve machine vision capabilities, enabling our customers to build better products for the end-users.”

iCatch Technology applying its years of experience in imaging to the automotive market, offers proprietary products—CI1 and CR1. CI1 is a 1080p automotive grade sensing camera module that can support up to two CMOS sensor having a high-performance image signal processor (ISP) that enables features such as high dynamic range (HDR), lens distortion correction (LDC), image dewarp and more. Additionally, having a neural processing unit (NPU) allows processing advanced driver-assistance systems (ADAS) functions, such as pedestrian detection (PD) and Driver Monitoring System (DMS), directly in the camera module. Their second offering, CR1, is an automotive grade SoC that is AEC-Q100 Grade 2 qualified. It can support up to two CMOS sensors, while ADAS functions supported by CR1 include Forward Collision Warning System (FCWS), Lane Departure Warning System (LDWS), PD, and DMS.

iCatch Technology also has started investing in new emerging technologies, such as 3D sensing and deep learning, for the next wave of opportunities in AIoT, industrial, and smart home. Additionally, with the commercialization of 5G in 2020, the communication infrastructure will be able to support new imaging applications, and the company will be prepared to become the pioneer of this revolution.