iCatch’s SoCs are powering a diversity of camera applications in
iCatch Join Forces with Academic to Showcase in ICIP 2019 an Intelligent Automotive Camera Solution for Both Viewing and Sensing
The iCatch’s Ci1 series supports automotive grade image processing with edge computing capability. It integrates a neural processing unit (NPU) to enable edge intelligence in camera module and to reduce the computational burden of the electronic control unit (ECU). Through real-time image analysis and computation, this NPU block can help the camera module to perform intelligent functions such as vehicle detection (VD), pedestrian detection (PD) and blind spot detection (BSD). Ci1 also has many advanced ISP features such as high dynamic range (HDR), lens distortion correction (LDC), motion-compensated temporal filtering (MCTF), and more functions which are essential to automotive imaging applications.
Professor Guo’s research group combines the embedded object detection and segmentation to develop a deep learning algorithm that is very suitable for ADAS functions including LDWS, FCWS/RCWS, and BSD. This new AI model reduces the model size and relaxes complexity of operation without accuracy degradation. Through implementing this new deep neural network object detection algorithm (NCTU SSD lite) into the NPU engine on Ci1 platform, it is possible to reduce power consumption and improve detection performance at the same time.
With the maturity of AI technology and the popularity of advanced driver assistance systems, image detection and recognition has been widely used in automotive applications. In addition to excellent image quality, instant object detection and analysis are also important. iCatch’s smart automotive imaging solution can provide the best image quality and smart edge intelligence together in the same camera device, which enables customers with more high-performance and cost effective solutions to choose.