Weber HsuPresident, iCatch Technology
Today, companies want to design an easy-toinstall outdoor wireless camera that requires long battery life, fast response time, fewer false alarms, and good image quality under low light and high contrast conditions.
Today, companies want to design an easy-toinstall outdoor wireless camera that requires long battery life, fast response time, fewer false alarms, and good image quality under low light and high contrast conditions. To preserve battery power, the outdoor camera usually needs to be in standby mode and wake up immediately when an event occurs. Thus, the fast response time is desired to inform the end-user immediately of an event through a wireless internet connection. However, there will be some chance that it is a false alarm due to light flashing or just a pet passing by. So, it will be good and can save power if there is an intelligence that can filter out these false alarms. On the other hand, good image quality under severe lighting conditions is always welcome.
iCatch has accumulated extensive experience in low power chip design and software management of portable and wearable camera devices for international brands. Throughout the integration of real-time operating system (RTOS) and hardware boot-up sequence, they can achieve a high-speed system boot-up time (typically 0.2 seconds) and low power consumption for the outdoor home security cameras. Furthermore, by adding AI edge computing capability, the iCatch chip can perform deep learning neural network processing acceleration to achieve real-time detection of pedestrians, intruders, pets, and parcels, etc. This will enhance the accuracy and effectiveness of the alarm and consequently save the communication bandwidth and storage capacity of the useless events. Their solution has also been equipped with enhanced noise filtering for low light conditions and high dynamic range (HDR) processing for the high contrast backlight conditions to restore high-quality images under these severe lighting environments.
iCatch has invested a lot in developing new technologies in multiple sensing and deep learning neural networks. They have created a 3D depthsensing solution by combining TOF (time-of-flight) depth image with RGB color image together so that the users can utilize the depth information associated with color images at the same time to explore better processing performance for complicated situations such as people counting in a region or distance measurement for access control. The same technology is also applied to process RGB and IR (infrared) images at the same time for automotive in-cabin monitoring. With this multiple sensing technology, iCatch is now working toward the fusion of radar signal with color image to support healthcare and automotive applications.
" iCATCH TECHNOLOGY HAS INVESTED A LOT IN DEVELOPING NEW TECHNOLOGIES IN MULTIPLE SENSING AND DEEP LEARNING NEURAL NETWORKS "
The investment in deep learning includes three parts, hardware engine, algorithm, and software tools. iCatch invested in deep neural network acceleration hardware design by cooperating with a third-party partner to integrate a neural processing unit (NPU) into the image processing chip design. Their algorithm team survey, optimize and train the selected convolutional neural network models to fit into the NPU hardware to achieve a real-time inference of various detection and recognition tasks for vehicles, pedestrians, objects, and biological features such as the face.
iCatch differentiates its solution from the competitors by focusing on wireless and battery-powered home camera applications. The solution is targeted for the emerging cloud-based DIY home security product and service companies, such as Arlo, Ring, Nest, SimpliSafe, and the like, not for commercial or business installation markets. Based on their unique features of superior image quality, low power consumption, fast boot-up time, and better alarm accuracy by AI processing, their solution can extend the camera’s battery life and enable twoway communication between the device and the end user’s mobile phone. Furthermore, the iCatch solution is based on a structuralized software stack. The customer can easily build their application on top of it through the software development kit or by iCatch’s customization service. The customers can create their highly differentiated solution by and on the same development platform.
Additionally, iCatch also provides a cloud-based application software reference kit and development service to help the customers to build their cloudbased applications quickly with enhanced video quality and optimized communication performance for the end users of their intelligent home or doorbell camera devices.
The NPU hardware is inserted into their chip with an architecturally optimized design. iCatch employed a plus-AI strategy. That means the AI edge computing engine is carefully integrated into their image processing SOC in both hardware and software aspects. So that the NPU engine can perform image and video analytics as well as detection and recognition processing very effectively. The NPU engine executes the inference computation software of deep convolutional neural network models and generates the detection results for the system’s further decision processing.
iCatch’s market in security camera and AIOT applications has just started. The market will expand as they acquire more customers through the marketing campaign and their current customers open new projects. For the days to come, iCatch plans to promote its solutions to more smart-home camera brand names (OEM) in North America and Asian areas, such as Korea, Japan, and Taiwan, by supporting the ODM manufacturers in Taiwan and Asian regions. Furthermore, they plan to expand the application of their intelligent image processing SOC solution to not only security and automotive, but also AR/VR, consumer wearable, and industrial machine vision applications. iCatch plans to elevate the solution offer of the next generation by supporting more variety of image sensors, such as audio, radar, and neuromorphic vision imager and adding more AI edge computation capability to perform more intelligent detection functions while preserving privacy at the edge device without sending sensitive personal information to the cloud. This is beneficial to home security and surveillance application as well as medical and healthcare monitoring for elderly people.
iCatch Technology has focused on image processing technology and SOC solution development since its establishment. For many years, the image processing system chips have been used in consumer digital still camera (DSC) products of the world’s famous brand names. By gradually adding AI computing capability into their chip solution, the customer base has also migrated to the smart home security and automotive imaging device manufacturers and global brand names. At the same time, the DSC market declined rapidly over the last few years. A reputation of high image quality and value-added customization service was established over the years to help customers differentiate their products and service from others.
In addition to the security and AIOT camera applications, iCatch’s intelligent image processing SOC solutions can also be used in automotive monitoring applications, such as dashboard camera and recorder, driver monitoring system (DMS), occupancy monitoring system (OMS), electronic rearview mirror monitoring system, etc. Furthermore, iCatch’s solution is also very suitable for auto-framing (center-stage) high-resolution conferencing camera applications for teleworking, remote education, telemedicine, as well as other people-centric video applications. Their vision is to enhance the convenience of human life through the ever-improving image, multiple sensing, and video processing technologies with intelligence.