(Hong Kong, December 30, 2024) - Intron Technology Holdings Limited ("Intron Technology" or "Group"; stock code: 1760.HK), a leading provider of automotive electronics solutions in China, has Launched the Horizon Journey®6-based raw data fusion perception solution. The company's latest innovation combines BEV and Transformer technologies with a raw data fusion approach, effectively enhancing the precision and efficiency in radar-camera perception systems. This advanced solution is further optimized by integrating a 4D mmWave radar, a multi-camera system and sensor fusion algorithm, establishing a high-performance, low-cost, and ultra-high versatility intelligent driving perception fusion system.
As advanced autonomous vehicles are required to navigate increasingly complex driving environments, the number and variety of sensors equipped on these vehicles have steadily increased. This has resulted in a surge in data volume and diversity in data formates, placing even higher demands on perception solutions. With traditional perception solutions falling short of solving the object recognition problem, detection accuracy, and long tail problem, the overall cost of vehicle production has increased. Building on years of experience in software and hardware development, engineering services, and a thorough understanding of industry pain points, Intron Technology provides a robust and cost-effective perception system with its breakthrough raw data fusion perception solution. This achievement follows Intron’s self-developed "Autonomous Driving Vision-Radar Fusion Algorithm," which placed third globally in the nuScenes 'Camera, Radar' nuScenes Detection Score (NDS) authoritative evaluation in November last year, with an impressive high score of 64.3%.
Intron Technology's raw data fusion perception solution is based on the latest generation of Horizon's automotive-grade chip- Horizon Journey®6 series, which fully taps and utilizes the high computing power of the chip, supports large-scale parallel computing and efficient data processing capabilities, and is equipped with a 4D mmWave radar and a camera system that is built in accordance with the advanced BEV+Transformer technology model. BEV converts the perception method of 2D image perspective plus ranging into a feature map generated by the fusion of 360° space and time, plus the in-depth analysis of the Transformer architecture to output static information (such as lane lines, traffic lights, road edges, etc.) and dynamic information (such as pedestrians, two-wheelers, cars, etc.), and combines with time-series frames to realize path planning, obstacle avoidance and dynamic prediction and within seconds adjusts for surrounding vehicles, obstacles, and other factors to achieve more accurate environmental understanding and scene analysis capabilities. Moreover, BEV+Transformer reduces the complex fusion process between multiple sensors and eliminates the dependence on high-precision maps and lidars, making the overall sensing scheme simpler, cost-effective, and more versatile.
Intron Technology's self-developed RC_OCC model inference results
(From left to right: 6V image, radar point cloud, occupancy prediction results)
The traditional late-fusion method involves the independent data processing of each sensor to obtain the individual result – such as the object list, and then performs fusion. Due to the limited sensing capability of a single sensor, final decision errors under certain conditions and misdetection can easily result. Intron Technology's raw data fusion perception solution places the fusion step in front, maximizes the collection and processing of effective information through a neural network in the perception model, with less information loss, and makes full use of the multi-modal, centering on the mmWave radar and images. The complementarity between the sensor data improves the perception accuracy, and only a set of "soft and hard integrated" hyper-fusion perception algorithms are needed to achieve accurate perception and prediction in intelligent driving scenarios such as object detection, online mapping, and occupancy prediction, reducing the complexity and latency of the entire perception architecture. At the same time, via the continuous training and optimization of the mmWave radar data platform and deep learning model, the solution achieves rapid iteration and upgrade, and continues to improve the robustness and safety of the system.
Amid the "second half" of the automotive intelligence era, Intron Technology is integrating and developing advanced technologies such as artificial intelligence and big data with its strong R&D capabilities and outstanding productization capabilities. It is also breaking barriers with its eco-system partners in the "algorithm-driven + cost-probing" process. In this way, we will drive the industry in achieving continuous breakthroughs in intelligent driving, and accelerate the inclusive process of middle and high-end intelligent driving!
-End-
About Intron Technology Holdings Limited (Stock code: 1760)
Intron Technology Holdings Limited is a leading automotive electronics solutions provider in China focuses on providing solutions targeting critical automotive electronic components applied in New Energy, Automated Driving, Connectivity, Body Control, Safety and Powertrain systems. The Group leverages its engineering and R&D expertise, combined with advanced semiconductor devices, to offer industry-leading productized solutions for customers, thereby fostering the sustainable development of Automotive Industry in China. Intron Technology has been included in the MSCI China Index and the FTSE Global Equity Index Series ("GEIS").
Media Inquiries:
Strategic Financial Relations Limited
| Veron Ng | Tel: 2864 4831 |
| Iris Au Yeung | Tel: 2114 4913 |
| Lilia Yang | Tel: 2864 4833 |
|
Email: sprg_intron@sprg.com.hk |
|
