AI Computing &System Architecture
Laser Die Alignment and Key Computing Carrier Technologies for Semiconductors
This technology combines an image-based alignment algorithm with an AI recognition mechanism. By leveraging FFT image preprocessing, contour extraction, and cluster filtering, it effectively locks onto alignment hole positions, supplemented by a CNN model for precise detection and point compensation, thereby significantly enhancing alignment stability and accuracy.
Combines high-precision image alignment algorithms, AI recognition, and high-speed computing to resolve sub-2um alignment issues relying solely on lasers, delivering high-precision positioning and intelligent inspection solutions for advanced packaging.
With the rapid growth in demand for advanced packaging and heterogeneous integration, high-precision overlapping alignment of laser dies within multilayer structures has become a critical technology. This technology combines an image-based alignment algorithm with an AI recognition mechanism. By leveraging FFT image preprocessing, contour extraction, and cluster filtering, it effectively locks onto alignment hole positions, supplemented by a CNN model for precise detection and point compensation, thereby significantly enhancing alignment stability and accuracy. Utilizing the AMD K262 high-performance computing platform, a single machine can simultaneously support 4-channel image processing, achieving real-time alignment at a resolution of 2000×2000 pixels to satisfy the demands for both high throughput and high precision. The overall solution offers the advantages of modularity and low-cost deployment, making it widely applicable to semiconductor laser die stacking processes, assisting the industry in improving yield and production efficiency, and accelerating the implementation of advanced packaging technologies.
