About Us
AccuChip Counting Device is a chip grain counting equipment manufactured jointly by CVIIL from New Jersey, USA, and MingCha VisionTech Company Limited from Hangzhou, China. The device counts the number of various grains on each product based on type by photographing and analyzing the images of wafers or samples. By applying artificial intelligence algorithms, the device can accurately count normal products and exclude defective products, such as the edge part of wafers, uneven cutting, etc.
The equipment consists of a sampling and photography module, a calculation module, and an intelligent software module. There are two ways to use the photography module: fixed photography and dynamic photography. Fixed photography takes a full-frame photo of each sample for analysis and counting, which is suitable for smaller samples. Dynamic photography, on the other hand, decomposes a large field of view into smaller fields of view, continuously takes photos, and then stitches the images into a larger one for analysis using algorithms. This method can capture large samples while maintaining high resolution. Both photography cameras are 25 million-pixel industrial cameras, and lenses with different focal lengths are configured depending on the required field of view size. The device also has an automatic focus function, making it quick and easy to switch lenses.
The calculation module is an image processing module that can process images with up to nearly one billion pixels and uses the latest deep learning algorithms. As a result, the device is equipped with the latest CPU and a GPU such as the Nvidia 3090 GPU, which is required for deep learning algorithms, resulting in fast and accurate target recognition and flaw elimination.
The calculation software is the core module of the device and was developed independently by our team. The software integrates multiple algorithms, including deep learning, object detection, image classification, pattern matching, morphological transformation, contour extraction, and combination matching, etc. to support the device's analysis of various products, including high-density products like LED chips and integrated products with multiple varieties of single crystals. Over the years, the software has been optimized through accumulated experience, leading to significant improvements in both human-machine interaction and algorithm efficiency, and is able to meet the various needs of customers.