Humans yearn for the starry sky, just like children yearn for the world outside their homes.
In addition to the planet Earth, there is another world in the universe. The universe is divided into a broad sense and a narrow sense. The broad sense of the universe refers to the general term of all things, and is the unity of time and space. The universe in a narrow sense refers to the space and matter beyond the earth's atmosphere. Exploring the universe can help us discover more cosmic laws.
Dawn Computing Service Explores the Dreamy and Mysterious Diffuse Galaxies
Low surface brightness galaxies are diffuse galaxies that are dreamy and mysterious. When observing these galaxies from the earth, in the environment of the night sky, their surface brightness is at least one magnitude lower than the surrounding background skylight. Searching and studying them will help humans better understand the structure and evolution of the Milky Way.
How to search for low-level galaxies in massive astronomical images? Recently, the expert group of Shandong University plans to build an efficient search platform for galaxies with low surface brightness through the support of Shuguang Computing Service, so as to accurately find candidates for galaxies with low surface brightness from T-level astronomical images within one day, and provide support for the research of astronomers and data processing experts.
The establishment of an efficient search platform will help improve the search efficiency of low-surface brightness galaxies in massive images, and also provide important reference and demonstration for the data processing work of large-scale domestic and foreign sky survey projects in the future.
The detection speed reaches within 1 second for each picture
The universe is infinite. For astronomical research, the amount of data generated is huge. At the same time, the resolution of astronomical images is very high, which will generate a large amount of unstructured data. The detection efficiency will directly become a constraint for research.
On the other hand, the Shandong University expert group has conducted preliminary research on the search for galaxies with low surface brightness. The target detection model based on the deep learning framework was built, and the performance was compared with the existing target detection models R-CNN, YOLO, etc. Research experiments have shown that deep learning-based object detection methods are very effective for searching low-face brightness galaxies from astronomical images. The built deep neural network model has very high requirements on the performance of the running equipment.
At present, the detection efficiency of the model is relatively low, and it takes about 4 seconds to detect a picture. The amount of astronomical photometry data is huge, and large-scale sky survey projects in the future can obtain massive image data of the order of TB or even hundreds of TB every day, and there is an urgent need to use computing service platforms to improve retrieval efficiency.
By deploying the Sugon computing service platform to improve detection efficiency, the detection speed will reach within 1 second for each picture. It is estimated that with the support of the Sugon computing service platform, all 930,000 images of the digital sky survey will be tested, and the results will be returned within 24 hours.
For a long time, we have thought that only the earth has life, and human beings are the most advanced and intelligent life forms in the universe, and exploring the universe may help us better understand the emergence and existence of life. The Sugon Computing Service will provide support for astronomical research and help humans better "look up at the starry sky".
