Nagarajan V, Marquardt B
JAN Scientific, Inc., Seattle, USA.
J Struct Funct Genomics. 2005;6(2-3):203-8. doi: 10.1007/s10969-005-1914-9.
Automatic imaging and scoring of crystallization drops is an essential step in high-throughput crystallography. Presently, white-light images of crystallization drops are acquired robotically and the images are analyzed and scored using pattern recognition algorithms. However, the scoring part remains unreliable as crystals and microcrystals are not always recognized by existing feature-extraction and recognition algorithms. We propose a fundamental shift in crystal monitoring through spectroscopic imaging of crystallization drops. This method converts the problem of automatic crystal detection from one of pattern recognition into one of intensity (concentration) analysis. The latter can be more robust and reliable.
结晶液滴的自动成像和评分是高通量晶体学中的关键步骤。目前，结晶液滴的白光图像通过机器人采集，然后使用模式识别算法对图像进行分析和评分。然而，由于现有特征提取和识别算法并不总能识别出晶体和微晶，评分部分仍然不可靠。我们建议通过结晶液滴的光谱成像对晶体监测进行根本性转变。该方法将自动晶体检测问题从模式识别问题转变为强度（浓度）分析问题。后者可能更稳健、更可靠。
