Process monitoring and control is an essential approach to improve additive manufacturing (AM) built quality. For the development of powder bed fusion (PBF) AM monitoring system, sensing process optical emission is a popular approach, as it provides rich information on melt pool condition which directly determines final part quality. However, the optical emission information is convoluted. And the lack of full understanding of it limits the further development of an optimal monitoring system. Therefore, the aim of this study is to explore the correlations between the optical emission and the processing condition to help enhance PBF process monitoring. A high-speed camera was used to acquire the images of the optical emission in the waveband of 800 nm – 1,000 nm. Several typical features were extracted and analyzed with the increase of laser power. The K-means clustering method was used to identify the hidden patterns of these features. The SVM model was built for quality identification. Five process patterns have been identified and therefore the collected dataset was partitioned into five subsets. The extracted features in each subset was characterized. It is found that plume area and plume orientation are the two most crucial features for processing condition monitoring. Number of spatters and spatter dispersion index are sensitive to some minor process vibrations which have little effect on built quality. Additionally, the time sequence information of the features can help improve the quality identification performance.