Metal additive manufacturing (AM) processes have transitioned from rapid prototyping applications to industrial adoption owing to their flexibility in product design, tooling, and process planning. Directed energy deposition (DED) is one of the most commonly used metal AM processes capable of producing large, high density parts, with a controlled microstructure. However, there are still ongoing challenges in maintaining a high level of reliability and repeatability when compared to conventional manufacturing processes. There is a need to define, identify and maintain regions of process stability in DED. In this study, a high-dynamic range camera and a physics-based model are used to monitor the melt pool, obtain process signatures, and predict deposition stability characteristics. The research efforts are focused on generating process maps to identify unstable process zones, with a reference to process physics, process signatures, and process outcomes using analytical modeling, in-situ melt pool monitoring, and ex-situ characterization, respectively. The goal is to classify the process signatures in pre-defined process zones (under-melt, conduction, keyhole, balling) to avoid instabilities, defects and anomalies using a low-cost high-dynamic range camera and kNN classifier, which has achieved 13% error rate. With this approach, decisions can be made to perform corrective actions (e.g. machining, re-manufacturing) or to scrap the manufactured part without ex-situ characterization.