Method for automatic monitoring selectively based in metadata of object employing analysis of images of deep learning

摘要

The embodiment relates to an automatic selection control method based on object metadata using deep learning image analysis. Specifically, this method obtains basic image analysis data based on deep learning by extracting object metadata from an image of a CCTV when monitoring is performed through a CCTV. In addition, when the object metadata is extracted, deep learning-based object classification is performed by classifying the CNN-based object from the extracted object metadata. Next, when the object is classified, the abnormal pattern is differently detected according to the extracted object metadata from the image distribution map of the object for each representative object calculated from the data in which the previous object appeared by the object metadata. So, when an abnormal pattern of object metadata is detected in this way, by assigning the priority of CCTV from the priority of one or more preset information among object information, information by purpose, information by location, and information by time period, the event is notified, It is characterized by notifying by control priority. Therefore, when controlling through CCTV through this, it increases the control efficiency by reducing the workload of the control personnel.