Edge computing based artificial intelligence video surveillance system and its operation method

摘要

Disclosed is a method of operating a video surveillance system. The operating method of the video surveillance system includes, by an edge computing device, receiving a first plurality of image frames from any one of a plurality of IP cameras, and the edge computing device uses a background image frame to provide the first plurality of video frames. Detecting an object from image frames of, and selecting image frames in which the object is detected, the edge computing device calculating a first energy of the object from the selected image frames, the edge computing device Determining whether the calculated first energy is greater than an arbitrary threshold value, and when the calculated first energy is greater than the arbitrary threshold value, the edge computing device transmits the image frames in which the object is detected to a server. Includes steps. The edge computing device calculating the first energy of the object from the selected image frames includes the detection of the first energy in the X-axis direction and the Y-axis direction in each of two image frames among the selected image frames. Calculating accumulated pixel values of the object, the edge computing device setting intermediate values of the accumulated pixel values calculated in each of the two image frames as centers of gravity of the detected object, the edge computing device Calculating a difference value between the centers of gravity set in each of the two image frames, the edge computing device calculating a motion change rate of the detected object by dividing the calculated difference value by a maximum value of a motion vector, The edge computing device counts pixel values of the detected object in any one of the selected image frames to calculate the size of the detected object, and the edge computing device determines the size of the detected object and Calculating a difference value between the average object size, the edge computing device calculating a size fit of the detected object by dividing the difference value between the detected object size and the average object size by the average object size, the The edge computing device calculates a size ratio of the detected object in any one of the selected image frames by dividing the calculated size of the object by the resolution of the one IP camera, and the edge computing device And calculating a sum of the detected motion change rate of the detected object, a size fit of the detected object, and a size ratio of the detected object as the first energy of the object.