Abstract: The present invention generally relates to a technology of detecting a stationary object in a static video, i.e., a video which is generated by a camera with a fixed view, such as a CCTV video. More specifically, the present invention relates to a technology of identifying a stationary object by use of inference information (Tensor, Activation Map) of Deep Learning object detector among objects which have been detected in a static video such as a CCTV video. According to the present invention, it is possible to lower the priority of stationary objects (e.g., parked vehicles) in processing a static video so that computing resource of the video analysis system can be effectively utilized in the process after object detection so as to enhance the performance of image analysis.

Abstract: The present invention generally relates to a technology of accurately mapping a specific point in CCTV video onto the corresponding point on a map. More specifically, the present invention relates to a technology of mapping a specific point in CCTV video onto the corresponding point on a map through coordinate transformation between pixel coordinate of canvas plane of the CCTV camera and global coordinate of physical space for the video surveillance system. The present invention may provide an advantage of improving the efficiency of video surveillance system. When an urgent situation is found in CCTV video, the location in which the urgent situation is occurring may be accurately specified, which renders an effective action possible.

Abstract: The operation method of integrated control system according to an embodiment of the present invention comprises the steps of: providing an event detection request to the video analysis server; detecting an event by analyzing videos from at least one camera which are allocated to each video analysis server in accordance with default configuration; selecting at least one correlated camera corresponding to event-detecting camera by analyzing table information stored in the monitoring client as well as the event detection result; newly allocating the correlated camera to the at least one video analysis server; and detecting an event in each of the video analysis server by analyzing videos from the newly-allocated correlated camera.

Abstract: The invention generally relates to a method of providing random access for video data in which a P-frame of the video data is configured to be random accessible in addition to I-frames. More specifically, the invention relates to a method of providing random access for video data in which a P-frame is configured to be random accessible based on a random access reference frame which is provided via a separate channel such as header area of video data packets or a metadata file, whereby the random access for video data may be effectively provided with overcoming the conventional video random access technology depending on the I-frame only. The invention makes it possible to configure a P-frame random accessible simply by accompanying a random access reference frame of small datasize, whereby video compression ratio may be improved due to smaller numbers of I-frames being included in video data than conventional arts.

Abstract: Disclosed is a method of detecting a moving object in video produced in a plurality of CCTV cameras. In the present invention, a video reconstruction is applied to the CCTV storage video in order to significantly reduce the amount of video data, and then an image processing is performed on the reconstructed storage video in order to detect a moving object and further to extract metadata thereof. According to the present invention, staff members of the integrated control center may efficiently search any possible critical objects in the CCTV storage video.

Abstract: The invention provides a method of detecting critical objects out of video produced in CCTV cameras. More specifically, the invention provides a method of detecting critical objects out of CCTV video, in which lightweight algorithms of foreground segmentation and blob analysis are applied to the CCTV video in order to detect moving objects and then object filtering is performed based on distribution characteristics of metadata in order to propose critical objects contained in the CCTV video, whereby a few staff members in the integrated control center may efficiently monitor display screen of a large number of CCTV cameras. The invention makes it possible to detect critical objects out of CCTV video with utilizing lightweight algorithms, which enables real-time processing by video analysis servers of normal computing power. Therefore, the efficiency of integrated control centers may be improved where massive CCTV videos are gathered from a large number of CCTV cameras.

Abstract: The present invention relates to a caching method of multi-core graphic processing unit (GPU) for improving image processing performance by efficiently storing video data into the cache memory out of the global memory. One aspect of the present invention is to provide a caching method of graphic processing unit (GPU) having multiple cores wherein at least a part of pixels out of A*B pixels of video data are cached into a cache memory in order to perform image processing on k pixels of N*N size (where, k, N, A and B are natural numbers; k=N*N; A>N; B>N), the method comprising: grouping the at least a part of pixels out of A*B pixels into k pixel groups; mapping the k pixel groups to k cores of the GPU one-to-one basis by utilizing index information of each of the k pixels; and storing video data of the k pixel groups in the cache memory with reference to the mapping result.