Hidden and Unknown Object Detection in Video

R. A. Simonyan

Object detection is applied to find such actual objects as faces, bicycles and buildings in images and videos. The algorithms executed in object detection normally use extracted features and learning algorithms to distinguish object category. It is often implemented in such processes as image retrieval, security, surveillance and automated vehicle parking system.Objects can be detected through a range of models, including Feature-based object detection, Viola-Jones object detection, SVM classification with histograms of oriented gradients (HOG) features, Image segmentation and blob analysis.For detection of hidden objects in the video the Object-class detection method is used, in which case the object or objects are defined in the video in advance [1][2].The proposed method is based on bitwise XOR comparison [3]. The method (system) detects moving as well as static hidden objects.The developed method detects objects with great accuracy; it detects also those hidden objects which have great color resemblance to the background images, which are undetectable for a human eye. There is no need to define or describe the searched object before the detection. Thus, the algorithm does not limit the search of the object depending on its type. The algorithm is developed to detect objects of any type and size. It is calculated so to work in case of weather change as well as at any time during a day irrespective of the brightness of the sun (which leads to the increase or the decrease of the intensity of the brightness of an image); in this way the method works dynamically. A system has been developed to execute the method. Object detection is applied to find such actual objects as faces, bicycles and buildings in images and videos. The algorithms executed in object detection normally use extracted features and learning algorithms to distinguish object category. It is often implemented in such processes as image retrieval, security, surveillance and automated vehicle parking system.Objects can be detected through a range of models, including Feature-based object detection, Viola-Jones object detection, SVM classification with histograms of oriented gradients (HOG) features, Image segmentation and blob analysis.For detection of hidden objects in the video the Object-class detection method is used, in which case the object or objects are defined in the video in advance [1][2].The proposed method is based on bitwise XOR comparison [3]. The method (system) detects moving as well as static hidden objects.The developed method detects objects with great accuracy; it detects also those hidden objects which have great color resemblance to the background images, which are undetectable for a human eye. There is no need to define or describe the searched object before the detection. Thus, the algorithm does not limit the search of the object depending on its type. The algorithm is developed to detect objects of any type and size. It is calculated so to work in case of weather change as well as at any time during a day irrespective of the brightness of the sun (which leads to the increase or the decrease of the intensity of the brightness of an image); in this way the method works dynamically. A system has been developed to execute the method.