Moving object shape detection by fragment processing

Abstract

The development of information technologies related to the analysis of visual information is inextricably linked with the methods of extracting various features or objects to facilitate their further analysis. This is due to the growing demands for visual data from the user. At the same time exactly object detection is one of the most fundamental and challenging tasks in locating objects in images and videos. Over the past, it has gained much attention to do more research on computer vision tasks such as object classification, counting of objects, and object monitoring. At the same time, researchers almost never paid attention to the fact of the shape of a moving object, and usually left this question for further analysis. At the same time, for example, for an object classification, having an object with clear shape outlines as input would be useful. This study provides video fragment processing for moving object shape detection. Our approach is based on dividing each frame into fragments that allow the present image frame as a square matrix for a formal description. The rectangular video frame has been transformed into a square matrix by SVD (singular value decomposition), where each element is a Ky Fan norm value used as a descriptor. Scene changes in the frame will affect Key Fan norm fluctuations. Comparing the fragment norm changes with other fragment norm changes will allow us to assess how significant these changes are. If the norm value exceeds the threshold value, we can include this fragment as part of the moving object. By combining such fragments together, we will detect moving object shapes. The threshold is dynamic and depends on time. In this study, we paid attention to calculating a threshold value for a fragment's reliable identification of a moving object. We also note that the experiments were conducted for the case when there is a stationary camera (surveillance camera) and some moving object in the field of view. And in this case, it was possible to obtain a clear contour of a complex shape for a moving object. More complex cases of simultaneous movement of both the object and the camera will be considered later.