DETECTION OF SMALL OBJECT – CRITICAL SURVEILLANCE

Abstract

Identifying small things in photos is a difficult yet crucial undertaking in areas such as surveillance, medical imaging, and automated inspection systems. This thesis presents a novel approach for detecting small objects by utilizing texture segmentation techniques that rely on Gray Level Co-occurrence Matrices (GLCM) and Gaussian Mixture Models (GMM).The first step in our methodology involves preparing the photographs, which includes resizing and converting them to grayscale in order to ensure uniformity. Subsequently, the images are partitioned into smaller subimages to permit meticulous texture analysis. The GLCM technique is utilized to extract important texture properties, including contrast, dissimilarity, correlation, and energy, from every subimage. These characteristics offer a thorough depiction of the textures that are present. Subsequently, a Gaussian Mixture Model employs texture features to cluster the subimages, therefore efficiently dividing the image into sections with clearly distinguishable textures. Clustering is utilized to identify and separate little things that are easily distinguishable from their surroundings because of their distinct textures. The efficacy of the suggested methodology was assessed using a compilation of synthetic and real-world photographs. The findings demonstrated the efficacy of our approach in accurately detecting and separating minute entities, even in intricate settings characterized by diverse textures. The findings indicate that this approach has the potential to be used in various applications, providing ample opportunities for further exploration and advancement in the fields of image processing and object detection.