Enhancing image categorization with the quantized object recognition model in surveillance systems

With surveillance systems using capturing and observing techniques, it is now much easier to identify things and individuals for security. However, variability in patterns due to device quality presents difficulties for these systems when attempting to categorize objects and individuals consistently. This research presents the Quantized Object Recognition Model (QORM), an approach to image categorization that seeks to minimize the effect of pattern fluctuations. The QORM uses a deep learning strategy to teach itself to identify and categorize various objects and people. Initially, quality equalizers are used for the segment-by-segment examination and differentiation of the patterns. The training inputs are compared with quantized segments based on characteristics like saturation and direction to identify objects and people. The non-classifiable images are identified by separate training on pattern variations that lead to errors. The difficulty of correctly categorizing objects and individuals in surveillance systems is discussed, and the Quantized Object Recognition Model is proposed as a solution. This study uses the Home Object 06 dataset, which has 10,000 images split into 100 categories of high-quality images. The system improves the accuracy by 8.06%, F1-Score by 9.01%, and sensitivity by 14.82%.