Automatic billing in retail stores demands identifying multiple items simultaneously in real time, which presents challenges due to factors like varying lighting during day and night, adverse environmental conditions, and dynamic perspectives. Similar shape and colour in pulses and cereals further confuse detection systems. Traditional object detection models perform well in controlled environments but struggle with inconsistent lighting, different distances, and movement. These issues cause delays and reduce customer satisfaction. To address this, the study develops a real-time automatic identification and classification system. Built on a customized YOLO-based pipeline with advanced pre-processing and augmentation, the model performs well under challenging lighting conditions. Under adverse conditions, the proposed model achieved 96.7% mean Average Precision (mAP), 94.8% precision, and 94.8% recall, showing an improvement of 2.1%, 2.6%, and 5.3% respectively over the baseline YOLOv9 model. This approach offers a scalable, cost-effective solution, enhancing retail efficiency and improving the customer experience.