No abstract

No abstract

Objectives:To identify a suitable object recognition method for video surveillance systems, especially in traffic monitoring, to track and sense multiple objects and classify them by employing conventional algorithms in order to boost accuracy. Methods: Neutrosophical Background Subtraction (NBS) and Grasshopper Optimized Convolutional Neural Network (GO-CNN) are employed to detect and track the objects in real time. The Neutrosophical Fuzzy Background Subtraction Method (NFBSM) is utilized to segment moving objects from the background in terms of truthful degree, false degree, and in-deterministic degree. CNN-based object tracking techniques, along with the grasshopper optimization algorithm (GO), are used for reliable object detection and hyper-parameter fine tuning. The NFBSM+GO-CNN model extracts the features from the dataset (collected from YOLO-V-3 from Kaggle) and selects a suitable function to perform object detection and classification of moving vehicles with a high accuracy rate. Four main features are used such as colour, texture, motion, and shape. PYTHON is used for implementation and comparative analysis, and the findings are compared with baseline models D-CNN and CNN-MOD. Findings: The proposed NFBSM+GO-CNN method outperforms with 98.2% accuracy, 96.3% improved precision, and 96.8% F-measure, which is comparatively higher than the baseline models in terms of real-time object detection and classification. Novelty: The outcome of NFBSM+GO-CNN clearly shows that the proposed model has the ability to detect the object in real-time and classify the vehicle in road traffic monitoring with a high accuracy rate. The proven results outperform existing models such as D-CNN and CNN-MOD.

No abstract