Vehicle Type Detection and Classification Using Improved Relief Feature Selection Algorithm and Bidirectional Gated Recurrent Unit

Nowadays, Traffic density estimation can be utilized to control traffic light signals for managing the traffic effectively. Vehicle Type Classification (VTC) is used in this traffic density. Because of the factors like minimal interclass variance and maximum intraclass variation, the VTC is challenging which is handled by proposed Improved Relief Feature Selection Algorithm and Bi-directional Gated Recurrent Unit (IRFSA-Bi-GRU) method. Bi-directional Gated Recurrent Unit (Bi-GRU) model is proposed for enhancing Vehicle Type Classification (VTC) within Intelligent Transport Systems. Multiview In Out Traffic Complexity Database (MIOTCD) is used for data collection then a camera response model is used for enhancing visual data quality. Kalman filtering and Gaussian mixture model techniques are utilized for vehicle recognition and tracking. Local Binary Pattern (LBP), Steerable Pyramid Transform (SPT), and Dual Tree Complex Wavelet Transform (DTCWT) are used for feature extraction and Improved Relief Feature Selection Algorithm (IRFSA) is used for feature selection. Finally, the Bi-directional Gated Recurrent Unit (Bi-GRU) model is used for classifying 11 vehicle categories on MIO-TCD. The obtained results exposed that the IRFSA-BiGRU method outperformed the existing methods by achieving an accuracy of 99.02%, 98.34% F-score, 98.78% MCC,99.93% specificity, 99.32% NPV, 98.75% precision, 98.14% recall on the MIO-TCD database.