Robust Object Recognition under Varied Circumstances using Machine Learning Algorithms

Object Recognition is one of the areas of computer vision (CV) applications that has advanced significantly in recent years. Finding a particular object from numerous things present in a scene is one of the greatest issues that exist in object detection. Thus, there is a need to develop a method for detecting and recognizing objects in various environments. The developed model can be integrated into three stages, namely pre-processing, feature extraction, and classification. Initially, the adaptive probabilistic Weiner Kuwahara filter (Pr-WeKUF) is introduced to eliminate the noise that exists in the image. In order to obtain the main characteristics present in the image, a hybrid squeeze excitation multi-attention transformer with Bi-GRU model (Hy-GrpSMUT) is utilized. Adaboost, random forest and extreme gradient boost have been introduced to classify objects in different environments. The performance of the proposed model is evaluated using two datasets: RTTS and PascalVOC2007. The performance study is designed to demonstrate the effectiveness of the proposed model. The experimental results show that the RTTS dataset achieves high accuracy, precision, recall, F1-score, and specificity of 98.8%, 98%, 98.2%, 98.12%, and 98.18%, respectively. Then, the PascalVOC2007 dataset achieves high accuracy, precision, F1-score, recall, and specificity of 98%, 97.9%, 97.9%, 97%, and 98.4%, respectively.