Advanced Deep Learning Model for Image Captioning Using Customized Vision Transformer with Global Optimization Algorithm