IWF-ECTIC: Improved Wiener filtering and ensemble of classification model for tourism image classification
Tejaswini A. Bhosale,S. Pushkar
DOI: https://doi.org/10.1007/s11042-024-20452-z
IF: 2.577
2024-11-28
Multimedia Tools and Applications
Abstract:Social media is so widely used; travelers frequently post texts, photos, and videos about their experiences online. These user-generated content (UGCs) significantly affect how travelers perceive Tourist Destination Image (TDI) and directly influence their decision-making. UGC photos represent passengers' visual preferences for a certain place. Considering the significance of photographs, a lot of researchers have tried to analyze them using machine learning methods. Nevertheless, a limitation of research efforts employing these techniques for tourism image analysis is that they cannot precisely categorize the unique photos found in specific tourist destinations using predetermined images. In order to get rid of these challenges, this work proposes a Tourism image classification model termed as Improved Weiner Filtering and Ensemble of Classification model for Tourism Image Classification (IWF-ECTIC) model which includes steps like preprocessing, segmentation, feature extraction, and classification. The preprocessing is done via improved wiener filtering which provides a more robust assessment of image quality. The segmentation is done by the Hierarchical Density-Based Spatial Clustering of Applications with Noise (HDBSCAN) model which can effectively handle outliers and noisy points. Subsequent to the segmentation, MLGBPHS (Modified Local Gabor Binary Pattern Histogram Sequence), MBP (Median Binary Pattern), color, shape and Improved Entropy-based features are extracted in the feature extraction stage to reduce the dimensionality of the data. Finally, ensemble classification is done by combining the Multihead Convolutional Neural Network with Attention Mechanism (MHCNN-AM), Bidirectional Long Short-Term Memory (Bi-LSTM), and Recurrent Neural Network (RNN) classifiers to classify the images. To improve the classification accuracy, optimal training will be done in MHCNN-AM, Bi-LSTM, and RNN classifier using the proposed Cuckoo Updated Chimp Optimization ( CUCO) algorithm. The output scores of the three classifiers are fused with the improved score level fusion process to obtain the final classified outcome (Beach, Garden, Hill station, Historical monuments, National park, Temples). Finally, the performance of the proposed IWF-ECTIC model is evaluated over conventional models with regard to positive, negative and other metrics. The proposed IWF-ECTIC obtained high precision 96.87% and accuracy 97.42% at 90th learning percentage while existing models acquired least ratings. These models assess the perception of different tourist destinations and offer practical solutions across various aspects of the tourism industry, from enhancing user experiences to informing strategic decisions. It offers a strong framework for managing a variety of visual content in a range of tourism locations. Thus the combination of improved Wiener filtering techniques with ensemble models can lead to more accurate and reliable tourism image classification in classifying tourist photos and analyzing tourism destination images.
computer science, information systems, theory & methods,engineering, electrical & electronic, software engineering