An Analysis Approach for Crowds with Varying Density and Complexity Using Image Data

Smart cities represent urban environments that leverage advanced technologies to optimize infrastructure and services. Crowd density analysis, a critical component of smart cities, employs sensors, cameras, and data analytics to monitor population density and movement in public spaces. Deep Learning (DL) techniques enable precise and automated crowd density evaluation, transforming traditional methods. This study introduces CILHOF-CDC, a CLAHE-InceptionV3-LSTM Hyperparameter Optimization Framework for Crowd Density Classification. The framework enhances image contrast using the CLAHE method, employs Inception V3 for feature extraction, and integrates a Bidirectional LSTM model for crowd density detection and classification. Reinforcement Learning (RL) is utilized for hyperparameter optimization to further improve system performance. Experimental evaluations on a crowd density image dataset demonstrate the efficacy of the proposed framework, achieving an impressive accuracy of 99.4%, outperforming recently developed DL models. Furthermore, comparative analyses with existing models in the literature reveal that CILHOF-CDC achieves superior performance in terms of classification accuracy and adaptability across diverse crowd density scenarios.