Air-Aware Crowd Density Prediction Using Hybrid CNN

Crowd density understanding serves as a fundamental requirement to maintain city safety and manage smart towns, as well as organize public gatherings. The traditional deep learning models perform density estimation through visual characteristics without considering environmental factors that control crowd movement patterns. The proposed hybrid CNN structure incorporates five air quality components, namely carbon monoxide (CO), particulate matter (PM2.5), nitrogen dioxide (NO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>), sulfur dioxide (SO<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</inf>), and ozone (O<inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">3</inf>) within a deep learning architecture for context-sensitive crowd estimation. Evaluation on ShanghaiTech data revealed that this model fulfilled training with 97.52% accuracy and validation with 97.57% accuracy to establish its strong predictive potential. The statistical outcome reveals precise density estimates through prediction errors where the mean squared error reaches 0.0043 and the mean absolute error reaches 0.0241. The predicted and actual density maps show a strong structural match because the Structural Similarity Index reaches 0.96. Environmental conditions demonstrate the most significant impact on human movement because PM2.5 particle concentration shows a relationship of r = 0.50 with crowd variations. This research creates a flexible framework for intelligent crowd analysis using deep learning and environmental intelligence, which can apply to urban planning, public safety monitoring, and event management operations.