Bio-Inspired Optimization with Transfer Learning Based Crowd Density Detection on Sparse Environment

Crowd density estimation is a major importance for applications including crowd control, public space planning, autonomous driving, visual surveillance, and warning visually distrait drivers previous accident. With strong scale, reflective, and translational symmetry, techniques to estimate the density of the crowd yield a promising outcome. But dynamic scenes with constantly evolving spatial and temporal domains and perspective distortion yet have difficulties. The dynamic nature of scenes and the complexity of demonstrating and integrating the feature space of objects of different magnitudes as predictive prototypes are the primary reason for this. This manuscript presents a Red Fox Algorithm with Transfer Learning based Crowd Density Detection (RFOTL-CDD) technique in Sparse Environment. The purpose of the RFOTL-CDD system lies in the effectual identification and classification of distinct types of crowds in a sparse environment. To achieve this, the presented RFOTL-CDD method uses a ResNet prototype for feature vector generation. For the identification and classification of a crowd, the RFOTL-CDD technique applies Naive Bayes (NB) classifiers. In this work, the RFO algorithm is utilized for boosting the performance of the ResNet method. The stimulation outcomes of the RFOTL-CDD technique can be well studied on a crowd dataset and the outcomes confirmed the supremacy of the RFOTL-CDD technique on crowd detection.