Intelligent Crowd Density Classification Using Improved Metaheuristics with Transfer Learning Model on Smart Cities

As a part of smart city schemes, many governments are developed intelligent methods to improve the quality of living by offering them intelligent services to enhance sustainability, livability, and workability. These intelligent methods gathered massive data counts, provide more opportunities for governments for analysing these data, and create data-driven decisions. Video surveillance in smart cities offers efficacy in city activities, safe society, and enhanced municipal services. Crowd density analysis is a most commonly used applications of object recognition, but crowd density classifying approaches face difficulties including non-uniform density, inter-scene deviation, intra-scene deviation, and occlusion. Therefore, this study develops an Intelligent Crowd Density Classification using Improved Metaheuristics with Transfer Learning (ICDC-IMTL) model on smart cities. In the presented ICDC-IMTL technique, the major aim is to recognize various kinds of crowd densities efficiently. To achieve this, the ICDC-IMTL technique makes use of the Guided filtering (GF) approach to preprocess the input images. The ICDC-IMTL algorithm exploits a neural architectural search network (NASNet) model and the hyperparameter tuning process is done by the improved artificial flora algorithm (IAFA) for feature extraction purposes. At last, the multilayer extreme learning machine (MLELM) model is applied for the crowd classification into distinct types. The stimulation validation of the ICDC-IMTL approach takes place on a crowd dataset comprising four distinct classes. The series of experiments highlighted the betterment of ICDC-IMTL algorithm over other existing DL technique.