In computer vision, detecting and classifying objects is of great importance in various fields such as transportation, healthcare, manufacturing and agriculture. On the other hand, in the field of transportation, there are different techniques are used to develop many applications related to pedestrian detection such as driver assistance system, intelligent video surveillance system, and emergency victim rescue. Therefore, this paper proposes an effective method for the problem of pedestrian detection from a drone-based images using Elephant Herding Optimization (EHO) to optimize the control and selection of the parameters and convergence speed in a deep learning architecture in order to improve finding a subset of features from a larger feature pool that provided better accurate classification and evaluation of pedestrian detection. The proposed method is evaluated by experimenting a number of pedestrian images obtained from VisDrone-Dataset and the results shows better and more effective detection and provide better and more effective classification in term of precision, 99%; recall, 98%; and F1 measure, 99%.

Keywords— Computer Vision, Object Recognition, Pedestrian detection, Deep learning, Elephant Herding Optimization (EHO)