Human Tracking System Using Optimization with Convolution Networks

With real-world applications in robotic vision, scene analysis, surveillance, image compression, medical imaging, and augmented reality, real-time Human Tracking Systems (HTSs) have emerged as a significant area of study in computer vision and image processing. To lessen the computational load and increase the accuracy of HTS, we present in this study a Faster Region-based Convolutional Neural Network coupled with Crow Search Optimization (FR-CNN-CSO). Additionally, to further enhance performance, an optimization strategy was combined to improve trajectory assessment, reduce individuality switches, and update functions according to environmental conditions. Moreover, the hybridized framework has provided an effective solution for improving tracking accuracy and compactional efficiency compared to conventional models. The suggested system is created using video input in a Python environment. Redundant data is eliminated from the gathered datasets during preprocessing. Histograms of Oriented Gradients (HOG) are then used for feature extraction. The extracted features are then fed into the FR-CNN model, which uses crow search optimization to effectively identify and track individuals. This strategy aims to reduce execution time and computational complexity while achieving excellent forecast accuracy. According to experimental data, the suggested approach performs better than traditional methods in terms of execution time, accuracy, precision, recall, and F-measure. The proposed FR-CNN-CSO algorithm illustrates a possible application in autonomous systems, smart surveillance systems, and provides a balance between adaptability and robustness.