Person Recognition Using Ear Images Based on Fractional Gannet Sparrow Optimization Enabled Deep Learning

Ear image-based person recognition is one of the technologies that has garnered more attention in recent years. Person recognition is potentially attributed to the advancement of capabilities and capturing techniques. Developing effective methods for ear recognition can assist in improving the efficiency of person recognition as ear images can be retrieved in a nonintrusive and contactless manner. Therefore, the paper proposes an innovative method based on deep learning for person recognition with ear images. At first, the input ear image is applied to image pre-processing by applying an Adaptive NonLocal Means (Adaptive NLM) filter. Consequently, feature extraction is executed, where features, such as Pyramid Histogram of Oriented Gradients (PHOG), Local Binary Pattern (LBP), Speeded Up Robust Features (SURF), and Scale-Invariant Feature Transform (SIFT)are extracted. At last, ear recognition is done by the Deep Maxout Network (DMN) tuned with the Fractional Gannet Sparrow Optimization (FGSO). Furthermore, the FGSO-DMN is evaluated by applying metrics, like f-measure, precision, and recall. The FGSO-DMN is found to record a maximum value of precision at 0.947, f-measure at 0.956, and recall at 0.966.