Classification of foot images according to person, age and gender with the local phase quantization

Foot images are an important biological feature of the human body and carry various characteristics of people. The texture in the footprint, shape, length, etc. It may be possible to identify a person by looking at different qualities. Although the hand structure has its own unique shape and skin texture, comparison of these with foot biometrics reveals a complex situation. The main reasons for this include close toes, the absence of typical lines in footprints, and the high noise content of turned footprints. However, although these details are not similar to the hand, they cause differences in foot images for each person. In addition to these, foot images also differ according to age, gender, race, shoes and age of starting to wear shoes. In this study, 6944 data, which are right and left foot images of 100 people, were collected. The features of these collected files were extracted by local phase quantization. For each image file, 1x256 vectors were produced. These processes were performed for all files and images were classified for person, age and gender with many different classification algorithms. While 99.42% accuracy rate was obtained for person recognition, 99.87% success was achieved for gender. Finally, 98.14% classification success was achieved for age. All these results show that recognition from foot images is possible with high success with the method here.