The rationale behind forensic science comparison methods is an important topic. The feature matching value framework proposed in 2022 is a new theory with competitive advantages, which may provide a scientific basis for feature comparison evaluation. In order to further clarify and enhance the scientific validity of this new theory, this paper studies its internal logic and scientific basis, and mathematically proves the form of feature matching value. Specifically, I carefully select the principles to be followed in the forensic science comparison method, establish them as a system of axioms (including 5 assumptions), and then derive the mathematical form of the concept of feature matching value from the axioms. Our work provides a scientific argument for the theory of forensic science comparison methodology, advancing from what it is to why.

With the intensive integration of deep learning and computer vision, a series of advanced technologies such as facial recognition, image (video) generation, and image classification, have made rapid progress. However, deep learning models are considered “black box models” due to their difficulty in explaining internal processes and predicting results, which poses a serious challenge to the interpretability of image evidence in the field of forensic science. Based on this, this review outlines an overview of interpretability issues based on deep learning. Emphasis was placed on the theoretical and methodological research on the interpretability of facial features based on deep learning both domestically and internationally, such as saliency maps method, perturbation-based method, and score/statistics-based method. Their applications in facial recognition and other related fields, especially in the field of forensic science portraits, were summarized. This review proposes the problems of facial feature interpretability methods based on deep learning models, and looks forward to the future development direction of facial feature interpretability based on deep learning.

The human ear possesses a complex three-dimensional structure and exhibits relatively high stability. Research on the morphological characteristics of ear images holds significant importance for individual identification in cases of occluded or incomplete facial images in videos and photographs. This paper categorizes the visible ear features in frontal and profile facial images into three levels: global feature, local feature, and detailed feature. Based on a self-compiled ear image dataset from the Institute of Forensic Science, Ministry of Public Security, P.R.C. (comprising 2078 individuals and 44826 images), the study further subdivides each feature type, providing illustrations and reference markers to depict the corresponding characteristics, thereby establishing a relatively comprehensive feature system. Using morphological comparison and statistical analysis methods, the frequencies of global features (36 feature items), local features (58 feature items), and detailed features were separately analyzed. The statistical results indicate that morphological features such as the earlobe and crus of the helix are the most frequently observed in ear images and remain relatively stable under varying lighting conditions, angles, and resolutions. Some features are relatively rare (e.g. multiple helical notches or an antitragus that is narrower at the top and wider at the bottom), and their presence can provide valuable support for individual identification.