May 9, 2023
Shoeprints have played an important role in criminal investigation for centuries. Yulin Ma(1906~1981), the most prominent shoeprint investigator in modern China who assisted the police in solving numerous large cases involving shoeprints and established himself as a pioneer in the study of shoeprint material evidence. Similarly, inScotland, shoeprints were utilized to solve a case in 1786, as evidenced by written and pictorial records.
In the past few decades, there has been a significant development and refinementin the application of shoeprint evidence. Researchers around the world are exploring various ways to maximize the potential of this type of evidence insolving crimes. The development and application of footwear impression evidence can be divided into four stages.
During this stage, physical evidence related to footwear impressions was stored in physical objects or captured through photography. However, this method of preservation was both space- and labor-intensive, as well as inconvenient to use.
Over the years, material evidence and documents were classified and stored according to sole patterns.
In order to preserve and compare footwear impression evidence more efficiently, a software system has been developed for the digital preservationand retrieval of footwear impression evidence. This system allows for the storage of footwear impression images and classification of footwear impression patterns according to their types, enabling quick retrieval of footwear impression patterns.
There are several software companies in the world that have developed footwear impression comparison systems to meet the needs of forensic and criminal investigators.These systems are widely used by forensic and criminal technical identification institutions globally.
The advantages of the following system include lower resource consumption and convenient data entry,making it suitable for the preservation and retrieval of a small number of footwear impression data.
However, the system also has its limitations. Different people may use different ways to describe the shoeprint patterns, which could lead to inaccurate retrieval of pattern types. Additionally, when there is a large amount of data, it takes aconsiderable amount of time to view the search results after each query.
Sol epatterns can be described through graphics and text, and can also be represented by pattern characteristics. After a simple processing, the pattern features of shoeprints can be displayed as a binary image. The shoeprint pattern comparison algorithm can then be used to extract the features from the binary image.
Once the sole pattern image has been binarized and the pattern features are extracted, the comparison can be performed using the pattern features, and the results can be sorted and displayed by similarity. After years of research and development, the algorithm has now been optimized , running in the system that has been used worldwide.
The feature comparison algorithm has the advantage of fast image feature comparison speed and the ability to sort comparison results based on similarity, making it convenient to use. However,the disadvantage is that each shoeprint requires processing of the binary image, and the quality of the binary image processing can have an impact on the accuracy of the query results.
In recent years, fully automated image matching algorithm has been widely appliedin various fields. In the field of shoeprint pattern matching, the image matching algorithm has also been adopted to achieve automatic matching of shoeprints to shoe models in the database.
The fully automated image matching algorithm utilizes big data and artificial intelligence to train a deep neural network on large shoeprint pattern dataset,which can effectively extract the highly abstract features of the shoeprint pattern. This process avoids the complex procedure of manually drawing the binary image, and directly extracts the features of the original image for matching, thus further improving the accuracy of the matching results compared to the binary image features.
The advantage of an automatic matching algorithm is the ability to automatically compare shoeprint patterns without human intervention. However, a disadvantageis that it can consume a large amount of computational resources.
[1] Forensic Footwear Evidence, page 3,
William J.Bodziak
[2] HOBBIT IMAGING SOLUTIONS
URL：https://hobbit-is.nl/?lang=en
[3] Deep multilabel CNN for forensic footwear impressiondescriptor identification
Marcin Budka a, Akanda WahidUl Ashraf a, Matthew Bennett a, Scott Neville b, Alun Mackrill b
URL：https://www.sciencedirect.com/science/article/pii/S1568494621004191#b5
.png){kind=small}
