Asst Prof Dr. Charturong Tantibundhit, Faculty staff at the Department of Electrical and Computer Engineering, Thammasat School of Engineering (TSE) has revealed that a TSE student has developed an innovation called “Automated Firearm Classification from Bullet”.
This innovation is the world’s first application that can inspect the bullet at the crime scene and display an accurate result using artificial intelligence (AI) technology.
The “Automated Firearm Classification from Bullet Markings Using Deep Learning” project is developed by a student and is advised by Asst Prof Dr. Charturong. The research also received cooperation from a Ph.D. student, Faculty of Science of Chulalongkorn University, the Royal Thai Police, and the Institute of Forensic Science.
Thailand has a relatively high gun crime rate with an average of 30,000 – 40,000 cases per year. Normally, in an investigation, the officer collects the physical evidence, the bullet, from the crime scene and sends it to the lab. There, the specialist checks for the markings on the bullet, and pass the result back to the officer who then uses the result for further investigation and case summary. The whole process takes around 30 days and has to be done in the special lab with expensive specialized equipment and the specialist. There is not a lot of the specialist in Thailand which make the whole process a lot harder.
This work aims to speed up the investigation process with AI, allowing the officer to process the class characteristics of the bullet at the crime scene with only a mobile phone and custom-built portable hardware. The officer can put the bullet into the hardware, and then use the application on the phone to take a 360 degrees panorama pictures with the help of the built-in rotating motor. Then the system will process the image and classify the firearm brand in under 62 seconds.
Asst Prof Dr. Charturong explained that this application is using AI to process and analyze the information. The research tested samples of 898 ammunition from guns of eight brands, that are commonly used in committing crimes. Accuracy of results is as high as 91-98 percent depending on the gun brand.
The application is currently on test runs in a real-life situation. The officer uses the result from the application to frame the investigation and cut down other possibilities which can mean tracing back to the criminal faster. In this state, the conventional method is still being used, accompany the application to maintain absolute accuracy.
This study is the first to design an application that can work on a smartphone, which is why the work has been published in the best quartile (Q1) of the IEEE Access journal. TSE is proud to conduct the study that is beneficial to society like the crime investigation and the judicial process.
In the future, the application will be able to work on the classification of more class characteristic subjects, such as the firearm model, serial number from bullet markings, and there might be a way to look for traces of the previous crime.
The in-depth information about this study can be accessed here.