A ballistic solution system and a method thereof

Abstract

The present invention relates to determining the brand / pattern of a licensed or unlicensed rifled gun involved in a crime by means of deep learning.

Description

[0001] A BALLISTIC SOLUTION SYSTEM AND A METHOD THEREOF

[0002] Technical Field of the Invention The present invention relates to determining the brand / pattern of a licensed or unlicensed rifled gun involved in a crime by means of deep learning.

[0003] State of the Art

[0004] The study of the changes that occur in the barrel with the firing of the cartridge, the movement of the cartridge bullet and its effects on the target is known as ballistics. Today, ballistic studies are carried out on bullets and casings fired from illegal or licensed guns. These studies are carried out by the shots taken during the crime scene investigation, and the visual examination of the pictures or residuals on the casing and core found. If the casing or core found belongs to a licensed weapon, the ballistic traces thereon are compared and matched with the existing database. In ballistic studies with illegal / unlicensed guns, this process is more difficult and the detection of the firearm relates to the expertise of the person.

[0005] These studies necessitates human supervision and experience. Therefore, carrying out these studies requires time and experienced manpower. A certain standard could not be established since the current practices are completely based on human supervision and require experience and training.

[0006] In the state of the art, the application numbered “US20030149543A1” was examined. The invention, which is the subject of the application, relates to the detection of the gun shot through the ballistic traces. In the invention, the gun is detected by comparing the traces of the suspect bullets and control bullets. In the state of the art, the application numbered “US9778007B1” was examined. The invention, which is the subject of the application, aims to match a firearm cartridge / bullet casing. In the invention, the tip of a firing pin of a firearm fired from the capsule of the casing is designed in nano-scale and the firing pin is patterned. In the state of the art, the application numbered “AU2017101803” was examined. The invention, which is the subject of the application, relates to the classification of different gun types by using deep learning in images. The invention provides classification of different gun images collected from databases with a deep learning-based system. Deep learning methods were used in applications in the state of the art, however, it is used for different purposes such as gun detection in the image and classification of different gun types. In rural areas or small cities, there may be a shortage of specialist personnel for ballistic examination, and images or evidence are sent to affiliated centers for examination. Therefore, there is a need for a system that can be performed without preliminary examination and expert support in rural areas and that accelerates the processes.

[0007] Consequently, the disadvantages disclosed above and the inadequacy of available solutions in this regard necessitated making an improvement in the relevant technical field.

[0008] Objects of the Invention

[0009] The present invention relates to determining the brand / pattern of a licensed or unlicensed rifled gun involved in a crime by means of deep learning.

[0010] The most important object of the present invention is to offer suggestions on potential gun brands and patterns with the photos of the casing part and / or core photos found at the crime scene. Thus, it accelerates the ballistic processes.

[0011] Another important object of the present invention is to carry out preliminary examination and examination in rural areas without the support of an expert. Thus, it provides ballistic detection in areas without trained personnel. Another important object of the present invention is to make the system platform independent and portable. Thus, it can be applied in any desired region without the need for special equipment. Another important object of the present invention is that it can be improved with artificial intelligence and deep learning methods. Thus, it reduces the need for experienced and educated people.

[0012] Another important aim of the invention is to provide predictions from different angles by processing different evidence pictures in separate ballistic deep learning modules, since the tip of a firing pin is characteristic for certain guns, the number of riffles / sets is decisive for certain gun groups.

[0013] Structural and characteristic features of the present invention as well as all advantages thereof will become clear through the attached figures and the following detailed description provided by making references thereto. Therefore, the assessment should be made by taking these figures and the detailed description into consideration.

[0014] Description of the Figures

[0015] Figure 1 illustrates a view of schematic flow diagram of the system according to the present invention.

[0016] Figure 2 illustrates a view of schematic flow diagram of the system of the processor according to the present invention.

[0017] Reference Numerals:

[0018] 100. A ballistic solution system 110. Electronic Device

[0019] 111. Application

[0020] 112. Processor

[0021] 112.1 Gun brand detection module

[0022] 112.2 Gun pattern detection module Description of the Invention

[0023] The present invention relates to determining the brand / pattern of a licensed or unlicensed rifled gun involved in a crime by means of deep learning.

[0024] Electronic device (110) is a device such as a computer, smartphone, tablet computer configured to exchange data over a data network. The electronic device (110) comprises at least one display and processor (112) therein.

[0025] The application (111 ) is executed on the electronic device (110), and it ensures that suggestions regarding potential gun brands and patterns are displayed to users through an interface. The application (111 ) enables users to upload photos of casing parts, capsule photos, and core photos taken with a macro-capable camera or lens at the crime scene they desire to analyze, through an interface. The application (111 ) enables users to select which images they upload are casing, capsule, and core images through an interface. The application (111) comprises presenting the brand pattern estimation made separately according to the casing trace, pin trace, and core trace to the ballistics expert through an interface.

[0026] The processor (112) is executed within the electronic device (110) and provides the processing of the evidence photograph data. The processor (112) enables to detect suspicious weapon brand and pattern possibilities by using deep learning and fine- tuning methods on the macro photos uploaded to the application (111). The processor (112) provides image processing with deep learning modeling on each category of macro images, which are divided into three subcategories as casing, capsule and core, obtained through the application (111). The processor (112) performs graphical operations such as sharpening and highlighting traces on each of the categories of casing, capsule, and core. It can observe the independent evaluation of each module or the statistically most reliable estimate separately. Processor (112) provides independent evaluation of each category and computation of the most statistically reliable estimate. The processor (112) ensures that the brand and pattern that it finds the most similar in the brand-pattern estimation is presented first. The processor (112) provides the most trusted brand / pattern prediction first. Confidence index is the possibility of estimation in the range of 0-1 that the processor (112) predicts for the relevant brand / pattern while performing estimation. In a preferred embodiment of the present invention, the processor (112) provides the examination of images in three categories (casing, capsule pin, and core) and the determination of the brand pattern estimation according to these three categories. However, when requested or when there are no macro images for each category, the processor (112) enables the desired module to be run for the desired category.

[0027] The gun brand detection module (112.1 ) runs on the processor (112), and inserts the macro images obtained after graphical processes into a pre-trained and fine-tuned deep CNN (Convolutional neural network) model, thereby providing separate brand detection for the categories of casing, capsule pin, and core. In a preferred embodiment of the present invention, the gun brand detection module (112.1 ) provides brand detection by inserting it into a pre-trained and fine-tuned deep CNN model such as VGG16 / VGG19. The gun brand detection module (112.1 ) ensures that the brand with the highest similarity value is selected.

[0028] The gun pattern detection module (112.2) runs on the processor (112), and enables the detection of the patterns of the brands detected for each of the casing, capsule pin, and core categories in the gun brand detection module (112.1) by using trained deep learning. The gun pattern detection module (112.2) enables model predictions to be made separately according to the casing trace, pin trace, and core trace. Thus, separate deep learning models are calculated for each of the categories of casing, capsule pin, and core. The purpose of inserting the traces of each category into different deep learning models is to ensure that different ballistic traces can be evaluated in independent models.

[0029] Since the system is completely deep learning-based, it determines the brand pattern of the rifled gun shot from different ballistic traces without surveillance, provides support to human surveillance and investigation processes, and accelerates ballistic studies after criminal acts. The system can operate on the part that is thrown out of the casing ejection cavity and remains intact and the core that comes out of the barrel. Each image type is processed in a different deep learning module, therefore, even if the core is found in a degraded state, it can make predictions on the pin and magazine cheek traces on the casing. Modules can be activated configuratively according to the type of evidence photograph available. Thus, it can operate even if there is an evidence image of a single category that can be obtained. The operating method of the ballistic solution system (100) comprises the process steps of;

[0030] • Categorizing (101 ) the evidence macro images taken at the crime scene by users through the application (111) as casing, pin or core trace,

[0031] • Implementing the graphics processing of sharpening and highlighting marks on each of the categories of casing, pin, and core traces by the processor (112),

[0032] • Applying deep learning separately by the processor (112) on each category of macro images categorized in the application (111 ),

[0033] • Selecting the brand with the highest similarity by identifying the brands belonging to each category with a pre-trained and fine-tuned CNN model on each category of macro images categorized in the application (111) by means of the gun brand detection module (112.1 ),

[0034] • detecting the patterns of the brands detected for each of the categories of casing, pin, and core in the gun brand detection module (112.1 ) by using trained deep learning by means of the gun pattern detection module (112.1 ),

[0035] • transmitting the brand pattern and similarity values detected by the processor (112) to the application (111),

[0036] • Presenting the brand pattern and similarity values determined by the application (111 ) to the user through an interface.

Claims

CLAIMS1. A ballistic solution system (100) that enables the detection of the brand / pattern of a licensed or unlicensed rifled weapon involved in a crime by deep learning, characterized by comprising- at least one electronic device (110) that exchanges data over a data network and provides at least one processor (112) therein, and that comprises;- at least one gun brand detection module (112.1 ) that inserts the macro images obtained after graphical process into a pre-trained and fine-tuned CNN model by running on the processor (123), thereby providing separate brand detection for the categories of casing, capsule pin, and core,- at least one gun pattern detection module (112.2) that runs on the processor (112) to detect the patterns of the brands detected for each of the trace categories of casing, pin, and core in the gun brand detection module (12.1 ) by using trained deep learning for each category,- at least one processor (112) that is executed in the electronic device (110) and provides processing of the evidence macro image data, and that enables image processing with deep learning modeling on each category of macro images, which are divided into three sub-categories as casing, capsule and core, obtained through the application (111),- at least one application (111 ) that is executed on the electronic device (110), that provides users with suggestions about potential gun brands and patterns detected.

2. An operating method of the ballistic solution system (100), characterized in that, it comprises the process steps of;Categorizing the evidence macro images taken at the crime scene by users through the application (111) as casing, pin or core trace,• Implementing the graphics processing of sharpening and highlighting marks on each of the categories of casing, pin, and core traces by the processor (112),• Applying deep learning separately by the processor (112) on each category of macro images categorized in the application (111 ),• Selecting the brand with the highest similarity by identifying the brands belonging to each category with a pre-trained and fine-tuned CNN model on each category of macro images categorized in the application (111) by means of the gun brand detection module (112.1 ),• detecting the patterns of the brands detected for each of the categories of casing, pin, and core in the gun brand detection module (112.1 ) by using trained deep learning by means of the gun pattern detection module (112.1 ),• transmitting the brand pattern and similarity values detected by the processor (112) to the application (111),• presenting the brand pattern and similarity values determined by the application (111 ) to the user through an interface.

3. A ballistic solution system (100) according to Claim 1 , characterized in that, the electronic device (110) is a computer, smartphone, or tablet computer.

4. A ballistic solution system (100) according to Claim 1 , characterized in that , it comprises an application (111 ) that allows users to upload macro photos of casing, pin and core traces found at the crime scene, which they desire to analyze, through an interface.

5. A ballistic solution system (100) according to Claim 1 , characterized in that, it comprises an application (111 ) that allows users to categorize the evidence macro images taken at the crime scene as a casing, pin or core trace through an interface.

6. A ballistic solution system (100) according to Claim 1, characterized in that, it comprises an application (111) that allows to present the brand pattern estimation made separately according to the casing, pin, and core traces to the ballistics expert through an interface.

7. A ballistic solution system (100) according to Claim 1, characterized in that, it comprises a processor (112) that enables implementing the graphic processing of sharpening and highlighting traces on each of the categories of casing, pin, and core.

8. A ballistic solution system (100) according to Claim 1, characterized in that, it comprises a processor (112) that enables each category (casing, pin, and core trace) to be evaluated independently and the brand-pattern similarity estimates are calculated and the most similar brand pattern is presented first (112).

Images