Bullet trace recognition system and device based on multi-dimensional feature extraction

Abstract

The invention relates to the technical field of bullet trace recognition, and discloses a bullet trace recognition system and device based on multi-dimensional feature extraction, and the system comprises a data collection module which is used for synchronously collecting multi-dimensional data of bullet traces, the multi-dimensional data comprises three-dimensional shape data, oxide layer color data, combustion residue color data, scratch trend data and surface change data caused by heat effect; the data preprocessing module is connected with the data acquisition module and is used for preprocessing and calibrating the acquired multi-dimensional data; the feature extraction module is connected with the data preprocessing module, and is used for accurately identifying fuzzy traces of old bullets by simultaneously acquiring information in multiple aspects such as appearances, colors, scratch trends and temperature changes of the bullets and combining feature enhancement and logical reasoning technologies, so that judgment errors caused by only seeing a single feature are avoided, and the accuracy of recognition is improved. And accurate identification under a complex condition is ensured.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of bullet trace identification, in particular to a bullet trace identification system based on multi-dimensional feature extraction and a device thereof. BACKGROUND

[0002] Bullet trace identification is a key technology, which plays a crucial role in combating gun-related crimes and ensuring social safety. Traditional bullet trace identification methods mainly rely on the comparison and analysis of two-dimensional microscopic images. The staff determines the source of the bullet by manually observing and judging the shape and size of the trace. However, this method has many drawbacks, such as strong subjectivity, low identification efficiency, and accuracy easily affected by human factors.

[0003] With the continuous progress of computer technology and image processing technology, bullet trace identification methods based on three-dimensional topographic data have gradually developed. This method collects three-dimensional topographic data of bullet head traces, uses computers for feature extraction and matching identification, thereby improving the accuracy and efficiency of identification to some extent. However, even so, the existing identification method based on three-dimensional topographic data still has limitations. It only focuses on three-dimensional topographic features, ignoring information in other dimensions (oxidation layer color, color of combustion residues, direction of scratches, surface changes caused by thermal effects), which makes the accuracy and reliability of identification need to be further improved.

[0004] The existing bullet trace identification device also has defects in data collection, which cannot realize the synchronous collection and comprehensive analysis of multi-dimensional data, and is difficult to meet the demand of accurate identification of bullet traces in actual application. In view of this, we propose a bullet trace identification system based on multi-dimensional feature extraction and a device thereof. SUMMARY

[0005] To solve the above-mentioned problems of the existing bullet trace identification method based on three-dimensional topographic data, which only focuses on three-dimensional features and ignores information in other dimensions, and the identification device cannot synchronously collect and comprehensively analyze multi-dimensional data, resulting in insufficient accuracy and reliability, and difficulty in meeting the demand of accurate identification, the present application provides a bullet trace identification system based on multi-dimensional feature extraction and a device thereof.

[0006] The bullet trace identification system based on multi-dimensional feature extraction and the device thereof provided by the present application adopt the following technical solutions:

[0007] A bullet trace recognition system and device based on multi-dimensional feature extraction, comprising: a data acquisition module for synchronously acquiring multi-dimensional data of bullet traces, the multi-dimensional data including three-dimensional topography data, oxidation layer color data, color data of combustion residues, scratch direction data, and surface change data caused by thermal effects;

[0008] A data preprocessing module connected with the data acquisition module for preprocessing and calibrating the acquired multi-dimensional data;

[0009] A feature extraction module connected with the data preprocessing module for extracting corresponding features from the preprocessed multi-dimensional data, respectively obtaining three-dimensional topography features, oxidation layer color features, combustion residue color features, scratch direction features, and thermal effect surface change features;

[0010] A feature matching module connected with the feature extraction module for matching the extracted multi-dimensional features with bullet trace sample features stored in a database and calculating a matching degree;

[0011] A result output module connected with the feature matching module for outputting a recognition result according to the matching degree.

[0012] Preferably, the three-dimensional topography features extracted by the feature extraction module include trace depth, slope, and curvature; the oxidation layer color features and combustion residue color features include color RGB value, hue, and saturation; the scratch direction features include scratch direction angle and length; and the thermal effect surface change features include temperature change gradient and hot spot distribution.

[0013] Preferably, the feature matching module calculates a comprehensive matching degree by using a weighted summation method, and the weight of each dimension feature is pre-set according to its importance in recognition.

[0014] Preferably, the result output module outputs corresponding bullet sample information when the matching degree is greater than a preset threshold, and outputs a result of not matching to a corresponding sample when the matching degree is less than or equal to the preset threshold.

[0015] Preferably, the device comprises a base, characterized in that a bearing platform is rotatably connected to the top of the base, a three-dimensional topography acquisition assembly is arranged on the top of the bearing platform, a color feature acquisition assembly is arranged on one side of the three-dimensional topography acquisition assembly, a scratch direction acquisition assembly is arranged on one side of the color feature acquisition assembly, a thermal effect surface change acquisition assembly is arranged on one side of the scratch direction acquisition assembly, the three-dimensional topography acquisition assembly, the color feature acquisition assembly, the scratch direction acquisition assembly, and the thermal effect surface change acquisition assembly are electrically connected with a control module, and a data transmission module is fixedly connected to the back of the control module.

[0016] The bearing platform is arranged on the base and used for placing the cartridge to be identified, and the bearing platform can rotate around its own axis.

[0017] The three-dimensional topography acquisition component, the color feature acquisition component, the scratch direction acquisition component and the thermal effect surface change acquisition component are arranged on the base and distributed around the bearing platform.

[0018] The three-dimensional topography acquisition component comprises a three-dimensional scanner, and a lens of the three-dimensional scanner is directed to the cartridge on the bearing platform.

[0019] The color feature acquisition component comprises a high-resolution color camera and a light supplement device, and a lens of the high-resolution color camera is directed to the cartridge on the bearing platform.

[0020] The scratch direction acquisition component is internally provided with a laser scanner, and the laser scanner is used for emitting a laser beam to irradiate on the trace surface of the cartridge.

[0021] The thermal effect surface change acquisition component comprises an infrared thermal imager.

[0022] The control module is connected with the bearing platform, the three-dimensional topography acquisition component, the color feature acquisition component, the scratch direction acquisition component and the thermal effect surface change acquisition component respectively, and is used for controlling the working states of the components.

[0023] Preferably, the control module can control the rotation speed and angle of the bearing platform, and the acquisition time and frequency of the acquisition components.

[0024] Preferably, the light supplement device is used for providing stable light conditions for color acquisition.

[0025] The method comprises the following steps:

[0026] S1: placing the cartridge to be identified on the bearing platform, starting the acquisition components and the bearing platform through the control module, rotating the cartridge by the bearing platform, synchronously acquiring the multi-dimensional data of the cartridge trace by the acquisition components, and transmitting the data to the data acquisition module through the data transmission module;

[0027] S2: pre-processing the multi-dimensional data received by the data acquisition module by the data preprocessing module, calibrating the data acquired by different acquisition components, and keeping the data consistent in space and time;

[0028] S3: extracting features from the pre-processed multi-dimensional data by the feature extraction module, extracting the depth, slope and curvature of the trace for the three-dimensional topography data, extracting the RGB value, hue and saturation of the color for the color data, extracting the direction angle and length of the scratch for the scratch direction data, and extracting the temperature change gradient and hot spot distribution for the thermal effect surface change data;

[0029] S4: The feature matching module matches the extracted multi-dimensional features with the sample features in the database, and calculates the comprehensive matching degree by using the weighted summation method, and the weight of each dimension feature is pre-set;

[0030] S5: The result output module outputs the recognition result according to the comprehensive matching degree, and outputs the corresponding bullet sample information when the matching degree is greater than the pre-set threshold value, and outputs the result of not matching the corresponding sample when the matching degree is less than or equal to the pre-set threshold value.

[0031] Preferably, the weights of the three-dimensional topographic features, color features, scratch direction features and thermal effect surface change features in step S4 are 0.3, 0.2, 0.25 and 0.25 respectively.

[0032] Preferably, the pre-set threshold value in step S5 is 0.8.

[0033] In summary, the present application has the following beneficial technical effects:

[0034] 1. By simultaneously collecting the shape, color, scratch direction and temperature change of the bullet and other information, and combining feature enhancement and logical reasoning technology, even the blurred traces of old bullets can be accurately identified, avoiding the judgment errors caused by looking at single features, and ensuring accurate identification in complex situations.

[0035] 2. The special light source and enhancement algorithm are adapted to old bullet identification, which can meet the needs of laboratory fine analysis and cope with on-site rapid inspection, and has a wide range of applications.

[0036] 3. The device has small size, high automation degree, standardized acquisition and processing process, and does not need to rely on too much experience during operation. In addition, the modular design is convenient for maintenance, the fuzzy matching function can also assist manual judgment, and is easy to popularize and apply. BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 is the overall process of the present application.

[0038] Figure 2 is the working process of the identification device of the present application.

[0039] Figure 3 is the detailed process of the feature extraction module of the present application.

[0040] Figure 4 is the feature matching and result output process of the present application.

[0041] Figure 5 is the overall structure diagram of the identification device of the present application.

[0042] Explanation of reference signs:

[0043] 1, base; 2, bearing platform, 3; three-dimensional topography acquisition assembly; 4, color feature acquisition assembly; 5, scratch direction acquisition assembly; 6, heat effect surface change acquisition assembly; 7, control module; 8, data transmission module. DETAILED DESCRIPTION

[0044] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0045] Please refer to Figures 1-5 The present application provides three technical solutions:

[0046] The system comprises a data acquisition module, a data preprocessing module, a feature extraction module, a feature matching module and a result output module, and each module realizes data interaction through a data bus or a wireless communication mode.

[0047] The data acquisition module adopts a multi-sensor synchronous acquisition architecture, comprising a three-dimensional topography sensor (scanning accuracy 0.01 mm), a hyperspectral camera (spectral range 400-700 nm, resolution 12 million pixels), a laser profile sensor (scanning frequency 10 kHz) and an infrared thermal imager (temperature measurement range -20~300℃, thermal sensitivity 0.05℃). Each sensor realizes acquisition timing alignment through a time synchronizer (accuracy ±1 ms), ensuring the consistency of multi-dimensional data in the time dimension.

[0048] The data preprocessing module comprises a noise removal unit and a data calibration unit. The noise removal unit removes high-frequency noise by Gaussian filtering (σ=1.5) for three-dimensional topography data, suppresses salt and pepper noise by median filtering (window 3x3) for color data, and removes noise by wavelet thresholding (db4 wavelet, soft threshold) for infrared data; the data calibration unit realizes spatial calibration through a calibration board (checkerboard + gray gradient board), unifies the data of each sensor to a three-dimensional coordinate system (error ≤0.02 mm) with the bullet axis as the origin, and realizes time calibration through timestamp alignment.

[0049] The feature extraction module: based on the preprocessed data, multi-dimensional feature analysis is performed:

[0050] Three-dimensional topography features: trace depth (maximum depth, average depth), slope (inclination angle along the trace direction, accuracy ±0.5°), curvature (principal curvature, Gaussian curvature, unit 1 / mm) are extracted by point cloud fitting algorithm;

[0051] Color feature: Image segmentation (Otsu threshold method) was performed on the oxide layer and combustion residue area to extract the RGB value (0-255), hue (0-360°), and saturation (0-100%);

[0052] Scratch direction feature: The oxide layer on the surface of the old bullet caused some scratches to be blurred. After enhancement processing by Hough transform, 9 effective scratches were detected, with direction angles scattered in 42°±8° (slightly deviated from 40° of the new bullet of the same model), average length 1.5 mm (actual physical length was estimated to be 1.8 mm due to the coverage of the oxide layer), and density 1.2 / mm² in the 12 mm² detection area.

[0053] Thermal effect surface change feature: Due to the influence of storage time, the thermal residual signal was weak, the temperature change gradient decreased to 0.3 ℃ / mm, and only one obvious hot spot was detected, with coordinates (x=4 mm, y=1 mm) and temperature 2 ℃ higher than the matrix. Compared with the aging model of the same type, it was confirmed that the hot spot position had a coincidence degree of 70% with the hot spot distribution area of the new bullet.

[0054] Feature matching module: Weighted Euclidean distance algorithm was used to calculate the matching degree. The preset weight of each dimension feature was: 0.3 for three-dimensional morphology feature, 0.2 for color feature, 0.25 for scratch direction feature, and 0.25 for thermal effect surface change feature. The specific calculation method was:

[0055] Comprehensive matching degree = 0.3 × three-dimensional matching degree + 0.2 × color matching degree + 0.25 × scratch matching degree + 0.25 × thermal effect matching degree

[0056] Among them, each single dimension matching degree was processed by Euclidean distance normalization of feature vector (range 0-1, value greater than 1, matching degree higher).

[0057] Result output module: The preset matching threshold was 0.8. When the comprehensive matching degree was greater than 0.8, the information of the corresponding bullet sample in the database was output (including the model of the gun, the production batch, the rifling parameters, etc.). When the comprehensive matching degree was less than or equal to 0.8, it was output that no corresponding sample was matched, and manual review was prompted.

[0058] The device includes a base 1, a bearing platform 2, a three-dimensional morphology acquisition assembly 3, a color feature acquisition assembly 4, a scratch direction acquisition assembly 5, a thermal effect surface change acquisition assembly 6, a control module 7, and a data transmission module 8.

[0059] Bearing platform 2: Driven by a servo motor (rotational speed 0-60 rpm adjustable), it can clamp a bullet with a diameter of 5-20 mm, and the rotation angle control accuracy is ±0.1°.

[0060] Three-dimensional morphology acquisition component 3: three-dimensional scanner (consistent with the parameters in the data acquisition module), scanning range covers the full length of the bullet (20-50 mm).

[0061] Color feature acquisition component 4: including high-resolution color camera (color temperature 5500K, brightness adjustable), camera lens is at an angle of 45° with the table surface of the bearing platform 2, and the light supplement device ensures that the light uniformity is >90%.

[0062] Scratch direction acquisition component 5: laser scanner emits a 650nm red laser beam to irradiate the surface of the bullet, and receives reflected light to obtain the scratch profile.

[0063] Thermal effect surface change acquisition component 6: infrared thermal imager (consistent with the parameters in the data acquisition module) lens is directly opposite to the table surface of the bearing platform 2, the field of view is 50°×35°, and the thermal distribution of the bullet surface can be completely photographed.

[0064] Control module 7: PLC controller is adopted, parameters can be set through a touch screen (10.1 inches), the rotation speed (such as 30 rpm) of the bearing platform 2, the acquisition frequency (such as 10 frames / second) of each acquisition component and the acquisition time length (such as 5 seconds) can be controlled.

[0065] Data transmission module 8: gigabit Ethernet interface is adopted to transmit the collected data to the system host in real time (transmission rate ≥100 Mbps).

[0066] Identification method

[0067] Step S1: place the 9mm pistol bullet to be identified on the bearing platform 2, set the rotation speed of the bearing platform 2 to 30 rpm through the control module 7, set the acquisition frequency of each acquisition component to 10 frames / second, and start the acquisition process. The bearing platform 2 drives the bullet to rotate at a constant speed, the three-dimensional morphology acquisition component 3 obtains the three-dimensional point cloud data of the bullet surface, the color feature acquisition component 4 synchronously photographs the color image of the oxidation layer and the combustion residue, the scratch direction acquisition component 5 scans the scratch profile, the thermal effect surface change acquisition component 6 records the surface temperature distribution, and the data is sent to the data acquisition module through the data transmission module 8.

[0068] Step S2: the data preprocessing module processes the received multidimensional data: Gaussian filtering is adopted to remove three-dimensional point cloud noise, median filtering is adopted to purify color image, and wavelet threshold denoising is adopted to optimize infrared data; through the calibration plate, the data of each sensor is calibrated to a unified coordinate system, and the timestamp alignment error is ≤1ms.

[0069] Step S3: The feature extraction module extracts features from the pre-processed data: three-dimensional topographic features include maximum depth 0.12 mm, average slope 35°, principal curvature 0.05 / mm; color features include oxide layer RGB value (150, 80, 50), hue 25°, saturation 60%; scratch direction features include direction angle 45°, average length 2 mm, density 0.8 / mm²; thermal effect features include temperature change gradient 0.5°C / mm, hot spot coordinates (x=5 mm, y=3 mm).

[0070] Step S4: The feature matching module matches the extracted features with the bullet sample features in the database, and calculates the three-dimensional matching degree 0.85, the color matching degree 0.78, the scratch matching degree 0.82, and the thermal effect matching degree 0.80. The comprehensive matching degree = 0.3x0.85 + 0.2x0.78 + 0.25x0.82 + 0.25x0.80 = 0.82.

[0071] Step S5: The result output module judges that the comprehensive matching degree 0.82>0.8, and outputs the matching sample information.

[0072] Example Two

[0073] The difference between this example and Example One is that the data preprocessing accuracy and feature matching efficiency are optimized, which is suitable for scenes with high speed requirements (such as on-site rapid investigation).

[0074] System optimization:

[0075] The data preprocessing module adds an adaptive noise removal algorithm, which dynamically adjusts the filtering parameters according to the data signal-to-noise ratio (such as three-dimensional data σ=1.0-2.0 adaptive), and the processing efficiency is improved by 30%;

[0076] The feature matching module uses KD tree to accelerate feature retrieval, which shortens the sample matching time;

[0077] Weight adjustment: the scratch direction feature weight is increased to 0.3, and the thermal effect feature weight is reduced to 0.2 (because the thermal effect is easily disturbed in the field environment), and the three-dimensional topography (0.3) and color (0.2) remain unchanged.

[0078] Identification method:

[0079] In the identification process of the bullet, the comprehensive matching degree calculation of step S4 is 0.3x0.88+0.2x0.75+0.3x0.86+0.2x0.70=0.81, which is greater than the threshold value 0.8, and the sample is successfully matched.

[0080] Example Three

[0081] This embodiment is aimed at the identification scene of old bullets (serious surface oxidation, blurred traces), focusing on optimizing the robustness of feature extraction.

[0082] System optimization:

[0083] The feature extraction module adds a trace enhancement algorithm: morphological dilation operation is used on three-dimensional topographic data to highlight shallow marks (depth <0.05mm), and histogram equalization is used on color data to enhance the contrast between the oxidation layer and the substrate;

[0084] The feature matching module introduces fuzzy logic reasoning, allowing feature values to float within a range of ±10% (such as a scratch length error ≤0.2mm), reducing sensitivity to fuzzy traces.

[0085] When identifying rifle bullets, the enhanced shallow mark depth in the three-dimensional topographic features extracted in step S3 can be identified to 0.03mm, and the fluorescence intensity (gray value 200) of the combustion residues in the color features is effectively extracted; the comprehensive matching degree of step S4 is 0.79, which is close to the threshold value 0.8, but is determined as "suspected matching" through fuzzy logic, outputting sample information and prompting manual verification (the final manual confirmation is correct matching).

[0086] The above describes an exemplary embodiment of a bullet trace identification system and device based on multi-dimensional feature extraction provided by the present disclosure with reference to the preferred embodiment, however, those skilled in the art can understand that various modifications and improvements can be made to the above specific embodiments without departing from the concept of the present disclosure, and various technical features and structures proposed by the present disclosure can be combined without exceeding the protection scope of the present disclosure, the protection scope of the present disclosure is determined by the appended claims.

Claims

1. A bullet trace identification system based on multidimensional feature extraction, characterized in that, include: The data acquisition module is used to synchronously acquire multidimensional data of bullet marks, including three-dimensional morphology data, oxide layer color data, combustion residue color data, scratch direction data, and surface change data caused by thermal effects. A data preprocessing module, connected to the data acquisition module, is used to preprocess and calibrate the acquired multidimensional data; The feature extraction module is connected to the data preprocessing module and is used to extract corresponding features from the preprocessed multidimensional data to obtain three-dimensional morphology features, oxide layer color features, combustion residue color features, scratch direction features, and thermal effect surface change features, respectively. The feature matching module, connected to the feature extraction module, is used to match the extracted multidimensional features with the bullet trace sample features stored in the database and calculate the matching degree. The result output module is connected to the feature matching module and is used to output the recognition result based on the matching degree.

2. The system according to claim 1, characterized in that, The three-dimensional morphological features extracted by the feature extraction module include the depth, slope, and curvature of the trace; the color features of the oxide layer and the color features of the combustion residue include the RGB value, hue, and saturation of the color; the scratch direction features include the direction angle and length of the scratch; and the thermal effect surface change features include the temperature change gradient and hot spot distribution.

3. The system according to claim 1, characterized in that, The feature matching module calculates the overall matching degree using a weighted summation method, and the weights of each feature dimension are pre-set according to their importance in the recognition.

4. The system according to claim 1, characterized in that, The result output module outputs the corresponding bullet sample information when the matching degree is greater than a preset threshold; and outputs the result that no matching sample was found when the matching degree is less than or equal to the preset threshold.

5. A bullet trace identification device based on multidimensional feature extraction, comprising a base (1), characterized in that, The top of the base (1) is rotatably connected to a support platform (2). The top of the support platform (2) is provided with a three-dimensional shape acquisition component (3). A color feature acquisition component (4) is provided on one side of the three-dimensional shape acquisition component (3). A scratch direction acquisition component (5) is provided on one side of the color feature acquisition component (4). A thermal effect surface change acquisition component (6) is provided on one side of the scratch direction acquisition component (5). The three-dimensional shape acquisition component (3), color feature acquisition component (4), scratch direction acquisition component (5), and thermal effect surface change acquisition component (6) are all electrically connected to the control module (7). A data transmission module (8) is fixedly connected to the back of the control module (7). The support platform (2) is mounted on the base and is used to place the bullets to be identified. The support platform (2) can rotate around its own axis. The three-dimensional shape acquisition component (3), color feature acquisition component (4), scratch direction acquisition component (5) and thermal effect surface change acquisition component (6) are all set on the base and distributed around the support platform (2); The three-dimensional shape acquisition component (3) includes a three-dimensional scanner, the lens of which is facing the bullet on the support platform (2); The color feature acquisition component (4) includes a high-resolution color camera and a supplementary lighting device, with the lens of the high-resolution color camera facing the bullet on the carrier platform (2); The scratch direction acquisition component (5) has a built-in laser scanner, which is used to emit a laser beam to irradiate the surface of the bullet mark. The thermal effect surface change acquisition component (6) includes an infrared thermal imager; The control module (7) is connected to the carrier platform (2), the three-dimensional shape acquisition component (3), the color feature acquisition component (4), the scratch direction acquisition component (5), and the thermal effect surface change acquisition component (6) respectively, and is used to control the working status of each component.

6. The apparatus according to claim 5, characterized in that, The control module (7) can control the rotation speed and angle of the support platform (2), as well as the acquisition time and frequency of each acquisition component.

7. The apparatus according to claim 5, characterized in that, The supplementary lighting device is used to provide stable lighting conditions for color acquisition.

8. A method for identifying bullet traces based on multidimensional feature extraction, characterized in that, Includes the following steps: S1: Place the bullet to be identified on the carrier platform (2), start each acquisition component and the carrier platform (2) through the control module (7), the carrier platform (2) drives the bullet to rotate, each acquisition component synchronously acquires multi-dimensional data of bullet traces, and transmits the data to the data acquisition module through the data transmission module (8); S2: The data preprocessing module preprocesses the multidimensional data received by the data acquisition module and calibrates the data acquired by different acquisition components to ensure that the data maintains consistency in space and time. S3: The feature extraction module extracts features from the preprocessed multidimensional data. For three-dimensional topography data, it extracts the depth, slope, and curvature of the traces; for color data, it extracts the RGB values, hue, and saturation of the colors; for scratch direction data, it extracts the direction angle and length of the scratches; and for thermal effect surface change data, it extracts the temperature change gradient and hotspot distribution. S4: The feature matching module matches the extracted multidimensional features with the sample features in the database, and calculates the comprehensive matching degree by weighted summation. The weights of each feature dimension are preset. S5: The result output module outputs the recognition result based on the comprehensive matching degree. When the matching degree is greater than the preset threshold, the corresponding bullet sample information is output; when the matching degree is less than or equal to the preset threshold, the result of not matching the corresponding sample is output.

9. The method according to claim 8, characterized in that, In step S4, the weights of the three-dimensional morphology features, color features, scratch direction features, and thermal effect surface change features are 0.3, 0.2, 0.25, and 0.25, respectively.

10. The method according to claim 8, characterized in that, The preset threshold in step S5 is 0.8.

Images