CN104123546A - Multi-dimensional feature extraction based bullet trace comparison method - Google Patents
Multi-dimensional feature extraction based bullet trace comparison method Download PDFInfo
- Publication number
- CN104123546A CN104123546A CN201410358260.6A CN201410358260A CN104123546A CN 104123546 A CN104123546 A CN 104123546A CN 201410358260 A CN201410358260 A CN 201410358260A CN 104123546 A CN104123546 A CN 104123546A
- Authority
- CN
- China
- Prior art keywords
- data
- dimensional
- bullet
- feature extraction
- mark
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a multi-dimensional feature extraction based bullet trace comparison method. Bullet trace amplified plane two-dimensional images are obtained by a digital photography method through the existing bullet automatic comparison retrieval system, the bullet trace height information is omitted, the real morphology of the bullet emission trace cannot be reflected, and accordingly the information which can provide the identification reference is less. According to the multi-dimensional feature extraction based bullet trace comparison method, multi-dimensional features of bullet data are extracted, the multi-dimensional feature advantages are combined to achieve the accurate comparison analysis of the bullet data, and a 3D-Zernike based three-dimensional feature extraction method is put forward to solve the description problem of the three-dimensional topology structure characteristics of three-dimensional bullet trace data. The multi-dimensional feature extraction based bullet trace comparison method comprises the following steps of extracting a mean curve; performing characteristics coarse matching; performing three-dimensional feature extraction and analysis in an interested area; performing comparison analysis. The multi-dimensional feature extraction based bullet trace comparison method is used for performing signal and information processing.
Description
Technical field:
The present invention relates to a kind of cartridge mark comparison method based on multi-dimension feature extraction.
Background technology:
Bullet automatic comparison searching system is to adopt the method for digital photography to obtain the planar image that shoot mark is amplified at present, the method has been ignored bullet trace elevation information, the real topography that can not reflect bullet bullet mark, therefore can provide the quantity of information of identification reference relatively less.Be mainly three parts: image data acquiring, characteristic quantity analysis and database and network connection system.The method of digital photography is used in the collection of view data, and bullet clamping, on the rotary table by Electric Machine Control, is carried out to Image Mosaics, processing and analysis [8,9] by computing machine after rotating a circle.The focusing of automatic adjusting, laser assisted, the worktable that this system can be carried out light intensity adjusted etc. automatically, do not need too much manual intervention, can obtain fast the two-dimensional image data of sample.For improving the accuracy of the verification of traces of bullet shell case, analysis, disclose the minutia that planar graph and two dimensional character parameter can not be expressed, the three-dimensional stereo topography of palpus quantitative collection bullet marking on fired bullets.Quantification is obtained the real topography of bullet marking on fired bullets, extracts bullet shell case vestige three-dimensional data feature, analyzes.2007, (the National Institute of Standards and Technology of National Institute of Standards and Technology, NIST) set about setting up national shoot mark database, begin one's study and carry out feasibility assessment, aspect the analysis of bullet three-dimensional surface, doing preliminary exploratory development.
At home, bullet trace check development is also towards quantification future development.Developed cartridge mark automatic recognition system, first adopted bullet roll extrusion tinfoil, on tinfoil, obtained the information of bullet surface, then by camera system, absorbed the marks on surface on tinfoil, finally automatically identified.Developed cartridge mark Computer Automatic Recognition system, this system comprises the untouchable extraction of shooting shell case vestige, and mark information analysis, processing two parts, be domestic first set cartridge mark Computer Automatic Recognition system.Generally speaking, restricted by each side's factor, domestic bullet marks recognition research is at the early-stage, needs the support of new technology, new method.
Along with global gun violence caseload increases gradually, gun violence molecule is fled about to commit crimes, series sexual crime, harm time is long, and damaging range is wide, has caused great harm to citizen's the security of the lives and property, in recent years, the event of China's commit crimes with firearms also increases gradually, and criminal character is serious, very harmful.Therefore for the detection rapidly of gun-related case, prevent as early as possible further gun violence, in addition, offender increases for the kind of the gun of committing a crime, bullet, the shell case quantity that need to make criminal identification are very big, cartridge mark is complicated, bullet, existing its determinacy of shell case vestige form, show again certain randomness and ambiguity, like this, in rifle case detection with provide in criminal material evidence process, the identification of transmitting gun kind and gun homogeneity has run into great difficulty, below three problems extremely important: the one, it is fast that evaluation speed is wanted; The 2nd, it is high that the accuracy of identifying is wanted; The 3rd, need to be by bullet trace datumization, building database, realizes the counter control of cartridge mark and identification fast, prevents that gun-related case from altering and the generation of case, the gun of identification transmitting fast and accurately, improve the detection rate of armed case and the hitting dynamics of increase gun violence.Development along with 3-D scanning technology, the three-dimensional Trace Data of bullet makes rifle bullet trace Computer Automatic Recognition become possibility, in order to improve case handling efficiency and the accuracy of gun-related case, effectively hit and containment gun violence, some advanced countries are all studied cartridge mark counter control technology in the world, realize the computing machine identification of cartridge mark.
According to the vestige on bullet, can judge the kind of ejecting gun, and the final gun of determining this bullet of transmitting.Therefore, check and analyzing cartridge vestige, by the detection of firearm management, gun-related case is played an important role, provide scientific basis for handling a case quickly and accurately, accelerate the speed of handling a case, the process that simplification is handled a case.And can be by used bullet and shell case building database, be connected among a complete information network, realize network management and the utilization of bullet information, rifle bullet trace automatic identifying method is the important component part that realizes cartridge mark automatic recognition system, along with 3-D scanning technology, the development of modern signal processing technology, the three-dimensional Trace Data of bullet makes rifle bullet trace Computer Automatic Recognition become possibility, yet in public security system, there is no now directly to apply the bullet file store of the three-dimensional foundation of bullet yet, more there is no shoot mark alignment algorithm and the software for three-dimensional data information.
Along with the development of 3-D scanning technology, the three-dimensional Trace Data of bullet makes rifle bullet trace Computer Automatic Recognition become possibility, for means and the technology of bullet trace check, mainly can be divided into qualitativeization check and two kinds of modes of quantification check.General stereomicroscope, comparison microscope, the electron microscope of adopting of qualitativeization check.The inexorable trend of cartridge mark detection technique development is the quantification check of vestige, and its top priority is the information of relevant bullet bullet mark of obtaining truly, all sidedly, and domestic current detection technique is main mainly with qualitative analysis, is subject to man's activity larger, and efficiency is low.
Summary of the invention:
The object of this invention is to provide a kind of cartridge mark comparison method based on multi-dimension feature extraction.
Above-mentioned object realizes by following technical scheme:
Extract the multidimensional characteristic of bullet data, associating multidimensional characteristic advantage realizes the accurate compare of analysis of shoot mark data.For the description problem of the three-dimensional topology architectural characteristic of three-dimensional shoot mark data, the three-dimensional feature extracting method based on 3D-Zernike is proposed, the cartridge mark comparison method based on multi-dimension feature extraction comprises the following steps:
Extract Mean curve;
Feature is slightly mated;
In area-of-interest, carry out three-dimensional feature extraction and analysis;
Compare of analysis.
The described cartridge mark comparison method based on multi-dimension feature extraction, described extraction Mean curve, adopt adaptive average value filtering shoot mark data to be carried out to the pre-service of noise remove, projection to pretreated shoot mark time rib three-dimensional data on XOZ axle, shoot mark tendency can well be reflected in drop shadow curve, the average of Qu Ge drop shadow curve on Z axis, the curve being formed by average.
The described cartridge mark comparison method based on multi-dimension feature extraction, described feature is slightly mated, because each bullet of sample data has rib data, i.e. c1, c2, c3 and c4 4 times.When measuring collection, by one rib shoot mark, rotate to another shoot mark, unified according to counterclockwise rotation or clockwise, and the inferior rib scratch data of sample are also according to measuring sequence, provide numbering c1, c2, c3 and c4, but each scratch data is not corresponding according to numbering, so there are following four kinds of matching schemes:
Scheme
:
Scheme
:
Scheme
:
Scheme
:
The described cartridge mark comparison method based on multi-dimension feature extraction described carries out three-dimensional feature extraction and analysis in area-of-interest, according to the preliminary matching result that obtains shoot mark, extracts on this basis three-dimensional feature and does further the matching analysis.The three-dimensional feature of three-dimensional data is different from the topological structure that two dimensional character can be good at response data, and the domestic research to cartridge mark is at present less, more there is no the research of three-dimensional shoot mark feature extraction aspect.Feature for three-dimensional shoot mark data, what three-dimensional invariant moment features herein adopted is 3D-Zernike descriptor, this feature is proposed to identify for objective by N. Canterakis, M. Novotni and R. Klein improve its algorithm subsequently, R. the people such as D. Mill á n finds that the advantageous property of this descriptor uses it for distinguishing of medical image medium vessels knurl, and domestic researchist uses it for the terrain match based on DEM.3D-Zernike descriptor can fully reflect the three-D space structure of target, and the descriptor of higher order more can be described the details of the three dimensions shape of target.This stereoscopic features not only can keep good translation, yardstick, and rotational invariance, and also it is little to have the redundancy of information representation, the high efficiency of information representation, robustness.3D-Zernike square can overcome different angles, the impact of the noises such as different light on data analysis.
The described cartridge mark comparison method based on multi-dimension feature extraction, described compare of analysis, calculate the Euclidean distance between the 3D-Zernike descriptor of different shoot mark data, the similarity degree of describing between two data by the size of Euclidean distance mates identification.
Beneficial effect of the present invention:
Bullet head trace acquisition and automatic identifying method are the important component parts that realizes cartridge mark automatic recognition system.Development along with technology such as 3-D scanning technology, modern signal processing, the three-dimensional Trace Data of bullet makes rifle bullet trace Computer Automatic Recognition become possibility, yet the bullet file store that in public security system, also the three-dimensional of direct application bullet useless is set up now, does not more have shoot mark alignment algorithm and the software for three-dimensional data information.The three-dimensional shoot mark data that this problem just collects first realize pre-service, feature extraction and the feature selecting of bullet data to the analytic process of the holonomic system of similarity comparative analysis, and for the description problem of the three-dimensional topology architectural characteristic of three-dimensional shoot mark data, the three-dimensional feature extracting method based on 3D-Zernike is proposed first.
By optical lens group, forming three-dimensional micro measurement system measures accurately to microcosmic field range, and then completed, bullet data accurately, utilize Digital Signal Processing, through the pre-service of bullet data, feature extraction, set up shoot mark property data base, through feature selecting, then bullet characteristic is carried out to correlativity and similarity comparative analysis.The present invention is pre-service, mathematical model structure, feature extraction, the feature selecting of the data of collection bullet, and comparative analysis, in the recognition system of one, has been replenished the domestic and international blank in this field.
Microtechnic for measurement in space field has based on scanning electron microscope with based on two kinds of methods of optical stereo microscope now.The research of scanning electron microscope and application are had to some mechanisms abroad in research, in calendar year 2001, there is the software and hardware system of comparative maturity to occur simultaneously, domesticly once in some R&D institution, did in this respect research, delivered some articles, but have no commercialization system, occurred.The bullet file store that in public security system, also the three-dimensional of direct application bullet useless is set up now, does not more have shoot mark alignment algorithm and the software for three-dimensional data information.
1. the present invention utilizes extraction Mean curve, and the whole audience formula measurement of acquisition system non-contact measurement, change visual field adds Point Measurement and combines, and measuring speed is fast, measures two dimensional image and the high accuracy three-dimensional data of high definition.This acquisition system can be used for shape and the deformation measurement in biology, medical science, industry, environmental protection, material, police criminal detection, micromechanics, microelectronics, Precision Machining field, has important researching value and wide application prospect.
2. the present invention utilizes feature slightly to mate, the edge extracting technology of carrying out bullet data analysis, noise analysis, noise model foundation, noise remove and doing for further part feature extraction etc., proposed to there is the filtering method that adaptive noise is differentiated, be called adaptive average value filtering.
3. three-dimensional feature extraction and analysis are carried out in utilization of the present invention in area-of-interest, and the forcing cone region of inferior rib data is to the extraction of the features such as Mean curve of bullet center Z-direction projection.And for the description problem of the three-dimensional topology architectural characteristic of three-dimensional shoot mark data, the three-dimensional feature extracting method based on 3D-Zernike is proposed first.Make full use of the translation of 3D-Zernike descriptor, yardstick, rotational invariance, the high efficiency of information representation, towards true three-dimension shoot mark data analysis and comparison, experimental verification the validity of this feature for three-dimensional shoot mark data description, 3D-Zernike descriptor reveals good homogeneity and otherness to the description list of the inferior rib data of shoot mark.
4. the present invention utilizes compare of analysis, sets up raw data base, property data base and compare of analysis results repository based on three-dimensional bullet data, builds three-dimensional shoot mark comparison software platform.
Accompanying drawing explanation:
Accompanying drawing 1 is the cartridge mark comparison method figure based on multi-dimension feature extraction of the present invention.
Accompanying drawing 2 is three-dimensional point set data plots of the shoot mark manifold that gathers of the present invention time rib 1.
Accompanying drawing 3 is three-dimensional point set data plots of the shoot mark manifold that gathers of the present invention time rib 2.
Accompanying drawing 4 is three-dimensional point set data plots of the shoot mark manifold that gathers of the present invention time rib 3.
Accompanying drawing 5 is three-dimensional point set data plots of the shoot mark manifold that gathers of the present invention time rib 4.
Accompanying drawing 6 is 3DZD figure that bullet 1 vestige of the present invention is corresponding.
Accompanying drawing 7 is 3DZD figure corresponding after bullet 1 of the present invention rotates to an angle.
Accompanying drawing 8 is 3DZD figure that bullet 2 vestiges of the present invention are corresponding.
Accompanying drawing 9 is 3DZD that bullet 1 vestige of the present invention is corresponding and the comparison diagram of other bullet trace character pair.
Embodiment:
Embodiment 1:
A kind of cartridge mark comparison method based on multi-dimension feature extraction, extract the multidimensional characteristic of bullet data, associating multidimensional characteristic advantage realizes the accurate compare of analysis of shoot mark data, description problem for the three-dimensional topology architectural characteristic of three-dimensional shoot mark data, the three-dimensional feature extracting method of proposition based on 3D-Zernike, the cartridge mark comparison method based on multi-dimension feature extraction comprises the following steps:
Extract Mean curve;
Feature is slightly mated;
In area-of-interest, carry out three-dimensional feature extraction and analysis;
Compare of analysis.
The block diagram of the cartridge mark comparison method based on multi-dimension feature extraction is as shown in Figure 1:
Embodiment 2
According to the cartridge mark comparison method based on multi-dimension feature extraction described in embodiment 1, described extraction Mean curve, adopt adaptive average value filtering shoot mark data to be carried out to the pre-service of noise remove, projection to pretreated shoot mark time rib three-dimensional data on XOZ axle, shoot mark tendency can well be reflected in drop shadow curve, the average of Qu Ge drop shadow curve on Z axis, the curve being formed by average.
Embodiment 3:
According to the cartridge mark comparison method based on multi-dimension feature extraction described in embodiment 1 or 2, described feature is slightly mated, because each bullet of sample data has rib data, i.e. c1, c2, c3 and c4 4 times.When measure gathering, by one rib shoot mark, rotate to another shoot mark, unified according to a direction rotation (or counterclockwise, or clockwise).And the inferior rib scratch data of sample, also according to measuring sequence, provide numbering c1, c2, c3 and c4.But each scratch data is not corresponding according to numbering, so there are following four kinds of matching schemes:
Scheme
:
Scheme
:
Scheme
:
Scheme
:
If respectively according to above-mentioned four kinds of schemes, solve and can be used for the amount of Expressive Features similarity and solve corresponding similarity, therefrom select the comparison scheme that similarity is the highest. due to feature extraction with to solve similarity consuming time more, and in 4 schemes, only has an optimum, if can establish comparison scheme by simple and effective algorithm, can save for 75% the time that solves.In this case, order matching algorithm is very necessary.
By to the analysis of training sample data and experiment, proposed the order Matching Model of estimating based on similar herein, similar to estimate be to using the whether close basis as considering of the direction of two pattern vectors, and vector magnitude is unimportant, two pattern vectors are more similar, it is similar estimate larger.Conventional similar estimate have included angle cosine, related coefficient, index similarity coefficient etc.Consider the distribution characteristics of data sample, adopt related coefficient to estimate to differentiate matching order as similar herein, wherein related coefficient is the vector angle cosine after data center.
If
,
, formula of correlation coefficient is:
（4-1）
Wherein,
be respectively
mean vector, related coefficient is constant for translation, rotation and the yardstick convergent-divergent of coordinate system
.
Matching strategy is as follows:
Step 1 pair filtered rib data matrix
with the unit of classifying as, carry out summation operation, obtain shoot mark three-dimensional data and exist
the Mean curve of projection in plane, specific implementation formula is as follows:
（4-2）
Wherein,
;
, 3 represent two times different rib data.
Step 2 is in order to improve the similar precision of estimating, right
data are carried out pre-service, obtain effectively reflecting the data of shoot mark curved surface features.Preprocessing process is as follows:
1) calculate
; (one)
2) right
middle element, if
, order
.
So just can extract the data that can reflect shoot mark curved surface features.
Step 3, based on formula of correlation coefficient (4-1), solves the similar matrix of estimating
.Wherein
, represent four kinds of different matching schemes;
, represent the matching degree of four the corresponding Mean curves of rib of different bullets under four kinds of matching schemes.Similarly estimate matrix
calculation expression as follows:
（4-3）
Wherein
,
be in Step2
, 0.9 is obtained by experiment.
Step 4 pair obtains estimates matrix
do following processing, obtain matching scheme
corresponding similar measure value:
（4-4）
Right
each row vector is carried out descending sort, because inferior rib shoot mark generally has 2-3, coincide better, so herein to mate similar estimate have two kinds of defining modes, principle is as follows:
1. get
in similarly estimate two maximum averages final coupling be similar estimates as it, formula is as follows:
（4-5）
2. get
in similarly estimate three maximum averages final coupling be similar estimates as it, formula is as follows:
（4-6）
Wherein
finger is sued for peace to data,
with
be 4 to take advantage of 1 column vector.
Two features in step 5 pair step 4 are described
with
carry out Fusion Features, thereby establish matching order, concrete steps are as follows:
1. if
with
the scheme of establishing is identical, for
, Best similarity is estimated scheme and is
, with reference to this scheme, carry out feature extraction comparison simultaneously.
2. if
with
the scheme of establishing is different, and is respectively
with
, defined feature is described:
Scheme
with
average under two states:
（4-7）
Scheme
with
average under two states:
（4-8）
If
, get
as optimum matching scheme, otherwise get
for optimum matching scheme.
Embodiment 4:
According to the cartridge mark comparison method based on multi-dimension feature extraction described in embodiment 1 or 2,3, described in area-of-interest, carry out three-dimensional feature extraction and analysis, according to the preliminary matching result that obtains shoot mark, extract on this basis three-dimensional feature and do further the matching analysis.The three-dimensional feature of three-dimensional data is different from the topological structure that two dimensional character can be good at response data.The domestic research to cartridge mark is at present less, more there is no the research of three-dimensional shoot mark feature extraction aspect.Feature for three-dimensional shoot mark data, what three-dimensional invariant moment features herein adopted is 3D-Zernike descriptor, this feature is proposed to identify for objective by N. Canterakis, M. Novotni and R. Klein improve its algorithm subsequently, R. the people such as D. Mill á n finds that the advantageous property of this descriptor uses it for distinguishing of medical image medium vessels knurl, and domestic researchist uses it for the terrain match based on DEM.3D-Zernike descriptor can fully reflect the three-D space structure of target, and the descriptor of higher order more can be described the details of the three dimensions shape of target.This stereoscopic features not only can keep good translation, yardstick, and rotational invariance, and also it is little to have the redundancy of information representation, the high efficiency of information representation, robustness.3D-Zernike square can overcome different angles, the impact of the noises such as different light on data analysis.
3D-Zernike square is to be defined on the basis of spheric harmonic function.
(1) spheric harmonic function
Spheric harmonic function form class on a spheroid is similar to Fourier transform, the spitting image of being for sine or the cosine function of a line or a ball, is defined as follows:
(5-1)
Wherein
for normalized factor
represent corresponding Legendre function.
Spheric harmonic function forms a vector:
(5-2)
This vector is for this
n-dimensional subspace n is complete invariable rotary.
(2) hamonic function polynomial expression
According to the transformational relation of rectangular coordinate system and spherical coordinate system
(5-3)
Hamonic function polynomial expression
be defined as:
(5-4)
For corresponding Legendre function, adopt integral formula, and convert thereof into rectangular coordinate, hamonic function polynomial expression can be write as to following form:
(5-5)
Here
that normalized factor is defined as:
(5-6)
In above-mentioned formula, be for
situation,
time, hamonic function polynomial expression meets symmetric relation:
(5-7)
(3) derivation of 3D-Zernike square:
3D-Zernike function
be defined as:
(5-8)
Here limit
, and
for even number.Can be by above-mentioned house property hamonic function polynomial expression
write as rectangular coordinate form as follows:
(5-9)
Here
and utilize a coefficient
carry out the normalization of bonding circle of position inner function:
(5-10)
Radial polynomial is defined as follows:
(5-11)
Normalized Relation formula is as follows:
(5-12)
For 3D-Zernike function, there is identical invariant relation with ball is humorous.If function is formed
the vector of dimension
, for each
rotation matrix P has following relation arbitrarily:
(5-13)
We can define the 3D-Zernike square of a three-dimensional body accordingly
can be expressed as:
(5-14)
In formula
for three-dimensional body function.
Same 3D-Zernike square
also there is the symmetric property the same with formula (3-12):
(5-15)
In order to keep rotational invariance, still need as spheric harmonic function, under each multiplicity l, corresponding one comprises
the vector of element
, definition 3D-Zernike descriptor is:
, be used stereoscopic features.
The computing method of 3D-Zernike descriptor are as follows:
(1) computing formula that quantizes
3D-Zernike function
can be expressed as:
(5-16)
Wherein
for complexity coefficient:
(5-17)
If
So 3D-Zernike square
can be expressed as:
(5-18)
Wherein
for three-dimensional geometry square, it is defined as:
(5-19)
By the definition of 3D-Zernike descriptor (being called for short 3DZD) above, calculating will be to all n, l, and m combination is calculated, and we also will be to all for the calculating of the 3D-Zernike descriptor of certain exponent number n here
in meet
for all n of even number, l, m combination is calculated.
Complexity coefficient
for all, meet
r, s, t combination adds up, and need to first calculate.Because this step is totally independent of certain specific objective, this coefficient and targeted species are irrelevant, so can calculate in advance in the situation that not considering target.Due to for for some (n, l, m) combination, the coefficient obtaining is 0 again, so as long as we are by the coefficient of those non-zeros and corresponding r, and s, t is stored in a coefficient list accordingly, in order to subsequent calculations, uses.
(2) step of calculating 3D-Zernike square and 3D-Zernike descriptor is as follows:
I normalization
One group of shoot mark time rib three-dimensional coordinate (x, y, z) according to input, first calculates target center of gravity, then true origin is transformed to center of gravity place, then carries out yardstick convergent-divergent, by objective mapping in unit ball.
Ii computational geometry square
Calculate all satisfied
all positive integer r, s, t combines corresponding geometric moment
Iii calculates 3D-Zernike square
According to (5-18) formula, calculate 3D-Zernike square.Here be noted that when summation as long as use the nonzero coefficient being stored in complexity coefficient list to reduce calculated amount.In addition for
situation, can, according to the character of 3D-Zernike square, utilize symmetric relation formula (5-15) to obtain corresponding result.
Iv calculates 3D-Zernike descriptor
Embodiment 5:
According to the cartridge mark comparison method based on multi-dimension feature extraction described in embodiment 1 or 2,3,4, described compare of analysis, calculate the Euclidean distance between the 3D-Zernike descriptor of different shoot mark data, the similarity degree of describing between two data by the size of Euclidean distance mates identification.
Embodiment 6:
According to the cartridge mark comparison method based on multi-dimension feature extraction described in embodiment 1 or 2,3,4,5,
Three-dimensional shoot mark feature extraction and analysis are tested:
In experiment, adopt three-dimensional shoot mark data collector to gather rib data 60 groups of 15 bullets times, as listed 6 groups of typical time rib data in Fig. 2.In experiment, get 10 rank 3D-Zernike descriptors, corresponding 4 groups the rib data of each bullet, 1 * 36 vector of each the corresponding 10 rank 3D-Zernike descriptors of rib data, the vector of each bullet correspondence 1 * 144.Under different angles, the shoot mark data of scanning exist certain rotation to change, and extract 3DZD feature, and between calculated characteristics, Euclidean distance is carried out analysis and comparison, and result is as shown in Fig. 6, Fig. 7, Fig. 8 and Fig. 9.Wherein Fig. 6 and the identical shoot mark three-dimensional data for gathering under different angles in Fig. 7, its 3DZD difference is very little as seen; Figure 8 shows that from 3DZD feature corresponding to the shoot mark data of another gun, shown in it and Fig. 6 and difference shown in Fig. 7 obvious.And as shown in Figure 9, for the comparative result of 3DZD corresponding to bullet 1 vestige and other bullet trace character pair, obviously can find out, from same two bullets (1 of gun, 13) characteristic of correspondence difference is less, and differs greatly with other shoot mark data characteristic of correspondence.
For the description problem of the three-dimensional topology architectural characteristic of three-dimensional shoot mark data, the three-dimensional feature extracting method based on 3D-Zernike is proposed.Make full use of the translation of 3D-Zernike descriptor, yardstick, rotational invariance, the high efficiency of information representation, towards true three-dimension shoot mark data analysis and comparison.Experimental verification the validity of this feature for three-dimensional shoot mark data description, 3D-Zernike descriptor reveals good homogeneity and otherness to the description list of the inferior rib data of shoot mark.
Claims (5)
1. the cartridge mark comparison method based on multi-dimension feature extraction, it is characterized in that: the multidimensional characteristic that extracts bullet data, associating multidimensional characteristic advantage realizes the accurate compare of analysis of shoot mark data, description problem for the three-dimensional topology architectural characteristic of three-dimensional shoot mark data, the three-dimensional feature extracting method of proposition based on 3D-Zernike, the cartridge mark comparison method based on multi-dimension feature extraction comprises the following steps:
Extract Mean curve;
Feature is slightly mated;
In area-of-interest, carry out three-dimensional feature extraction and analysis;
Compare of analysis.
2. the cartridge mark comparison method based on multi-dimension feature extraction according to claim 1, it is characterized in that: described extraction Mean curve, adopt adaptive average value filtering shoot mark data to be carried out to the pre-service of noise remove, projection to pretreated shoot mark time rib three-dimensional data on XOZ axle, shoot mark tendency can well be reflected in drop shadow curve, the average of Qu Ge drop shadow curve on Z axis, the curve being formed by average.
3. the cartridge mark comparison method based on multi-dimension feature extraction according to claim 1 and 2, it is characterized in that: described feature is slightly mated, because each bullet of sample data has rib data 4 times, be c1, c2, c3 and c4, when measuring collection, by one rib shoot mark, rotate to another shoot mark, unified according to counterclockwise rotation or clockwise, the inferior rib scratch data of sample are also according to measuring sequence, provide numbering c1, c2, c3 and c4, but each scratch data is not corresponding according to numbering, so there are following four kinds of matching schemes:
Scheme
:
Scheme
:
Scheme
:
Scheme
:
.
4. according to the cartridge mark comparison method based on multi-dimension feature extraction described in claim 1 or 2 or 3, it is characterized in that: described in area-of-interest, carry out three-dimensional feature extraction and analysis, according to the preliminary matching result that obtains shoot mark, extract on this basis three-dimensional feature and do further the matching analysis, the three-dimensional feature of three-dimensional data is different from the topological structure that two dimensional character can be good at response data, 3D-Zernike descriptor fully reflects the three-D space structure of target, the descriptor of higher order more can be described the details of the three dimensions shape of target, this stereoscopic features not only can keep good translation, yardstick, rotational invariance, and the redundancy with information representation is little, the high efficiency of information representation, robustness, 3D-Zernike square overcomes different angles, the impact of the noises such as different light on data analysis.
5. according to the cartridge mark comparison method based on multi-dimension feature extraction described in claim 1 or 2 or 3 or 4, described compare of analysis, calculate the Euclidean distance between the 3D-Zernike descriptor of different shoot mark data, the similarity degree of describing between two data by the size of Euclidean distance mates identification.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410358260.6A CN104123546A (en) | 2014-07-25 | 2014-07-25 | Multi-dimensional feature extraction based bullet trace comparison method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410358260.6A CN104123546A (en) | 2014-07-25 | 2014-07-25 | Multi-dimensional feature extraction based bullet trace comparison method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN104123546A true CN104123546A (en) | 2014-10-29 |
Family
ID=51768951
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410358260.6A Pending CN104123546A (en) | 2014-07-25 | 2014-07-25 | Multi-dimensional feature extraction based bullet trace comparison method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104123546A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104680543A (en) * | 2015-03-18 | 2015-06-03 | 哈尔滨工业大学 | Method for estimating direction variation of digital surface models based on 3D-Zernike (three-dimensional-Zernike) moment phase analysis |
| CN106485258A (en) * | 2016-10-21 | 2017-03-08 | 中北大学 | A kind of line array CCD bullet location drawing that is based on is as rapid extraction processing method |
| CN109191502A (en) * | 2018-08-14 | 2019-01-11 | 南京工业大学 | A kind of method of automatic identification shell case trace |
| CN110555389A (en) * | 2019-08-09 | 2019-12-10 | 南京工业大学 | bullet line-bore trace identification method based on ridgelet transformation and rotation matching |
| CN112085045A (en) * | 2020-04-07 | 2020-12-15 | 昆明理工大学 | Linear trace similarity matching algorithm based on improved longest common substring |
| CN112424581A (en) * | 2018-07-13 | 2021-02-26 | Ntn株式会社 | Condition monitoring apparatus and condition monitoring system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101488223A (en) * | 2008-01-16 | 2009-07-22 | 中国科学院自动化研究所 | Image curve characteristic matching method based on average value standard deviation descriptor |
| US20090234587A1 (en) * | 2000-03-10 | 2009-09-17 | Smith Detection Inc. | Measuring and analyzing multi-dimensional sensory information for identification purposes |
| CN103246875A (en) * | 2013-05-09 | 2013-08-14 | 东南大学 | Three-dimensional facial recognition method based on elasticity matching of facial curves |
| CN103323817A (en) * | 2013-06-25 | 2013-09-25 | 中国人民解放军理工大学 | Airborne synthetic aperture radar sea surface wind vector retrieval method |
| CN103729882A (en) * | 2013-12-30 | 2014-04-16 | 浙江大学 | Point cloud relative pose estimation method based on three-dimensional curve matching |
-
2014
- 2014-07-25 CN CN201410358260.6A patent/CN104123546A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090234587A1 (en) * | 2000-03-10 | 2009-09-17 | Smith Detection Inc. | Measuring and analyzing multi-dimensional sensory information for identification purposes |
| CN101488223A (en) * | 2008-01-16 | 2009-07-22 | 中国科学院自动化研究所 | Image curve characteristic matching method based on average value standard deviation descriptor |
| CN103246875A (en) * | 2013-05-09 | 2013-08-14 | 东南大学 | Three-dimensional facial recognition method based on elasticity matching of facial curves |
| CN103323817A (en) * | 2013-06-25 | 2013-09-25 | 中国人民解放军理工大学 | Airborne synthetic aperture radar sea surface wind vector retrieval method |
| CN103729882A (en) * | 2013-12-30 | 2014-04-16 | 浙江大学 | Point cloud relative pose estimation method based on three-dimensional curve matching |
Non-Patent Citations (3)
| Title |
|---|
| 赵红毅: "一种枪弹痕迹自动检测优化仿真研究", 《计算机仿真》 * |
| 韩毅 等: "一种枪弹痕迹自动比对算法", 《计算机仿真》 * |
| 高凤娇 等: "基于 3D-Zernike 的三维弹痕特征提取技术", 《自动化技术与应用》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104680543A (en) * | 2015-03-18 | 2015-06-03 | 哈尔滨工业大学 | Method for estimating direction variation of digital surface models based on 3D-Zernike (three-dimensional-Zernike) moment phase analysis |
| CN104680543B (en) * | 2015-03-18 | 2017-08-25 | 哈尔滨工业大学 | Digital surface model orientation discrepancy method of estimation based on 3D Zernike square phase analysises |
| CN106485258A (en) * | 2016-10-21 | 2017-03-08 | 中北大学 | A kind of line array CCD bullet location drawing that is based on is as rapid extraction processing method |
| CN106485258B (en) * | 2016-10-21 | 2019-03-05 | 中北大学 | One kind being based on line array CCD bullet location drawing picture rapidly extracting processing method |
| CN112424581A (en) * | 2018-07-13 | 2021-02-26 | Ntn株式会社 | Condition monitoring apparatus and condition monitoring system |
| CN112424581B (en) * | 2018-07-13 | 2023-09-22 | Ntn株式会社 | Condition monitoring device and condition monitoring system |
| US12007310B2 (en) | 2018-07-13 | 2024-06-11 | Ntn Corporation | State monitoring device and state monitoring system |
| CN109191502A (en) * | 2018-08-14 | 2019-01-11 | 南京工业大学 | A kind of method of automatic identification shell case trace |
| CN109191502B (en) * | 2018-08-14 | 2021-08-24 | 南京工业大学 | Method for automatically identifying cartridge case trace |
| CN110555389A (en) * | 2019-08-09 | 2019-12-10 | 南京工业大学 | bullet line-bore trace identification method based on ridgelet transformation and rotation matching |
| CN110555389B (en) * | 2019-08-09 | 2022-02-25 | 南京工业大学 | Bullet line-bore trace identification method based on ridgelet transformation and rotation matching |
| CN112085045A (en) * | 2020-04-07 | 2020-12-15 | 昆明理工大学 | Linear trace similarity matching algorithm based on improved longest common substring |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104123546A (en) | Multi-dimensional feature extraction based bullet trace comparison method | |
| CN102236794B (en) | Recognition and pose determination of 3D objects in 3D scenes | |
| Jégou et al. | Improving bag-of-features for large scale image search | |
| Amanatiadis et al. | Evaluation of shape descriptors for shape-based image retrieval | |
| Guo et al. | A novel local surface feature for 3D object recognition under clutter and occlusion | |
| CN106919944B (en) | ORB algorithm-based large-view-angle image rapid identification method | |
| CN103109307B (en) | For making the method and apparatus of Three-dimension object recognition image data base | |
| CN104090972A (en) | Image feature extraction and similarity measurement method used for three-dimensional city model retrieval | |
| JP2014081347A (en) | Method for recognition and pose determination of 3d object in 3d scene | |
| JP2012168941A (en) | Method for representing 3d objects or objects with descriptors | |
| CN114972459B (en) | Point cloud registration method based on low-dimensional point cloud local feature descriptor | |
| Guo et al. | 3D free form object recognition using rotational projection statistics | |
| CN110930456A (en) | Three-dimensional identification and positioning method of sheet metal part based on PCL point cloud library | |
| CN109920050B (en) | Single-view three-dimensional flame reconstruction method based on deep learning and thin plate spline | |
| CN104851095A (en) | Workpiece image sparse stereo matching method based on improved-type shape context | |
| Ye et al. | Place recognition in semi-dense maps: Geometric and learning-based approaches | |
| Tao et al. | A pipeline for 3-D object recognition based on local shape description in cluttered scenes | |
| Mouine et al. | Combining leaf salient points and leaf contour descriptions for plant species recognition | |
| Zhou et al. | Histograms of Gaussian normal distribution for 3D feature matching in cluttered scenes | |
| Wang et al. | Target recognition and localization of mobile robot with monocular PTZ camera | |
| CN107330436B (en) | Scale criterion-based panoramic image SIFT optimization method | |
| Ahmad et al. | A fusion of labeled-grid shape descriptors with weighted ranking algorithm for shapes recognition | |
| Zhao et al. | Research on Feature Matching of an Improved ORB Algorithm | |
| Bo et al. | A robust image registration algorithm used for panoramic image mosaic | |
| Lin et al. | Improving the efficiency and accuracy of sift image matching |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20141029 |