CN110378271A - A kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter - Google Patents
A kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter Download PDFInfo
- Publication number
- CN110378271A CN110378271A CN201910622166.XA CN201910622166A CN110378271A CN 110378271 A CN110378271 A CN 110378271A CN 201910622166 A CN201910622166 A CN 201910622166A CN 110378271 A CN110378271 A CN 110378271A
- Authority
- CN
- China
- Prior art keywords
- equipment
- algorithm
- image
- noise
- gait recognition
- 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.)
- Granted
Links
- 230000005021 gait Effects 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000012216 screening Methods 0.000 title claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 85
- 238000012360 testing method Methods 0.000 claims abstract description 23
- 230000000694 effects Effects 0.000 claims abstract description 7
- 238000011084 recovery Methods 0.000 claims description 13
- 238000011156 evaluation Methods 0.000 claims description 9
- 230000006870 function Effects 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 4
- 230000035945 sensitivity Effects 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 4
- 238000013135 deep learning Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 claims description 3
- 230000007613 environmental effect Effects 0.000 claims description 3
- 238000001303 quality assessment method Methods 0.000 claims description 3
- 238000000926 separation method Methods 0.000 claims 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims 1
- 230000008569 process Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000013475 authorization Methods 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000008092 positive effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 239000002344 surface layer Substances 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/217—Validation; Performance evaluation; Active pattern learning techniques
- G06F18/2193—Validation; Performance evaluation; Active pattern learning techniques based on specific statistical tests
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G06T5/70—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/215—Motion-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/20—Analysis of motion
- G06T7/254—Analysis of motion involving subtraction of images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
- G06V40/23—Recognition of whole body movements, e.g. for sport training
- G06V40/25—Recognition of walking or running movements, e.g. gait recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30196—Human being; Person
Abstract
A kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter, comprising the following steps: (1), to hardware device do objective parameter and assess；(2), the human region in gait test library is demarcated；(3), the test library classification based on noise；(4), the test library information flag based on fuzziness；(5), Gait Recognition test is carried out to any particular algorithms；(6), the signal-to-noise ratio under corresponding camera is calculated, corresponding threshold signal-to-noise ratio range is obtained；(7), step (6) are carried out to all devices to determine, is rejected；(8), corresponding maximum fuzzy tolerance section is obtained；(9), step 8 is carried out to surplus equipment to determine, reject.The present invention can filter out the highest capture apparatus of cost performance；Capture apparatus can be more objectively selected, and obtains optimal recognition effect；Algorithm and equipment can adjust mutually, this makes Gait Recognition system building more flexible, reduce hardware cost.
Description
Technical field
The present invention relates to technical field of image processing.
Background technique
It is that many smart device manufacturers are extremely paid close attention to for the problem that Gait Recognition acquires the optimal screening problem of equipment,
Smart machine is related to biological characteristic validation and authorization, while improving accuracy of identification as far as possible, needs to reduce unnecessary hardware
Cost accomplishes benefit.
In terms of lectotype selection, assessed at this stage just for equipment itself, such as resolution ratio, frame per second, from surface layer meaning
On, the equipment that resolution ratio is higher, frame per second is higher thinks that quality is better, and there is no to lectotype selection for the relevant parameter of algorithm
There is positive effect.
It can not normally be identified in Algorithm for gait recognition for the low quality video of shooting process appearance at this stage,
Think that quality standard is not achieved, then select another money acquisition equipment, i.e. process algorithm itself is unrelated with hardware selection.
For the integrating process of Gait Recognition system, often there are multiple alternative Algorithm for gait recognition or deposited
Algorithm interface, shooting hardware (such as camera) condition be it is independent with algorithm, cannot achieve the highest hardware of cost performance and set
Standby screening, hardware cost are high.
Summary of the invention
In order to solve the above problem caused by Gait Recognition acquisition equipment and algorithm independence, the present invention provides one kind to be based on
The Gait Recognition equipment screening technique of quality dimensions assessment parameter.
Present invention technical solution used for the above purpose is: a kind of gait based on quality dimensions assessment parameter
Identify equipment screening technique, comprising the following steps:
(1), assume that the recognizer of Gait Recognition system shares N number of, to be selected hardware device and has M, based on actually answering
Shooting environmental is emulated with environment construction, objective parameter is done to hardware device and is assessed, the fuzzy parameter of camera lens is obtained and makes an uproar
The horizontal parameter of sound；
(2), existing gait test library is arranged, the human region in library is demarcated；
(3), the test library classification based on noise, comprising the following steps:
3-1, an image recovery is done to the arbitrary image in test library, reduction does not include the image of noise, utilizes recovery
Image afterwards makes the difference with original image, difference, that is, noise figure of reservation；
3-2, statistical noise energy calculate two norm of difference of 3-1；
3-3, statistics original image energy, calculate two norms of image after the recovery of 3-1；
3-4., signal noise ratio (snr) of image=20log (original image energy/noise energy) is calculated；
3-5, repeat 3-1~3-4, calculate the signal-tonoise information snr of all images, to signal-to-noise ratio do maximum value with
Minimum Data-Statistics obtain signal-to-noise ratio value range, fix the subregion of step-length to entire scope, obtain different signal-to-noise ratio areas
Between, the label in section where then finding all figure signal-to-noise ratio；
(4), the test library information flag based on fuzziness, comprising the following steps:
4-1, image done to the image after denoising in step 3-1 restore, reduction clearly image, calculate original image with it is extensive
The Fourier transformation of image, is then divided by, obtains the frequency domain representation of degenrate function after multiple；
4-2, the low-pass cut-off frequencies for counting degenrate function；
4-3. repeats 4-1~4-2, calculates the cutoff frequency information of all images, then does most to recording frequency
Big value and minimum Data-Statistics, obtain frequency value range, fix the subregion of step-length to entire scope, obtain different frequency zones
Between, the label in section where then finding all figure frequencies；
(5), to any particular algorithms carry out Gait Recognition test, obtain the algorithm wrong identification sample and algorithm with it is all kinds of
The sensibility of objective condition, steps are as follows:
5-1, using signal-to-noise ratio section as abscissa, with section recognition failures sample quantity be ordinate, count signal-to-noise ratio
Introduce the statistic histogram of recognition failures；
5-2. is classified as abscissa with different fuzzinesses, and the quantity with section recognition failures sample is ordinate, counts mould
The statistic histogram of paste degree introducing recognition failures；
(6), for all wrong identification samples, to the noise level of hardware device objective evaluating and respective corresponding step
The image energy of 3-1 denoising calculates the signal-to-noise ratio snr under corresponding camera*, correspond to the critical letter of algorithms of different in step 5-1
It makes an uproar than range, takes algorithm minimum to noise sensitivity in all algorithms, obtain corresponding threshold signal-to-noise ratio range, it is assumed that snr*
Less than the low_snr of the algorithm, then it is assumed that the noise of equipment level is excessively high, can be larger for doing face identification error, it is not recommended that
It uses；It is assumed that snr*In the critical range of the algorithm, it is believed that the equipment can do recognition of face, but discrimination also promotes sky
Between；It is assumed that snr*Greater than the high_snr of the algorithm, it is believed that the equipment can do recognition of face, and its noise level knows face
The discrimination of other algorithm has little effect；
(7), step (6) are carried out to all devices to determine, is rejected it is not recommended that the capture apparatus used；
(8), for noise level meets the equipment of algorithm discrimination, the maximum fuzzy of algorithms of different in step 5-2 is taken
Tolerance section takes algorithm minimum to blur detection in all algorithms, obtains corresponding maximum fuzzy tolerance section, it is assumed that hard
The fuzzy parameter of part equipment evaluation is the fuzziness F of equipment evaluation*Inverse, if the fuzziness F of equipment evaluation*Greater than the algorithm
High_F, then it is assumed that the image that the equipment is taken is excessively fuzzy, can larger for doing Gait Recognition error, it is not recommended that use；
It is assumed that F*In the tolerance section of the algorithm, it is believed that the equipment can do Gait Recognition, but there are also rooms for promotion for discrimination；It is assumed that
F*Less than the low_F of the algorithm, which can do Gait Recognition, and its Fuzzy Level is several to the discrimination of Algorithm for gait recognition
Do not influence；
(9), step 8 is carried out to surplus equipment to determine, reject it is not recommended that the capture apparatus used.
Further include step (10), if there is no equipment to meet condition after rejecting, is handled by following three kinds of methods: a, root
The robustness of algorithms of different, method are improved according to quality assessment result are as follows: increase pretreatment denoising or deblurring；B, expansion is utilized
The method for filling trained low-quality image library reduces algorithm susceptibility based on deep learning；C, expand hardware and screen range, increase
Hardware cost reduces the fuzzy and noise level of hardware.
Further include step (11), if there are also equipment residues after rejecting, passes through equipment cost and actual application environment
Discrimination decides whether to select to influence discrimination small equipment.
In the step (2), gait test library personnel amount is at ten thousand grades, everyone gait frame number is more than 500 frames, mark
Determining method is artificial calibration or automatic Calibration, and automatic calibration method is the automatic calibration method of DarkNet model.
In step 5-1, SNR ranges are the threshold signal-to-noise ratio range [low_snr of algorithmi,high_snri], i ∈ [1,
N], when section definition is thinner, the range is smaller.
In the step 5-2, the fuzzy interval before fuzziness is maximum fuzzy tolerance section [low_Fi,high_Fi], i
∈ [1, N], when section definition is thinner, the range is smaller.
In the step 3-1, image recovery method is that denoising is filtered from coding and/or edge self-adaption.
In the step 4-1, image recovery method is super-resolution algorithms and/or deblurring algorithm.
Gait Recognition equipment screening technique based on quality dimensions assessment parameter of the invention, based on Algorithm for gait recognition
The influence that image is imaged in susceptibility and actual photographed hardware device, can filter out the highest capture apparatus of cost performance；It utilizes
The quality tolerance of Algorithm for gait recognition, can more objectively select capture apparatus, and obtain optimal recognition effect；Algorithm
It can be adjusted mutually with equipment, this makes Gait Recognition system building more flexible, reduces hardware cost.
Detailed description of the invention
Fig. 1 is signal-to-noise ratio of the present invention and error sample distribution statistics result legend.
Fig. 2 is cutoff frequency of the present invention and error sample distribution statistics result legend.
Specific embodiment
It is mainly noise and clarity (or fuzziness) two aspects, noise that capture apparatus, which is embodied in the parameter in quality,
It is to be obscured mainly caused by the point spread function (PSF) of camera lens as caused by hardware circuit.Therefore, this programme is by commenting
Algorithm for gait recognition is surveyed to the sensibility of noise, fuzziness, it is specified that capture apparatus parameter.
The present invention is based on the Gait Recognition equipment screening techniques of quality dimensions assessment parameter, comprising the following steps: it is assumed that step
The recognizer of state identifying system, which shares N number of, to be selected hardware device, M, for specific application environment:
(1), based on actual application environment building emulation shooting environmental, objective parameter is done to hardware device and is assessed, hardware is set
Standby objective evaluation method can be with referenced patent " a kind of camera Auto-Test System " (application number 201821039835.8), mainly
Obtain the fuzzy parameter (such as MTF50) and noise level parameter (such as signal-to-noise ratio) of camera lens.
(2), existing gait test library is arranged, at ten thousand grades, everyone gait frame number at least exists personnel amount
More than 500 frames, it is therefore an objective to provide the gait figure that can be photographed in various actual environments, then be marked to the human region in library
Calmly, scaling method can be artificial calibration, be also possible to automated process calibration, and automated process can be oneself of DarkNet model
Dynamic scaling method (face recognition algorithms can not do automatic segmentation in order to prevent).
(3), the test library classification based on noise:
3-1. does an image to the arbitrary image in test library and restores, and target is that reduction does not include noise as much as possible
Image, the technology that can be used include: to denoise from coding techniques (Lu X, Tsao Y, Matsuda S, et al.Speech
enhancement based on deep denoising autoencoder[C]//Interspeech.2013:436-
440.), edge self-adaption filtering technique (Side Window Filtering.CVPR 2019), using after recovery image with
Original image makes the difference, it is believed that difference, that is, noise figure of reservation；
3-2. statistical noise energy, specific method are two norms of difference for calculating 3-1；
3-3. count original image energy, specific method be calculate 3-1 recovery after image two norms；
3-4. calculates signal noise ratio (snr) of image=20log (original image energy/noise energy)；
3-5. repeats 3-1~3-4, calculates the signal-tonoise information snr of all images, then does maximum to signal-to-noise ratio
Value and minimum Data-Statistics, obtain signal-to-noise ratio value range, fix the subregion of step-length to entire scope, obtain different signal-to-noise ratio
Section, the label in section where then finding all figure signal-to-noise ratio.
(4), the test library information flag based on fuzziness:
4-1. does an image to the image after denoising in 3-1 and restores, and target is to restore clearly image as much as possible, can
It include: super-resolution technique (Yang J, Wright J, Huang T S, et al.Image super- with the technology used
resolution via sparse representation[J].IEEE transactions on image
Processing, 2010,19 (11): 2861-2873.), deblurring algorithm (Nah S, Hyun Kim T, Mu Lee K.Deep
multi-scale convolutional neural network for dynamic scene deblurring[C]//
Proceedings of the IEEE Conference on Computer Vision and Pattern
Recognition.2017:3883-4891.), the Fourier transformation of image, is then divided by, is moved back after calculating original image and restoring
Change the frequency domain representation of function；
Low-pass cut-off frequencies (the frequency of amplitude-frequency response energy attenuation to dominant frequency energy 50% of 4-2. statistics degenrate function
Value)；
4-3. repeats 4-1~4-2, calculates the cutoff frequency information of all images, then does most to recording frequency
Big value and minimum Data-Statistics, obtain frequency value range, fix the subregion of step-length to entire scope, obtain different frequency zones
Between, the label in section where then finding all figure frequencies.
(5), to any particular algorithms carry out Gait Recognition test, obtain the algorithm wrong identification sample and algorithm with it is all kinds of
The sensibility of objective condition:
For 5-1. using signal-to-noise ratio section as abscissa, the quantity with section recognition failures sample is ordinate, counts signal-to-noise ratio
The statistic histogram of recognition failures is introduced, generally, signal-to-noise ratio is bigger, and wrong identification sample size is lower, as shown in Figure 1, frame
Constituency is threshold signal-to-noise ratio range, and when being more than certain signal-to-noise ratio, then high signal-to-noise ratio influences less discrimination, this programme
The SNR ranges are referred to as the threshold signal-to-noise ratio range [low_snr of the algorithmi,high_snri], i ∈ [1, N] works as section definition
Thinner, the range is smaller；
5-2. is classified as abscissa with different fuzzinesses, and the quantity with section recognition failures sample is ordinate, counts mould
Paste degree introduces the statistic histogram of recognition failures, and generally, fuzziness is higher, and low-pass cut-off frequencies are lower, and fuzziness can recognize
For the inverse for being low-pass cut-off frequencies, wrong identification sample is higher, as shown in Fig. 2, frame constituency is maximum fuzzy tolerance section, when
When more than certain fuzziness, identification sample size will appear geometric progression increase, and this programme claims fuzzy before the fuzziness
Section is maximum fuzzy tolerance section [low_Fi,high_Fi], i ∈ [1, N], when section definition is thinner, the range is smaller.
(6), for all wrong identification samples, to the noise level of any hardware equipment objective evaluating with it is respectively corresponding
The image energy of 3-1 denoising calculates the signal-to-noise ratio snr under corresponding camera*, correspond to the threshold signal-to-noise ratio of algorithms of different in 5-1
Range takes the algorithm (low_snr of critical range minimum to noise sensitivity in all algorithmsiSmaller, algorithm is to noise-sensitive
Spend lower), obtain corresponding threshold signal-to-noise ratio range, it is assumed that snr*Less than the low_snr of the algorithm, then it is assumed that the noise of equipment
It is horizontal excessively high, it can be larger for doing face identification error, it is not recommended that use；It is assumed that snr*In the critical range of the algorithm, recognize
Recognition of face can be done for the equipment, but there are also rooms for promotion for discrimination；It is assumed that snr*Greater than the high_snr of the algorithm, it is believed that
The equipment can do recognition of face, and its noise level has little effect the discrimination of face recognition algorithms.
(7), 6 are carried out to all devices to determine, is rejected it is not recommended that the capture apparatus used.
(8), for noise level meets the equipment of algorithm discrimination, the maximum fuzzy of algorithms of different in 5-2 is taken to tolerate
Section takes the algorithm (high_F of tolerance minimum to blur detection in all algorithmsiBigger, algorithm is to fuzzy sensitivity
Spend lower), obtain corresponding maximum fuzzy tolerance section, it is assumed that the MTF50 of hardware device assessment is equipment evaluation fuzziness F*
Inverse (non-frequency-domain index requires transformation into frequency domain low-pass cut-off frequencies), if equipment evaluation fuzziness F*Greater than the algorithm
High_F, then it is assumed that the image that the equipment is taken is excessively fuzzy, can be larger for doing Gait Recognition error, it is not recommended that use；It is false
Locking equipment assesses fuzziness F*In the tolerance section of the algorithm, it is believed that the equipment can do Gait Recognition, but discrimination is also
Room for promotion；It is assumed that equipment evaluation fuzziness F*Less than the low_F of the algorithm, it is believed that the equipment can do Gait Recognition, and its
Fuzzy Level has little effect the discrimination of Algorithm for gait recognition.
(9), 8 are carried out to surplus equipment to determine, is rejected it is not recommended that the capture apparatus used.
(10) if, there is no equipment to meet condition after rejecting, can be with: different calculations are improved according to quality assessment result
The robustness of method, specific practice can be increase pretreatment denoising or deblurring, also can use expansion training low-quality spirogram
As the method in library reduces algorithm susceptibility based on deep learning, hardware screening range can also be expanded, increase hardware cost, drop
The fuzzy and noise level of low hardware.
(11) if, through rejecting after, there are also equipment residue, examined by the discrimination of equipment cost and actual application environment
Consider the equipment for whether selecting to have little effect discrimination.
The present invention is by testing mass parameter to the susceptibility of Algorithm for gait recognition, based on Gait Recognition module in system
Robustness, to the sensibility of each condition, select most suitable hardware, keep the algorithm of system hardware-related, reduce hardware at
This.
The present invention is described by embodiment, and those skilled in the art know, is not departing from spirit of the invention
In the case where range, various changes or equivalence replacement can be carried out to these features and embodiment.In addition, in religion of the invention
It leads down, can modify to these features and embodiment to adapt to particular situation and material without departing from essence of the invention
Mind and range.Therefore, the present invention is not limited to the particular embodiment disclosed, fallen with claims hereof
Embodiment in range belongs to protection scope of the present invention.
Claims (8)
1. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter, it is characterised in that: the following steps are included:
(1), assume that the recognizer of Gait Recognition system shares N number of, to be selected hardware device and there are M, be based on practical application ring
Border building emulation shooting environmental, does objective parameter to hardware device and assesses, obtain the fuzzy parameter and noise water of camera lens
Flat parameter；
(2), existing gait test library is arranged, the human region in library is demarcated；
(3), the test library classification based on noise, comprising the following steps:
3-1, an image recovery is done to the arbitrary image in test library, reduction does not include the image of noise, after recovery
Image makes the difference with original image, difference, that is, noise figure of reservation；
3-2, statistical noise energy calculate two norm of difference of 3-1；
3-3, statistics original image energy, calculate two norms of image after the recovery of 3-1；
3-4., signal noise ratio (snr) of image=20log (original image energy/noise energy) is calculated；
3-5, repeat 3-1~3-4, calculate the signal-tonoise information snr of all images, maximum value and minimum are done to signal-to-noise ratio
Data-Statistics obtain signal-to-noise ratio value range, fix the subregion of step-length to entire scope, obtain different signal-to-noise ratio sections, so
The label in section where finding all figure signal-to-noise ratio afterwards；
(4), the test library information flag based on fuzziness, comprising the following steps:
4-1, an image recovery is done to the image after denoising in step 3-1, restore clearly image, after calculating original image and restoring
The Fourier transformation of image, is then divided by, and obtains the frequency domain representation of degenrate function；
4-2, the low-pass cut-off frequencies for counting degenrate function；
4-3. repeats 4-1~4-2, calculates the cutoff frequency information of all images, then does maximum value to recording frequency
With minimum Data-Statistics, frequency value range is obtained, fixes the subregion of step-length to entire scope, obtains different frequency separations,
Then the label in section where finding all figure frequencies；
(5), to any particular algorithms carry out Gait Recognition test, obtain the algorithm wrong identification sample and algorithm with it is all kinds of objective
The sensibility of condition, steps are as follows:
5-1, using signal-to-noise ratio section as abscissa, with section recognition failures sample quantity be ordinate, statistics signal-to-noise ratio introduce
The statistic histogram of recognition failures；
5-2. is classified as abscissa with different fuzzinesses, and the quantity with section recognition failures sample is ordinate, counts fuzziness
Introduce the statistic histogram of recognition failures；
(6), for all wrong identification samples, to the noise level of hardware device objective evaluating and respective corresponding step 3-1
The image energy of denoising calculates the signal-to-noise ratio snr under corresponding camera*, correspond to the critical noise of algorithms of different in step 5-1
Than range, algorithm minimum to noise sensitivity in all algorithms is taken, obtains corresponding threshold signal-to-noise ratio range, it is assumed that snr*It is small
In the low_snr of the algorithm, then it is assumed that the noise of equipment level is excessively high, can be larger for doing face identification error, it is not recommended that make
With；It is assumed that snr*In the critical range of the algorithm, it is believed that the equipment can do recognition of face, but discrimination also promotes sky
Between；It is assumed that snr*Greater than the high_snr of the algorithm, it is believed that the equipment can do recognition of face, and its noise level knows face
The discrimination of other algorithm has little effect；
(7), step (6) are carried out to all devices to determine, is rejected it is not recommended that the capture apparatus used；
(8), for noise level meets the equipment of algorithm discrimination, the maximum fuzzy of algorithms of different in step 5-2 is taken to tolerate
Section takes algorithm minimum to blur detection in all algorithms, obtains corresponding maximum fuzzy tolerance section, it is assumed that hardware is set
The fuzzy parameter of standby assessment is the fuzziness F of equipment evaluation*Inverse, if the fuzziness F of equipment evaluation*Greater than the algorithm
High_F, then it is assumed that the image that the equipment is taken is excessively fuzzy, can be larger for doing Gait Recognition error, it is not recommended that use；It is false
Determine F*In the tolerance section of the algorithm, it is believed that the equipment can do Gait Recognition, but there are also rooms for promotion for discrimination；It is assumed that F*
Less than the low_F of the algorithm, which can do Gait Recognition, and its Fuzzy Level to the discrimination of Algorithm for gait recognition almost
Do not influence；
(9), step 8 is carried out to surplus equipment to determine, reject it is not recommended that the capture apparatus used.
2. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter according to claim 1, special
Sign is: further including step (10), if not having equipment to meet condition after rejecting, is handled by following three kinds of methods: a, root
The robustness of algorithms of different, method are improved according to quality assessment result are as follows: increase pretreatment denoising or deblurring；B, expansion is utilized
The method for filling trained low-quality image library reduces algorithm susceptibility based on deep learning；C, expand hardware and screen range, increase
Hardware cost reduces the fuzzy and noise level of hardware.
3. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter according to claim 1, special
Sign is: further including step (11), if there are also equipment residues after rejecting, passes through equipment cost and actual application environment
Discrimination decides whether to select to influence discrimination small equipment.
4. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter according to claim 1, special
Sign is: in the step (2), gait test library personnel amount is at ten thousand grades, everyone gait frame number is more than 500 frames, mark
Determining method is artificial calibration or automatic Calibration, and automatic calibration method is the automatic calibration method of DarkNet model.
5. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter according to claim 1, special
Sign is: in step 5-1, SNR ranges are the threshold signal-to-noise ratio range [low_snr of algorithmi,high_snri], i ∈ [1,
N], when section definition is thinner, the range is smaller.
6. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter according to claim 1, special
Sign is: in the step 5-2, the fuzzy interval before fuzziness is maximum fuzzy tolerance section [low_Fi,high_Fi], i
∈ [1, N], when section definition is thinner, the range is smaller.
7. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter according to claim 1, special
Sign is: in the step 3-1, image recovery method is that denoising is filtered from coding and/or edge self-adaption.
8. a kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter according to claim 1, special
Sign is: in the step 4-1, image recovery method is super-resolution algorithms and/or deblurring algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910622166.XA CN110378271B (en) | 2019-09-05 | 2019-09-05 | Gait recognition equipment screening method based on quality dimension evaluation parameters |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910622166.XA CN110378271B (en) | 2019-09-05 | 2019-09-05 | Gait recognition equipment screening method based on quality dimension evaluation parameters |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110378271A true CN110378271A (en) | 2019-10-25 |
| CN110378271B CN110378271B (en) | 2023-01-03 |
Family
ID=68250909
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910622166.XA Active CN110378271B (en) | 2019-09-05 | 2019-09-05 | Gait recognition equipment screening method based on quality dimension evaluation parameters |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110378271B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111598132A (en) * | 2020-04-21 | 2020-08-28 | 烽火通信科技股份有限公司 | Portrait recognition algorithm performance evaluation method and device |
| CN116248842A (en) * | 2023-05-12 | 2023-06-09 | 北京城建智控科技股份有限公司 | Image encryption method based on camera device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104156919A (en) * | 2014-08-04 | 2014-11-19 | 陕西科技大学 | Motion-blurred image restoration method based on wavelet transform and Hopfield neural network |
| WO2015172235A1 (en) * | 2014-05-15 | 2015-11-19 | Tandemlaunch Technologies Inc. | Time-space methods and systems for the reduction of video noise |
| CN109509168A (en) * | 2018-08-30 | 2019-03-22 | 易诚博睿(南京)科技有限公司 | A kind of details automatic analysis method for picture quality objective evaluating dead leaf figure |
-
2019
- 2019-09-05 CN CN201910622166.XA patent/CN110378271B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2015172235A1 (en) * | 2014-05-15 | 2015-11-19 | Tandemlaunch Technologies Inc. | Time-space methods and systems for the reduction of video noise |
| CN104156919A (en) * | 2014-08-04 | 2014-11-19 | 陕西科技大学 | Motion-blurred image restoration method based on wavelet transform and Hopfield neural network |
| CN109509168A (en) * | 2018-08-30 | 2019-03-22 | 易诚博睿(南京)科技有限公司 | A kind of details automatic analysis method for picture quality objective evaluating dead leaf figure |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111598132A (en) * | 2020-04-21 | 2020-08-28 | 烽火通信科技股份有限公司 | Portrait recognition algorithm performance evaluation method and device |
| CN111598132B (en) * | 2020-04-21 | 2023-06-09 | 烽火通信科技股份有限公司 | Portrait recognition algorithm performance evaluation method and device |
| CN116248842A (en) * | 2023-05-12 | 2023-06-09 | 北京城建智控科技股份有限公司 | Image encryption method based on camera device |
| CN116248842B (en) * | 2023-05-12 | 2023-08-01 | 北京城建智控科技股份有限公司 | Image encryption method based on camera device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110378271B (en) | 2023-01-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106846316A (en) | A kind of GIS inside typical defect automatic distinguishing method for image | |
| CN109087510A (en) | traffic monitoring method and device | |
| CN110378324A (en) | A kind of face recognition algorithms appraisal procedure based on quality dimensions | |
| Su et al. | A novel forgery detection algorithm for video foreground removal | |
| CN111210399B (en) | Imaging quality evaluation method, device and equipment | |
| CN114926463A (en) | Production quality detection method suitable for chip circuit board | |
| CN106529545A (en) | Speckle image quality recognition method and system based on image feature description | |
| CN105812618B (en) | A kind of method for testing motion and motion detection apparatus | |
| CN110415208A (en) | A kind of adaptive targets detection method and its device, equipment, storage medium | |
| CN113658192B (en) | Multi-target pedestrian track acquisition method, system, device and medium | |
| CN110378271A (en) | A kind of Gait Recognition equipment screening technique based on quality dimensions assessment parameter | |
| CN115841488B (en) | PCB hole inspection method based on computer vision | |
| CN109492647A (en) | A kind of power grid robot barrier object recognition methods | |
| CN115015286B (en) | Chip detection method and system based on machine vision | |
| CN116912475B (en) | Display screen foreign matter detection method and device, electronic equipment and storage medium | |
| CN110378270A (en) | A kind of face recognition device screening technique based on quality dimensions assessment parameter | |
| CN105913008A (en) | Crowd exceptional event detection method based on hypothesis examination | |
| CN116977931A (en) | High-altitude parabolic identification method based on deep learning | |
| CN106709516B (en) | Fuzzy image detection method based on naive Bayes method | |
| CN113989626B (en) | Multi-class garbage scene distinguishing method based on target detection model | |
| CN109448012A (en) | A kind of method for detecting image edge and device | |
| CN114863337A (en) | Novel screen anti-photographing recognition method | |
| Miljanovic et al. | Detection of windows in facades using image processing algorithms | |
| CN113436120A (en) | Image fuzzy value identification method and device | |
| Gao et al. | Spatio-temporal salience based video quality assessment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A Method for Screening Gait Recognition Devices Based on Quality Dimension Evaluation Parameters Effective date of registration: 20230726 Granted publication date: 20230103 Pledgee: Dalian Branch of Shanghai Pudong Development Bank Co.,Ltd. Pledgor: YICHENG GAOKE (DALIAN) TECHNOLOGY Co.,Ltd. Registration number: Y2023980049989 |
|
| PE01 | Entry into force of the registration of the contract for pledge of patent right |