CN113344007B - Log fingerprint extraction and matching method - Google Patents

Log fingerprint extraction and matching method Download PDF

Info

Publication number
CN113344007B
CN113344007B CN202110612617.9A CN202110612617A CN113344007B CN 113344007 B CN113344007 B CN 113344007B CN 202110612617 A CN202110612617 A CN 202110612617A CN 113344007 B CN113344007 B CN 113344007B
Authority
CN
China
Prior art keywords
image
log
images
gray
sequence
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.)
Active
Application number
CN202110612617.9A
Other languages
Chinese (zh)
Other versions
CN113344007A (en
Inventor
张兰
唐晨
李向阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Science and Technology of China USTC
Original Assignee
University of Science and Technology of China USTC
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by University of Science and Technology of China USTC filed Critical University of Science and Technology of China USTC
Priority to CN202110612617.9A priority Critical patent/CN113344007B/en
Publication of CN113344007A publication Critical patent/CN113344007A/en
Application granted granted Critical
Publication of CN113344007B publication Critical patent/CN113344007B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures

Landscapes

  • Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Image Analysis (AREA)

Abstract

The invention discloses a log fingerprint extraction and matching method, which comprises the following steps: 1. preprocessing a log image: removing the log image background and converting the log image background into a gray scale space, 2, extracting log image features: calculating and counting the LBP value of each non-boundary point on the image, and taking the statistic as the log image characteristic, 3. log characteristic matching: and sampling the two log images to the same dimension, correcting the rotation angle, and calculating the similarity of the two log images in a multi-scale manner. The invention can extract the log fingerprint by a log fingerprint extraction technology, and can judge whether the log in the image comes from the same log or not based on the extracted fingerprint for any two log images, thereby avoiding the fraudulent behavior of online transaction and providing an effective solution for the online transaction of the log.

Description

Log fingerprint extraction and matching method
Technical Field
The invention belongs to the field of data fingerprint and entity transaction application, and particularly relates to a log fingerprint extraction and matching method.
Background
At present, the situation of few exits, no provinces and no state of departure makes the transaction under the line and across the line become extremely difficult. The log, as a representative of a large number of commodities in the traditional offline transaction, is influenced by the current situation, so that the sales volume of the whole log sales industry falls violently. The traditional log transaction process comprises the following steps: a customer selects logs in a log warehouse; paying a deposit and signing a contract after selecting proper logs; log suppliers send logs to sellers by logistics. In the present situation, the customer can not go to the log warehouse to pick up the logs, which is the main reason of the falling of the log sales. The log transaction on line is not like other commodities, and the transaction on line can not be carried out in late time, which mainly includes the following reasons: goods are susceptible to fraud, such as: and the package is dropped by the seller or the logistics party. Moreover, as one log is extremely large, a buyer cannot judge whether the log is the same log according to the photos; the freight cost is extremely high and even exceeds the price of the log, so that the reasonable return of goods in seven days similar to online shopping is not practical, and the buyer does not trust the online selling mode; the identification technology aiming at the logs is not specially used, the technologies such as face identification and fingerprint identification are very mature, and a person or an object can be uniquely identified through the extracted features at any time and any place. The existing main methods for log identification are: painting methods, bar codes, two-dimensional codes, microwave sensors, and radio frequency identification. The painting method, which is the most frequently used identification method for log transaction at present, marks the logs by painting numbers on the logs, but the marks are not unique, a seller can perform fraud by painting the same marks on another log, and the marks of the logs can be changed and removed during transportation. Bar codes and two-dimensional codes are only suitable for the finished product sale stage after log processing, although microwave sensors and radio frequency identification can identify the geometric shapes and internal structures of different logs, the price of required instruments is very high, the volume of the instruments is large, and the instruments are inconvenient to carry, so that the normalization of the instruments for log identification is not practical.
Disclosure of Invention
The invention aims to solve the defects of the prior art and provides a log fingerprint extraction and matching method, so that the log fingerprint can be extracted by a log fingerprint extraction technology, and for any two log images, whether logs in the images are from the same log can be judged based on the extracted fingerprints, so that the online transaction fraud behavior is avoided, and an effective solution is provided for the online transaction of the logs.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention relates to a log fingerprint extraction and matching method which is characterized by comprising the following steps:
step 1: preprocessing a log image;
step 1.1: background removal:
step 1.1.1: acquiring a raw wood image G and converting the raw wood image G from an RGB color space to an HSV color space, thereby obtaining color characteristics of the raw wood image G, comprising: a hue H component, a saturation S component, and a brightness V component;
step 1.1.2: setting a threshold value of a hue H component to TH based on color characteristics of the raw wood image G1And the threshold value of the saturation S component is TH2；
Step 1.1.3: according to the threshold TH1And TH2Judging the area of each pixel point of the log image G:
if the hue H component of any pixel point in the log image G is smaller than the threshold TH1And the saturation S component is greater than the threshold TH2If so, judging that the corresponding pixel points are the pixel points of the log area, and keeping the pixel values of the pixel points of the log area unchanged; otherwise, judging that the corresponding pixel point is the pixel point of the background area, and making the pixel value of the pixel point of the background area be '0'; thereby completing the background removal processing of the log image G and obtaining the log image G' with the background removed;
step 1.2: image cutting:
performing morphological treatment and smoothing treatment on the log image G' with the background removed, cutting a log area, and performing scale normalization treatment on the cut log area to obtain a log area image Gnew；
Step 1.3: gray level processing:
the log area image GnewConversion from RGB color space to gray scale space to obtain gray scale image Ggray；
And 2, step: extracting image features;
step 2.1: LBP sequence extraction:
recording a grayscale image GgrayUpper arbitrary ith non-boundary point xiGray value gray ofi；
Clockwise recording with ith non-boundary point xiCircle with radius r as center of circle and gray scale image GgrayThe ith pixel point sequence s obtained after the intersectioni＝[s1i,s2i,s3i,…,sni]Wherein s isniRepresenting the nth intersection point in the ith pixel point sequence;
the ith pixel point sequence siEach cross of (2)Point average and grayiMake a comparison, and will be greater than the grayiThe pixel value of the intersection point is set to be 1 and is less than or equal to grayiThe pixel value of the intersection point is set to be 0, so that a 0-1 sequence b with the length of n is obtainediCalculating the 0-1 sequence biIs a rotation invariant LBP sequence bi’；
Will rotate the invariant LBP sequence bi' converting into 10-system value sequence and then normalizing into an interval [0,256 ], thereby obtaining the ith non-boundary point xiLBP value of (a);
step 2.2: counting frequency:
statistical grayscale image GgrayObtaining LBP values of all non-boundary points, thereby obtaining a statistic gamma about distribution frequency of different LBP values, and using the statistic gamma as a gray level image characteristic;
and 3, step 3: matching image features;
step 3.1: preprocessing of feature matching:
for two log images I to be matchedaAnd IbTwo raw wood images IaAnd IbIs resampled to (min (w)a,wb),min(ha,hb) To obtain two sample images I of the same sizea' and Ib'; wherein, wa,wbRespectively two log images IaAnd IbWidth of (h)a,hbRespectively two raw wood images IaAnd IbThe height of (a); if the resampling is downsampling, downsampling by a local area averaging method; if the resampling is upsampling, performing upsampling by a bilinear interpolation method;
step 3.2: and (3) similarity calculation:
extracting two sampling images I according to the processes of step 1 and step 2a' and Ib' respective characteristics gammaaAnd gammabAnd calculating two features gammaaAnd gammabThe Wasserstein distance of (1), thereby taking the Wasserstein distance as two sampling images Ia' and Ib' similarity;
step 3.3: correcting a rotation angle:
fixed sample image Ia', image I is to be sampledb' rotate alpha degrees each time and rotate m times in total to obtain a rotating image sequence { I1α’,I2α’,I3α’,…,Imα' }, calculating sampling images I respectivelya' similarity with each rotated image in the sequence of rotated images, and taking the image I with the minimum similaritymin' as a sampled image I2' rotation angle correction image; wherein α represents an interval angle;
step 3.4: multi-scale similarity comparison:
sampling image Ia' and Imin' in accordance with {1: k1,1:k2,1:k3,…,1:ktDown-sampling at the ratio of (k) to obtain t sets of images, and then processing the images according to the ratio of (k)1:1,k2:1,k3:1,…,k t1 to obtain other t groups of images, respectively extracting the characteristics of each group of images and then carrying out similarity calculation, thereby obtaining a sampling image I under different scalesa' and Imin' similarity between them;
step 3.5: weighted average is carried out on all the calculated similarity to be used as a log image IaAnd IbThe actual similarity sim of (2); the actual similarity sim and the two set threshold values Th3、Th4Carrying out comparison; if sim<Th3Then the raw wood image I is consideredaAnd IbFrom the same log; if sim>TH4Then the raw wood image I is consideredaAnd IbFrom different logs; otherwise, it means that the log image I cannot be judgedaAnd IbWhether from the same log.
Compared with the prior art, the invention has the beneficial effects that:
1. the traditional log selling industry can only carry out offline transaction due to high transportation cost and easy fraudulent behavior of online transaction, and the existing technology has no technical means specially aiming at logs. The invention provides a log characteristic extraction technology aiming at logs specially, and the log characteristics are used as the fingerprints of the logs, so that the fraud behavior of online transaction is prevented, and a rotation angle correction and multi-scale matching algorithm are designed to enhance the robustness of the log fingerprints, so that the method is simple and effective, and a solution is provided for online log transaction.
2. In the invention, step 3.2, the similarity among the characteristic distributions of the log images is calculated and extracted by adopting the Wasserstein distance, and the difference among the log images is effectively measured, so that whether the images come from the same log can be accurately determined.
3. In the invention, in the steps 3.1, 3.3 and 3.4, the image is resampled first, and the consistency of the pixel number of the matched image is ensured when the matched image is matched; then, the rotation angle of the image is corrected, and the inconsistency of deflection angles caused by the rotation of the logs before and after transportation is further eliminated; and finally, performing multi-scale feature matching on the log image, and matching log features in different scales, thereby improving the robustness of feature matching.
Drawings
FIG. 1 is an RGB color image of the present invention with respect to a log;
FIG. 2 is an image of the present invention with the background of the log removed;
FIG. 3 is an exemplary illustration of an LBP sequence for obtaining pixel values according to the present invention;
FIG. 4 is an exemplary diagram of the calculation of pixel value rotation invariant LBP sequence values in accordance with the present invention;
FIG. 5 is a log map obtained by replacing the LBP values with pixel values according to the present invention;
FIG. 6 is a graph of the results of an experiment performed when calculating the threshold value according to the present invention.
Detailed Description
In this embodiment, a log fingerprint extraction and matching method includes the following steps:
step 1: preprocessing a log image;
step 1.1: background removal:
step 1.1.1: the raw wood image G is acquired as shown in fig. 1, and converted from the RGB color space to the HSV space, thereby obtaining the color characteristics of the raw wood image G, including: a hue H component, a saturation S component, and a brightness V component;
step 1.1.2: setting a threshold value of the hue H component to TH based on the color characteristics of the raw wood image G10.1 and a saturation S component of TH2＝0.2；
Step 1.1.3: according to a threshold value TH1And TH2Judging the area of each pixel point of the log image G:
if the hue H component of any pixel point in the log image G is less than the threshold TH1And the saturation S component is greater than the threshold TH2That is, a single pixel in the image simultaneously satisfies TH1<0.1 and TH2>When 0.2, judging that the corresponding pixel point is the pixel point of the log area, and keeping the pixel value of the pixel point of the log area unchanged; otherwise, judging that the corresponding pixel point is the pixel point of the background area, and making the pixel value of the pixel point of the background area be 0; thereby completing the background removal processing of the log image G and obtaining the log image G' with the background removed;
step 1.2: image cutting:
performing morphological processing and smoothing processing on the log image G' with the background removed, cutting out a log area, and performing scale normalization processing on the cut log area to obtain a log area image GnewAs shown in fig. 2, this is done to prevent the background area from interfering with the extraction and matching of log features and ensure the stability of the matching result. There are many other methods for background removal, which can also be used in the background removal step of the method;
step 1.3: gray level processing:
the log area image GnewConverting from RGB color space to grayscale space to obtain grayscale image Ggray；
And 2, step: extracting image features;
step 2.1: LBP sequence extraction:
recording a grayscale image GgrayUpper arbitrary ith non-boundary point xiGray value gray ofi；
Clockwise recording with the ith non-boundary point xiCircle with radius r as center of circle and gray scale image GgrayThe ith pixel point sequence s obtained after intersectioni＝[s1i,s2i,s3i,…,sni]Wherein s isniRepresenting the nth intersection point in the ith pixel point sequence; in this embodiment, r is 1 and intersects with the grayscale image to obtain a pixel point sequence with a length of 8.
The ith pixel point sequence siEach intersection of (a) with a grayiMake a comparison and will be greater than grayiThe pixel value of the intersection point is set to be 1 and is less than or equal to grayiThe pixel value of the intersection point is set to be 0, so that a 0-1 sequence b with the length of n is obtainediCalculating the rotation invariant LBP (local Binary Pattern) sequence, which is denoted as bi’；
Will rotate the invariant LBP sequence bi' converting into 10-system value sequence and then normalizing into an interval [0,256 ], thereby obtaining the ith non-boundary point xiThe LBP value of (c); for example, a circle with the pixel point as the center and 1 as the radius intersects with the image, the circle is set to be "0" or "1" according to the relative size of the pixel point and the intersected pixel point, as shown in fig. 3, and the LBP sequence is calculated to be 11100001. the sequence is rotated, as the length of the sequence is 8, all the cases can be traversed by rotating 7 times, and the minimum binary sequence in the 8 cases, namely 00001111 is the LBP sequence with unchanged rotation, as shown in fig. 4, is converted into a decimal value of 15 as the LBP value.
The pixel size of the whole log image is converted into its corresponding LBP value, and the result is shown in fig. 5. It can be seen that the features of the log, such as the grain, are very clear, which also illustrates the rationality of the method for extracting the features of the log.
Step 2.2: counting frequency:
statistical grayscale image GgrayObtaining LBP values of all non-boundary points so as to obtain a statistic gamma about the distribution frequency of different LBP values, and using the statistic gamma as a gray level image characteristic;
and 3, step 3: matching image features;
step 3.1: preprocessing of feature matching:
for two log images I to be matchedaAnd IbTwo raw wood images IaAnd IbIs resampled to (min (w)a,wb),min(ha,hb) To obtain two sample images I of the same sizea' and Ib'; wherein, wa,wbRespectively two raw wood images IaAnd IbWidth of (h)a,hbRespectively two log images IaAnd IbThe height of (a); if the resampling is downsampling, downsampling by a local area averaging method; if the resampling is upsampling, upsampling by a bilinear interpolation method;
step 3.2: and (3) calculating the similarity:
extracting two sampling images I according to the processes of step 1 and step 2a' and Ib' respective characteristics gammaaAnd gammabAnd calculating two features gammaaAnd gammabThe Wasserstein distance of (1), thereby taking the Wasserstein distance as two sampling images Ia' and Ib' similarity. The Wasserstein distance is the most suitable function for measuring the similarity of two distributions, is integrated into the Scipy library of Python, and can be directly called.
Step 3.3: correcting a rotation angle:
fixed sample image Ia', image I is to be sampledb' rotate alpha degrees each time and rotate m times in total to obtain a rotating image sequence { I1α’,I2α’,I3α’,…,Imα' } calculating the sampling images I respectivelya' similarity with each rotated image in the sequence of rotated images, and taking the image I with the minimum similaritymin' as a sampled image I2' rotation angle correction image; where α represents the spacing angle. In this embodiment, α is 2 and m is 11, because for a sequence of length 8, the rotation invariant LBP sequence can only guarantee 22.5 ° of invariance, while the experiment proves that the rotation angles of different images of the same log differ by 1 °, the similarity is very high, and thereforeThe angle difference value between one image and the matched image is within 1 degree by rotating the image sequence obtained by 11 times, so that the rotation angle can be ensured to be within 1 degree, and the rotation angle is corrected;
step 3.4: multi-scale similarity comparison:
sampling image Ia' and Imin' in accordance with {1: k1,1:k2,1:k3,…,1:ktDown-sampling at the ratio of (k) to obtain t sets of images, and then processing the images according to the ratio of (k)1:1,k2:1,k3:1,…,kt1} to obtain t further groups of images, in practice t is taken as 3, k1,k2,k3Respectively taking 2, 3 and 4 to expect that the local detail features are matched in a low scale and the global contour features are matched in a high scale through the change of the scale, so that the matching accuracy is improved, the features of all groups of images are respectively extracted and then similarity calculation is carried out, and then the sampling images I under different scales are obtaineda' and Imin' similarity between them;
step 3.5: weighted average is carried out on all the calculated similarity, and the weighted average is used as a log image IaAnd IbActual similarity sim of (c); the actual similarity sim and the two set threshold values Th3、Th4In a specific implementation, as shown in fig. 6, a total of 150 sets of images from the same log are calculated, denoted by · and a total of 300 sets of images from different logs are calculated, denoted by ×, and two horizontal lines are drawn at discrete (i.e., image ordinate) of 600 and 1000, respectively. It can be seen that when the disparity is below 600, the image pairs are all from the same log, and when the disparity is above 1000, the image pairs are all from different logs, so TH is taken3＝600，TH41000; if sim<Th3I.e. sim<600, the log image I is consideredaAnd IbFrom the same log; if sim>TH4I.e. sim>1000, then consider the log image IaAnd IbFrom different logs; otherwise, it means that the log image I cannot be judgedaAnd IbWhether or not toFrom the same log.

Claims (1)

1. A log fingerprint extraction and matching method is characterized by comprising the following steps:
step 1: preprocessing a log image;
step 1.1: background removal:
step 1.1.1: acquiring a raw wood image G and converting the raw wood image G from an RGB color space to an HSV space, thereby obtaining color characteristics of the raw wood image G, comprising: a hue H component, a saturation S component, and a brightness V component;
step 1.1.2: setting a threshold value of a hue H component to TH based on color characteristics of the raw wood image G1And the threshold value of the saturation S component is TH2；
Step 1.1.3: according to the threshold TH1And TH2Judging the area of each pixel point of the log image G:
if the hue H component of any pixel point in the log image G is less than the threshold TH1And the saturation S component is greater than the threshold TH2If so, judging that the corresponding pixel points are the pixel points of the log area, and keeping the pixel values of the pixel points of the log area unchanged; otherwise, judging that the corresponding pixel point is the pixel point of the background area, and making the pixel value of the pixel point of the background area be '0'; thereby completing the background removal processing of the log image G and obtaining the log image G' with the background removed;
step 1.2: image cutting:
performing morphological treatment and smoothing treatment on the log image G' with the background removed, cutting a log area, and performing scale normalization treatment on the cut log area to obtain a log area image Gnew；
Step 1.3: gray level processing:
the log area image GnewConverting from RGB color space to grayscale space to obtain grayscale image Ggray；
Step 2: extracting image features;
step 2.1: LBP sequence extraction:
recording a grayscale image GgrayUpper arbitrary ith non-boundary point xiGray value gray ofi；
Clockwise recording with ith non-boundary point xiCircle with radius r as center of circle and gray scale image GgrayThe ith pixel point sequence s obtained after the intersectioni＝[s1i,s2i,s3i,…,sni]Wherein s isniRepresenting the nth intersection point in the ith pixel point sequence;
the ith pixel point sequence siEach intersection of (2) with a grayiMake a comparison and will be greater than grayiThe pixel value of the intersection point is set to be 1 and is less than or equal to grayiThe pixel value of the intersection point is set to be 0, so that a 0-1 sequence b with the length of n is obtainediCalculating the 0-1 sequence biIs a rotation invariant LBP sequence bi’；
Will rotate the invariant LBP sequence bi' converting into 10-system value sequence and then normalizing into an interval [0,256), thereby obtaining the ith non-boundary point xiLBP value of (a);
step 2.2: counting frequency:
statistical grayscale image GgrayObtaining LBP values of all non-boundary points so as to obtain a statistic gamma about the distribution frequency of different LBP values, and using the statistic gamma as a gray level image characteristic;
and 3, step 3: matching image features;
step 3.1: preprocessing of feature matching:
for two raw wood images I to be matchedaAnd IbTwo raw wood images IaAnd IbThe size of (c) is resampled to (min (w)a,wb),min(ha,hb) To obtain two sample images I of the same sizea' and Ib'; wherein, wa,wbRespectively two raw wood images IaAnd IbWidth of (h), ha,hbRespectively two raw wood images IaAnd IbThe height of (d); if the resampling is downsampling, downsampling by a local area averaging method; if it isIf the resampling is upsampling, the upsampling is carried out by a bilinear interpolation method;
step 3.2: and (3) similarity calculation:
extracting two sampling images I according to the processes of step 1 and step 2a' and Ib' respective characteristics gammaaAnd gammabAnd calculating two features gammaaAnd gammabThe Wasserstein distance of (a), thereby taking the Wasserstein distance as two sampling images Ia' and Ib' similarity;
step 3.3: correcting a rotation angle:
fixed sample image Ia', image I to be sampledb' rotate alpha deg. each time, rotate m times altogether, and obtain a sequence of rotated images { I }1α’,I2α’,I3α’,…,Imα' } calculating the sampling images I respectivelya', similarity with each rotated image in the sequence of rotated images, and taking the image I with the smallest similaritymin' as a sampled image I2' rotation angle correction image; wherein α represents an interval angle;
step 3.4: and (3) multi-scale similarity comparison:
sampling image Ia' and Imin' in accordance with {1: k1,1:k2,1:k3,…,1:ktGet t groups of images by down sampling according to the proportion of { k }1:1,k2:1,k3:1,…,kt1 to obtain other t groups of images, respectively extracting the characteristics of each group of images, and then carrying out similarity calculation, thereby obtaining sampling images I under different scalesa' and Imin' similarity between them;
step 3.5: weighted average is carried out on all the calculated similarity, and the weighted average is used as a log image IaAnd IbActual similarity sim of (c); the actual similarity sim is compared with two set thresholds Th3、Th4Comparing; if sim<Th3Then the raw wood image I is consideredaAnd IbFrom the same log; if sim>TH4Then the raw wood image I is consideredaAnd IbFrom different logs; otherwise, it means that the log image I cannot be judgedaAnd IbWhether from the same log.
CN202110612617.9A 2021-06-02 2021-06-02 Log fingerprint extraction and matching method Active CN113344007B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110612617.9A CN113344007B (en) 2021-06-02 2021-06-02 Log fingerprint extraction and matching method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110612617.9A CN113344007B (en) 2021-06-02 2021-06-02 Log fingerprint extraction and matching method

Publications (2)

Publication Number Publication Date
CN113344007A CN113344007A (en) 2021-09-03
CN113344007B true CN113344007B (en) 2022-07-15

Family

ID=77472991

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110612617.9A Active CN113344007B (en) 2021-06-02 2021-06-02 Log fingerprint extraction and matching method

Country Status (1)

Country Link
CN (1) CN113344007B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115464557A (en) * 2022-08-15 2022-12-13 深圳航天科技创新研究院 Method for adjusting mobile robot operation based on path and mobile robot

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9973521B2 (en) * 2015-12-28 2018-05-15 International Business Machines Corporation System and method for field extraction of data contained within a log stream
US10262187B1 (en) * 2016-11-10 2019-04-16 Synaptics Incorporated Systems and methods for spoof detection based on local binary patterns
CN111695498B (en) * 2020-06-10 2023-04-07 西南林业大学 Wood identity detection method

Also Published As

Publication number Publication date
CN113344007A (en) 2021-09-03

Similar Documents

Publication Publication Date Title
CN109299720B (en) Target identification method based on contour segment spatial relationship
CN107633192B (en) Bar code segmentation and reading method based on machine vision under complex background
CN108596197B (en) Seal matching method and device
CN107292310B (en) Visual positioning and automatic reading method for circular pointer type dial plate
CN110210477B (en) Digital instrument reading identification method
CN111476109A (en) Bill processing method, bill processing apparatus, and computer-readable storage medium
CN108960237B (en) Reading identification method for pointer type oil level indicator
CN114998323B (en) Deformed steel bar abnormity determination method based on attention mechanism
CN113344007B (en) Log fingerprint extraction and matching method
CN108932449B (en) Bar code binarization method and system based on clustering
CN111161260A (en) Hot-rolled strip steel surface defect detection method and device based on deep learning
CN115082472B (en) Quality detection method and system for hub mold casting molding product
CN117037132A (en) Ship water gauge reading detection and identification method based on machine vision
CN104463091B (en) A kind of facial image recognition method based on image LGBP feature subvectors
CN116503462A (en) Method and system for quickly extracting circle center of circular spot
CN109359653B (en) Cotton leaf adhesion lesion image segmentation method and system
CN112396580B (en) Method for detecting defects of round part
CN116309573B (en) Defect detection method for printed characters of milk packaging box
CN111292346B (en) Method for detecting contour of casting box body in noise environment
CN111652287A (en) Hand-drawing cross pentagon classification method for AD (analog-to-digital) scale based on convolution depth neural network
CN113591875B (en) High-precision pointer type instrument identification method
CN115690104A (en) Wafer crack detection method and device and storage medium
CN109118240A (en) Subway scene fee payment method
Yi et al. A clustering-based algorithm for automatic detection of automobile dashboard
CN114596271A (en) Method for extracting corrosion characteristics of high-light-reflection surface

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