CN109784389A - A kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm - Google Patents
A kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm Download PDFInfo
- Publication number
- CN109784389A CN109784389A CN201910003873.0A CN201910003873A CN109784389A CN 109784389 A CN109784389 A CN 109784389A CN 201910003873 A CN201910003873 A CN 201910003873A CN 109784389 A CN109784389 A CN 109784389A
- Authority
- CN
- China
- Prior art keywords
- image
- sample
- adaboost
- coal
- classifier
- 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
- 239000003245 coal Substances 0.000 title claims abstract description 44
- 238000001514 detection method Methods 0.000 title claims description 13
- 239000011435 rock Substances 0.000 claims abstract description 22
- 238000000034 method Methods 0.000 claims abstract description 19
- 239000000428 dust Substances 0.000 claims abstract description 11
- 239000000843 powder Substances 0.000 claims description 7
- 238000004088 simulation Methods 0.000 claims description 7
- 239000006002 Pepper Substances 0.000 claims description 6
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 230000003455 independent Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 239000004744 fabric Substances 0.000 claims 1
- 238000005286 illumination Methods 0.000 abstract description 2
- 231100001261 hazardous Toxicity 0.000 abstract 1
- 238000005065 mining Methods 0.000 description 4
- 238000012952 Resampling Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000003066 decision tree Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The recognition methods of the invention discloses a kind of coal and rock based on Adaboost and Gabor algorithm, this method collects the image of coal and rock as original sample collection first, and the textural characteristics of sample are extracted with Gabor, then it is trained with Adaboost cascade classifier, then it is detected with the image that trained classifier detects needs, marks the region of the rock in image.Since the relatively hazardous image of coal working face is not easy to obtain, and illumination condition is severe in image, and dust is more, therefore carries out artificial synthesized data to collected original image, increases data set quantity to reach, improves classifier robustness and generalization ability.
Description
Technical field
The invention belongs to coal petrography identification field more particularly to a kind of coal petrography identifications based on Adaboost and Gabor algorithm
Detection method.
Background technique
The mining way in current China is strided forward to unmanned fully mechanized mining mode, and the unmanned operating of coalcutter, it is necessary to
Cutting could be accurately carried out by coal petrography identification, avoids the damage of coalcutter and the waste of coal.With based on machine learning
The fast development of image recognition technology, carrying out coal petrography knowledge method for distinguishing using image becomes feasible.It is compared to some other coal
Rock identification technology is for example: radar detection, infrared acquisition, sound-detection, vibration detecting etc. have the following advantage: 1, at low cost
It is honest and clean, it is easy to be promoted;2, image is visual, there is good comprehensibility;3, it can be imaged in existing fully-mechanized mining working
Upgraded on the basis of head, existing equipment can be efficiently used.Itself with regard to image recognition, coal petrography identification there is also
Image data acquisition is difficult, can not effectively be trained;Fully mechanized coal face illumination condition is severe;The problems such as fully mechanized coal face dust is big.Cause
This, needing one kind, efficiently robustness is good, the strong coal petrography recognition detection method of generalization ability, existing in the prior art to solve
Problems.
Summary of the invention
Goal of the invention: in view of the above problems, the present invention provides a kind of coal based on Adaboost and Gabor algorithm and rock
Recognition detection method extracts image texture by Gabor, and carries out the reliable fast of coal petrography type using Adaboost classifier
Speed identification.
Technical solution: to achieve the purpose of the present invention, the technical scheme adopted by the invention is that: one kind being based on Adaboost
With the coal petrography recognition detection method of Gabor algorithm, method includes the following steps:
(1) the total N0 of image for collecting coal and rock with class label, constitute raw data set X0;Wherein, the figure of coal
As being labeled as 1, the image tagged of rock is -1;
(2) influence for applying noise simulation downhole powder dust noise to image, becomes several for the data set X0 in step (1)
According to collection X;Wherein, there are coal and rock image N in X, and be predetermined size size by image normalization;
(3) textural characteristics of coal and rock image after being extracted plus made an uproar with Gabor algorithm;
(4) textural characteristics extracted in step (3) are input in adaboost classifier and are trained;
(5) image to be detected is detected with trained adaboost classifier, exports coal petrography recognition result.
Further, in step (2), apply the influence of noise simulation downhole powder dust noise to image, it will be in step (1)
Data set X0 become data set X, the method is as follows: different degrees of salt-pepper noise simulation downhole powder dust is added to original image and is made an uproar
The influence of sound forms new image data set, salt-pepper noise specifically:
(2.1) m Signal to Noise Ratio (SNR) is preset;
(2.2) the total number-of-pixels Q for calculating original image, brings first Signal to Noise Ratio (SNR) into following formula and obtains adding
The number of pixels made an uproar:
P=Q* (1-SNR);
(2.3) it obtains the position that add the P pixel made an uproar at random in original image, specifies identified P position picture
Element value is 0;
(2.4) it successively brings the 2-m Signal to Noise Ratio (SNR) into above-mentioned formula, repeats step (2.2)-(2.3), obtain original
The image of the corresponding m plus noise of image;
(2.5) it repeats step (2.2)-(2.4) all pictures are carried out plus made an uproar, the later image of output plus noise.
Further, in step (3), the textural characteristics of coal and rock image after being extracted plus made an uproar with Gabor algorithm, method
It is as follows:
(3.1) 5 scales are set as to Gabor algorithm parameter, 6 directions totally 30 filters, wherein gabor is filtered
The form of device are as follows:
Wherein x, y are two independents variable of two-dimensional function, x '=xcos θ+ysin θ, y '=- xsin θ+ycos θ, scale
Parameter lambda takes 2,3,4,5,6, directioin parameter θ to take 0 °, and 60 °, 120 °, 180 °, 240 °, 300 °, ψ is phase offset, and σ is Gaussian function
Number standard deviation, γ is length-width ratio;
(3.2) image after each plus noise obtains 30 texture images through 30 filters, to this 30 texture images
Seek gray average respectively, 30 gray averages of gained constitute the vector of a 30*1, the feature using the vector as original image to
Amount.
Further, in step (4), the textural characteristics extracted in step (3) are input in adaboost classifier
It is trained, the method is as follows:
(4.1) sample set the X={ (x being made of textural characteristics in step 3 is defined1,y1),(x2,y2),... (xi,
yi),...,(xN,yN), wherein N is the quantity of image in data set, and y is label y ∈ { -1,1 };
(4.2) weight of initialization sample is distributed as D1(x)=1/N, wherein D1It (x) is first Weak Classifier of training
Sample weights distribution function;
(4.3) remember adaboost Weak Classifier number be T, remember t=1,2 ..., T;
Training set t-th of Weak Classifier of training that (4.3.1) is made of N number of 30*1 feature vector and corresponding label y,
It is denoted as ht;
(4.3.2) calculates htError in classification εt=PX~Dt(ht(x) ≠ y (x)), i.e., it is weak under the distribution of current sample weights
Classification and image true tag inconsistent probability of the classifier to image, the sample for the training set that x refers to;
(4.3.3) should make the h in (4.3.2)tError in classification εt< 0.5, if εt>=0.5, then abandon current class
Device simultaneously re-starts training with resampling method;I.e. using sample weights as probability, training set is sampled.
(4.3.4) should make figure penalties function to the optimal effectiveness of image classification to obtain
It minimizes, whereinAnd it is 0 that function minimization, which corresponds to derivative, so figure penalties function should be made
It is 0 to weight derivative, i.e.,α is solved by above formulatObtain Weak Classifier htWeight αt
=0.5*In (1- εt)-0.5*In(εt);
(4.3.5) reduces the weight for correct sample of classifying in a upper circulation, and the weight for increasing classification error sample makes t+1
Data distribution under secondary circulation are as follows:Wherein, standardizing factor ZtFor vector
In the sum of each element;
(4.3.6) recycles (4.3.1) to (4.3.5) step until the full T Weak Classifier of training;
Wherein, PX~Dt() is that sample x obeys DtThe probability for meeting condition () under distribution, DtIt is the lower number of the t times circulation
According to distribution;Y (x) is true tag corresponding to sample x, htIt (x) is t-th of Weak Classifier to the output label of sample x;
T Weak Classifier group is combined into strong classifier by way of linear combination by (4.3.7)
The utility model has the advantages that compared with prior art, technical solution of the present invention has following advantageous effects:
1, the adaptability to coal working face dust atmosphere can be effectively improved;
2, quickly and accurately coal petrography identification can provide authentic data to the cutting of coalcutter, avoid cutting tooth because cutting rock
It is destroyed caused by stone, there is very big economic value.
3, low in cost, without complex device, and visual coal petrography identification interface is suitble to the use of front-line workers.
Detailed description of the invention
Fig. 1 is the basic flow chart of Coal-rock identification method of the present invention;
Fig. 2 is Adaboost algorithm flow chart of the present invention;
Fig. 3 is the gabor texture of certain rock specimens;
Fig. 4 is the gabor texture of certain coal sample.
Specific embodiment
Further description of the technical solution of the present invention with reference to the accompanying drawings and examples.
According to a kind of embodiment form, a kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm is provided,
The following steps are included:
Step 1, the total N0 of image for collecting coal and rock with class label, constitute raw data set X0;Wherein, coal
Image tagged be 1, the image tagged of rock is -1;
Step 2, the influence that downhole powder dust noise is simulated using synthetic data generation are become the data set X0 in step 1
For data set X;Wherein, there are coal and rock image N in X, and N is greater than N0, and the size for being 30*30 by image normalization,
The specific method is as follows:
The interference of different degrees of salt-pepper noise simulation fully-mechanized mining working dust is added to original image, forms new picture number
According to collection.Wherein, it is 10 times of raw data set or more that new data set, which includes the quantity of image,.
Wherein, salt-pepper noise specifically:
(2.1) m Signal to Noise Ratio (SNR) is preset;Such as signal-to-noise ratio can take 0.6 or 0.7 or 0.8 or 0.9;
(2.2) the total number-of-pixels Q for calculating original image, brings first Signal to Noise Ratio (SNR) into following formula and obtains adding
The number of pixels made an uproar:
P=Q* (1-SNR);
(2.3) it obtains the position that add the P pixel made an uproar at random in original image, specifies identified P position picture
Element value is 0;
(2.4) it successively brings the 2-m Signal to Noise Ratio (SNR) into above-mentioned formula, repeats step (2.2)-(2.3), obtain original
The image of the corresponding m plus noise of image;
(2.5) it repeats step (2.2)-(2.4) all pictures are carried out plus made an uproar, the later image of output plus noise.
Step 3 extracts the textural characteristics for adding coal and rock image after making an uproar with Gabor algorithm, and the specific method is as follows:
(3.1) 5 scales are set as to Gabor algorithm parameter, 6 directions totally 30 filters, wherein gabor filtering
The form of device are as follows:
Wherein, x, y are two independents variable of two-dimensional function;X '=xcos θ+ysin θ;Y '=- xsin θ+ycos θ;Scale
Parameter lambda takes 2,3,4,5,6;Directioin parameter θ takes 0 °, 60 °, 120 °, 180 °, 240 °, 300 °;Phase offset ψ takes 0 °;Gaussian function
Number standard deviation sigma=0.56 λ;Length-width ratio γ takes 0.5.
(3.2) each that the image after making an uproar is added to obtain 30 texture images through 30 filter filterings, to this 30 texture maps
As seeking gray average respectively, 30 gray averages of gained constitute the vector of a 30*1, using the vector as the feature of original image
Vector.
The texture feature vector extracted in step 3 is input in adaboost classifier and is trained by step 4, has
Body method is as follows:
(4.1) sample set the X={ (x being made of texture feature vector in step 3 is defined1,y1),(x2,
y2),......,(xN,yN), wherein N is the quantity of image in data set, and y is label y ∈ { -1,1 };
(4.2) weight of initialization sample is distributed as D1(x)=1/N, wherein D1It (x) is first Weak Classifier of training
Sample weights distribution function;
(4.3) the Weak Classifier number for remembering adaboost is T, to t=1,2 ..., T;
(4.3.1) is to training set one Weak Classifier of training being made of N number of 30*1 feature vector and corresponding label y
It is denoted as ht, Weak Classifier htCan be in C4.5 decision Tree algorithms, BP neural network, the common two classification sides in SVM SVM
Optional one in method;
(4.3.2) calculates htError in classification εt=PX~Dt(ht(x) ≠ y (x)), i.e., it is weak under the distribution of current sample weights
Classification and image true tag inconsistent probability of the classifier to image, the sample for the training set that x refers to;
(4.3.3) should make the h in (4.3.2)tError in classification εt< 0.5, if εt>=0.5, then abandon current class
Device simultaneously re-starts training with resampling method;
(4.3.4) should make figure penalties function to the optimal effectiveness of image classification to obtain
It minimizes, whereinAnd it is 0 that function minimization, which corresponds to derivative, so figure penalties function should be made
It is 0 to weight derivative, i.e.,α is solved by above formulatObtain Weak Classifier htWeight
αt=0.5*In (1- εt)-0.5*In(εt), the Weak Classifier that it enables classification accuracy rate high obtains in integrated classifier
Bigger franchise.
Whether (4.3.5) correctly adjusts the weight of sample according to last round of middle Weak Classifier to the classification of sample, makes score
The correct sample of class, by less concern, and makes the sample of classification error in the training of next round in next round training
More paid close attention to.Therefore the weight that reduce correct sample of classifying in a circulation, increases the weight of classification error sample
So that the data distribution under the t+1 times circulation are as follows:Wherein, standardizing factor ZtFor vectorIn the sum of each element, its effect is that the sum of weight for making all samples is 1.Recycle (4.3.1) extremely
(4.3.5) step is until the full T Weak Classifier of training.
Wherein, PX~Dt() is that sample x obeys DtThe probability for meeting condition () under distribution;DtIt is the lower number of the t times circulation
According to distribution;Y (x) is label corresponding to sample x;htIt (x) is t-th of Weak Classifier to the output label of sample x.
(4.3.6) obtains T Weak Classifier by T circulation, by T Weak Classifier group by way of linear combination
It is combined into strong classifier
Step 5. detects image to be detected with trained adaboost classifier, exports coal petrography recognition result.
Claims (4)
1. a kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm, which is characterized in that this method includes following
Step:
(1) the total N0 of image for collecting coal and rock with class label, constitute raw data set X0;Wherein, the image mark of coal
It is denoted as 1, the image tagged of rock is -1;
(2) influence for applying noise simulation downhole powder dust noise to image, becomes data set X for the data set X0 in step (1);
Wherein, there are coal and rock image N in X, and be predetermined size size by image normalization;
(3) texture feature vector of coal and rock image after being extracted plus made an uproar with Gabor algorithm;
(4) texture feature vector extracted in step (3) is input in adaboost classifier and is trained;
(5) image to be detected is detected with trained adaboost classifier, exports coal petrography recognition result.
2. a kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm according to claim 1, feature
It is, in step (2), applies the influence of noise simulation downhole powder dust noise to image, the data set X0 in step (1) is become
Data set X, the method is as follows: the influence of different degrees of salt-pepper noise simulation downhole powder dust noise is added to original image, is formed new
Image data set, salt-pepper noise specifically:
(2.1) m Signal to Noise Ratio (SNR) is preset;
(2.2) the total number-of-pixels Q that calculates original image brings first Signal to Noise Ratio (SNR) into following formula and obtains adding and makes an uproar
Number of pixels:
P=Q* (1-SNR);
(2.3) it obtains the position that add the P pixel made an uproar at random in original image, specifies identified P position pixel value
It is 0;
(2.4) it successively brings the 2-m Signal to Noise Ratio (SNR) into above-mentioned formula, repeats step (2.2)-(2.3), obtain original image
The image of corresponding m plus noises;
(2.5) it repeats step (2.2)-(2.4) all pictures are carried out plus made an uproar, the later image of output plus noise.
3. a kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm according to claim 2, feature
It is, in step (3), the texture feature vector of coal and rock image after being extracted plus made an uproar with Gabor algorithm, the method is as follows:
(3.1) 5 scales are set as to Gabor algorithm parameter, 6 directions totally 30 filters, the wherein shape of gabor filter
Formula are as follows:
Wherein, x, y are two independents variable of two-dimensional function, x '=xcos θ+ysin θ, y '=- xsin θ+ycos θ, scale parameter λ
2,3,4,5,6, directioin parameter θ is taken to take 0 °, 60 °, 120 °, 180 °, 240 °, 300 °, ψ is phase offset, and σ is Gaussian function standard
Difference, γ are length-width ratio;
(3.2) image after each plus noise obtains 30 texture images through 30 filter filterings, to this 30 texture images
Seek gray average respectively, 30 gray averages of gained constitute the vector of a 30*1, the feature using the vector as original image to
Amount.
4. a kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm according to claim 3, feature
It is, in step (4), the texture feature vector extracted in step (3) is input in adaboost classifier and is trained,
Method is as follows:
(4.1) sample set the X={ (x being made of texture feature vector in step (3) is defined1,y1),(x2,y2),...(xi,
yi),...,(xN,yN), wherein N is the quantity of image in data set, xiRefer to sample texture feature vector, yiFor label yi
∈{-1,1};
(4.2) weight of initialization sample is distributed as D1(x)=1/N, wherein D1It (x) is the sample of first Weak Classifier of training
Weight distribution function;
(4.3) remember adaboost Weak Classifier number be T, remember t=1,2 ..., T;
Training set t-th of Weak Classifier of training that (4.3.1) is made of N number of 30*1 feature vector and corresponding label y, is denoted as ht;
(4.3.2) calculates htError in classification εt=PX~Dt(ht(x) ≠ y (x)), i.e., the weak typing under the distribution of current sample weights
Classification and image true tag inconsistent probability of the device to image, the sample for the training set that x refers to;
(4.3.3) should make the h in (4.3.2)tError in classification εt< 0.5, if εt>=0.5, then it abandons current class device and lays equal stress on
Newly it is trained;
(4.3.4) should make figure penalties function to the optimal effectiveness of image classification to obtainIt is minimum
Change, whereinAnd it is 0 that function minimization, which corresponds to derivative, so figure penalties function should be made to power
Weight derivative is 0, i.e.,α is solved by above formulatObtain Weak Classifier htWeight,
αt=0.5*In (1- εt)-0.5*In(εt);
(4.3.5) reduces the weight for correct sample of classifying in a upper circulation, and the weight for increasing classification error sample makes the t+1 times
Data distribution under circulation are as follows:Wherein, standardizing factor ZtFor vector
In the sum of each element;
(4.3.6) circulation (4.3.1) is to (4.3.5) step until T Weak Classifier of training;
Wherein, PX~Dt() is that sample x obeys DtThe probability for meeting condition () under distribution, DtIt is the lower data point of the t times circulation
Cloth;Y (x) is true tag corresponding to sample x, htIt (x) is t-th of Weak Classifier to the output label of sample x;
T Weak Classifier group is combined into strong classifier by way of linear combination by (4.3.7)
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910003873.0A CN109784389A (en) | 2019-01-03 | 2019-01-03 | A kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910003873.0A CN109784389A (en) | 2019-01-03 | 2019-01-03 | A kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109784389A true CN109784389A (en) | 2019-05-21 |
Family
ID=66499800
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910003873.0A Pending CN109784389A (en) | 2019-01-03 | 2019-01-03 | A kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109784389A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113052208A (en) * | 2021-03-10 | 2021-06-29 | 神华神东煤炭集团有限责任公司 | Coal rock identification method based on vision, storage medium and electronic equipment |
CN113486817A (en) * | 2021-07-09 | 2021-10-08 | 中国矿业大学 | Coal face coal rock identification method based on laser scanning |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102521575A (en) * | 2011-12-16 | 2012-06-27 | 北京天诚盛业科技有限公司 | Iris identification method based on multidirectional Gabor and Adaboost |
CN104281838A (en) * | 2014-09-23 | 2015-01-14 | 西南交通大学 | Lug piece breakage detection method for high-speed rail overhead line system supporting device based on HOG features and two-dimensional Gabor wavelet transformation |
-
2019
- 2019-01-03 CN CN201910003873.0A patent/CN109784389A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102521575A (en) * | 2011-12-16 | 2012-06-27 | 北京天诚盛业科技有限公司 | Iris identification method based on multidirectional Gabor and Adaboost |
CN104281838A (en) * | 2014-09-23 | 2015-01-14 | 西南交通大学 | Lug piece breakage detection method for high-speed rail overhead line system supporting device based on HOG features and two-dimensional Gabor wavelet transformation |
Non-Patent Citations (2)
Title |
---|
WDMAD: "集成学习之Boosting—AdaBoost原理", 《知乎专栏HTTPS://ZHUANLAN.ZHIHU.COM/P/37358517?UTM_SOURCE=WECHAT_SESSION》 * |
郝清玉等: "基于图像多小波变换的煤岩界面识别", 《工矿自动化》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113052208A (en) * | 2021-03-10 | 2021-06-29 | 神华神东煤炭集团有限责任公司 | Coal rock identification method based on vision, storage medium and electronic equipment |
CN113052208B (en) * | 2021-03-10 | 2023-08-25 | 神华神东煤炭集团有限责任公司 | Vision-based coal rock identification method, storage medium and electronic equipment |
CN113486817A (en) * | 2021-07-09 | 2021-10-08 | 中国矿业大学 | Coal face coal rock identification method based on laser scanning |
CN113486817B (en) * | 2021-07-09 | 2023-09-08 | 中国矿业大学 | Coal face coal rock identification method based on laser scanning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106971152B (en) | Method for detecting bird nest in power transmission line based on aerial images | |
Zhao et al. | Cloud shape classification system based on multi-channel cnn and improved fdm | |
CN109522853B (en) | Face datection and searching method towards monitor video | |
CN108520226B (en) | Pedestrian re-identification method based on body decomposition and significance detection | |
CN110929593B (en) | Real-time significance pedestrian detection method based on detail discrimination | |
Ke et al. | A review of methods for automatic individual tree-crown detection and delineation from passive remote sensing | |
CN110837768B (en) | Online detection and identification method for rare animal protection | |
CN109829467A (en) | Image labeling method, electronic device and non-transient computer-readable storage medium | |
CN102609720B (en) | Pedestrian detection method based on position correction model | |
CN104166841A (en) | Rapid detection identification method for specified pedestrian or vehicle in video monitoring network | |
CN108549864A (en) | Area-of-interest filter method towards vehicle-mounted thermal imaging pedestrian detection and device | |
CN107808375B (en) | Merge the rice disease image detecting method of a variety of context deep learning models | |
TWI687159B (en) | Fry counting system and fry counting method | |
CN102509085A (en) | Pig walking posture identification system and method based on outline invariant moment features | |
CN104517104A (en) | Face recognition method and face recognition system based on monitoring scene | |
CN103208008A (en) | Fast adaptation method for traffic video monitoring target detection based on machine vision | |
Tamilselvi et al. | Unsupervised machine learning for clustering the infected leaves based on the leaf-colours | |
CN105447859A (en) | Field wheat aphid counting method | |
CN102542244A (en) | Face detection method and system and computer program product | |
CN104036250A (en) | Video pedestrian detecting and tracking method | |
CN107480585A (en) | Object detection method based on DPM algorithms | |
CN104239886A (en) | Image analysis based lawn and background boundary extraction method | |
CN109784389A (en) | A kind of coal petrography recognition detection method based on Adaboost and Gabor algorithm | |
CN108509861A (en) | A kind of method for tracking target and device combined based on sample learning and target detection | |
CN102592140A (en) | Method for monitoring condition of cyanobacteria bloom |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190521 |
|
RJ01 | Rejection of invention patent application after publication |