CN105488528A - Improved adaptive genetic algorithm based neural network image classification method - Google Patents
Improved adaptive genetic algorithm based neural network image classification method Download PDFInfo
- Publication number
- CN105488528A CN105488528A CN201510846339.8A CN201510846339A CN105488528A CN 105488528 A CN105488528 A CN 105488528A CN 201510846339 A CN201510846339 A CN 201510846339A CN 105488528 A CN105488528 A CN 105488528A
- Authority
- CN
- China
- Prior art keywords
- population
- rbf neural
- training
- network
- neural network
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24133—Distances to prototypes
- G06F18/24137—Distances to cluster centroïds
- G06F18/2414—Smoothing the distance, e.g. radial basis function networks [RBFN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
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 an improved adaptive genetic algorithm based neural network image classification method. The method comprises the following steps: extracting texture features of sample images by adopting a gray-level co-occurrence matrix based feature extraction algorithm to obtain texture features of a training sample and a test sample; by taking the texture feature of the training sample as an input of an RBF neural network, training the RBF neural network with a genetic optimization based neural network learning method to generate a trained RBF neural network; and inputting the texture feature of the test sample into the trained RBF neural network, and performing an image classification test. For the deficiency that a k-means clustering algorithm and other algorithms are sensitive to initial value selection, the method can better avoid the ''prematurity'' convergence of a genetic algorithm, simplify the network structure of a neural network classifier and improve the generalization capability of the network and the correct classification efficiency of the images.
Description
Technical field
The present invention relates to a kind of image classification method, belong to intelligent computation, technical field of image processing, particularly a kind of neural network image sorting technique based on improving expert inquiry method.
Background technology
Genetic algorithm (GeneticAlgorithm, GA) be a kind of random search algorithm using for reference the highly-parallel that organic sphere natural selection and evolutionary mechanism grow up, provide a kind of general framework solving complicated optimum problem, do not rely on the specific field of problem, there is very strong robustness.Be widely used in nonlinear system to unify the problem such as optimizing of control program.There are some shortcomings in standard genetic algorithm: when solving some actual complex problems, be easy to occur " precocity " convergence, when solution in colony does not also reach optimum solution, become closely similar between individuality, lose evolvability, cause population very rapid convergence to locally optimal solution instead of globally optimal solution, the performance of algorithm and effect of optimization are greatly reduced.The main cause that " precocity " produces is the forfeiture of population diversity, and keeps the diversity of population to play vital effect for evolution, is the power of Evolution of Population.Population diversity is kept to be solve genetic algorithm " precocity " problem effective ways.
Mostly based on large-scale calculations, between their calculated amount and computational accuracy, there is serious contradiction in traditional Image Classfication Technology.In recent years, the nerual network technique of high speed development is a kind of new way addressed this problem.Radial basis function (RadialBasisFunction, RBF) a kind of locality of reacting to external world based on the neuronal cell of human brain of neural network and the feedforward neural network that proposes is a kind of typical in solving the neural network of pattern classification, identification.Compared with other feedforward neural networks, it has best approximation capability and global optimum's characteristic, and structure is simple, the features such as training speed is very fast.Therefore, RBF neural is very suitable for Images Classification, is the powerful solving Images Classification problem.For k-means clustering algorithm etc., responsive deficiency is chosen to initial value, the self-adapted genetic algorithm based on population diversity is used in the parameter optimization of RBF neural.
Summary of the invention
Object of the present invention is intended to solve above-mentioned technological deficiency.
For achieving the above object, the present invention proposes a kind of neural network image sorting technique based on genetic optimization, comprises the following steps:
S1: adopt the feature extraction algorithm based on gray level co-occurrence matrixes, extracts the textural characteristics of sample image, obtain training sample with the texture feature vector of test sample book;
S2: using the input of the textural characteristics of training sample as RBF neural, adopts the network learning method Training RBF Neural Network based on genetic optimization, generates the RBF neural trained; Training RBF Neural Network step is as follows:
S2.1: determine population scale, maximum iteration time correlation parameter, carries out real coding to the parameter of RBF neural, produces initial population at random;
S2.2: the variance and the entropy that calculate population, and population " precocity " level index, according to the fitness function of setting, calculates the fitness value of each individuality in population;
S2.3: use roulette selection strategy to carry out selection operation to population according to the fitness value of individuality;
S2.4: judge population variance and entropy, carries out cross and variation operation according to formula;
S2.5: after genetic manipulation, produces population of new generation, returns step 2.2 until reach the condition of convergence of setting.
S2.6: the structure optimizing RBF neural, obtains the RBF neural sorter after the training of optimized parameter and structure
S3: the RBF neural input of the textural characteristics of test sample book trained, carries out Images Classification test.
Beneficial effect
According to the neural network image sorting technique based on improvement Adaptive Genetic optimized algorithm of the embodiment of the present invention, in method, corresponding improvement strategy is proposed to the system of selection etc. of the cross and variation probability of self-adapted genetic algorithm.The diversity of population is the necessary condition having the whole feas ible space of efficient search, and population diversity gene internal diversity used is measured, and it shows the degree of convergence of population at this gene structure, is the key criteria of measure algorithm evolvability power.According to the diversity index of " evolution " population in period, Evolving State is divided four kinds of situation discussion, under different Evolving States, there is different cross and variation probability Adjusted Option, each Adjusted Option all adapts with population EA hardware now, better can instruct Evolution of Population.Emulation experiment shows, the algorithm idea of proposition, " precocity " of genetic algorithm can be avoided preferably to restrain, improve the efficiency of evolution of genetic algorithm.
For k-means clustering algorithm etc., responsive deficiency is chosen to initial value, self-adapted genetic algorithm based on population diversity is used in the parameter optimization of RBF neural, proposes a kind of RBF neural image classification method based on improving expert inquiry method.In the method, self-adapted genetic algorithm utilizes Measurement of Population Diversity index to improve standard genetic algorithm, overcomes " precocity " convergence problem; Then, adopt self-adapted genetic algorithm to optimize the parameter of RBF neural, overcome k-means clustering algorithm etc. and responsive deficiency is chosen to initial value; Finally, also the structure of RBF neural is also optimized.Emulation experiment shows that this algorithm not only can simplified network structure, and improves the generalization ability of network and the correct classification rate of image.
Accompanying drawing explanation
The present invention above-mentioned and/or additional aspect and advantage will become obvious and easy understand from the following description of the accompanying drawings of embodiments, wherein:
Fig. 1 is the process flow diagram of the neural network image sorting technique based on improving expert inquiry method of the embodiment of the present invention; And
Fig. 2 is the RBF neural training process schematic diagram of one embodiment of the invention; And
Fig. 3 is the unmanned plane Aerial Images schematic diagram to be sorted of one embodiment of the invention.
Embodiment
Be described below in detail embodiments of the invention, the example of described embodiment is shown in the drawings, and wherein same or similar label represents same or similar element or has element that is identical or similar functions from start to finish.Being exemplary below by the embodiment be described with reference to the drawings, only for explaining the present invention, and can not limitation of the present invention being interpreted as.
As shown in Figure 1, according to the neural network image sorting technique that the present invention is based on improving expert inquiry method, a few class unmanned plane Aerial Images is classified.
In order to solve the independent landing problem of unmanned plane, need to classify to UAV Landing point geomorphologic map picture, to identify the landforms of applicable unmanned plane independent landing.Choose this 4 class unmanned plane Aerial Images of sand ground, meadow, water and forest to test.This 4 class geomorphologic map picture is more representative, is the main study subject of example.Other landforms directly can not separate class in the method, and are directly classified as class of should not landing, and do not add concern, therefore without disaggregated classification, are so also to accelerate algorithm process speed.
The general more complicated of unmanned plane during flying environment, the image photographed also is that various atural object is interlaced, be difficult to obtain only containing the Aerial Images of single landforms, for obtaining qualified training test sample book, dividing processing need be carried out to the original image of unmanned plane shooting, thus obtain the unmanned plane Aerial Images storehouse only containing single landforms, should not choose at the mixed edge of area and classification of handing over of all kinds of atural object, to ensure that data have typicalness, thus can classify accurately.The image of the above-mentioned 4 class landforms after segmentation as shown in Figure 3.
Adopt the unmanned plane Aerial Images of above-mentioned 4 classes after cutting to carry out embodiments of the invention, concrete implementation step is as follows:
S1: image texture characteristic extracts:
Obtain unmanned plane Aerial Images to be sorted, each class image selects 50 width as training sample, chooses 30 width images in addition as test sample book.Adopt the feature extraction algorithm based on gray level co-occurrence matrixes, extract the textural characteristics of sample image, obtain 400 training sample texture feature vectors and 120 test sample book texture feature vectors, table 1 is the texture feature vector after four width example images extract.
The textural characteristics value of table 1 unmanned plane Aerial Images
S2: using the input of the texture feature vector of training sample as RBF neural, adopt the network learning method Training RBF Neural Network based on genetic optimization, generate the RBF neural sorter trained, Training RBF Neural Network step is as follows, here to input the proper vector of one group of (400) training sample image:
S2.1: determine population scale 100, maximum iteration time 2000, the input layer number of RBF neural is 4; Output layer nodes is set to 2, namely represents 4 class unmanned plane Aerial Images to be sorted respectively with 00 ~ 11, and 00 represents sand ground, 01 represents meadow, 10 and represent waters, 11 and represent forest; Initial node in hidden layer is set to 8.
Carry out real coding to the parameter of RBF neural, produce initial population at random, table 2 show in more detail individual coded format, is initially random number.
The coded format of table 2 individuality
Wherein c
ifor the center of i-th hidden layer node of RBF neural, i=1,2 ..., 8; σ
ifor the width of i-th hidden layer node of RBF neural; w
ijfor i-th hidden layer node of RBF neural is to the connection weights between a jth output layer node, j=1,2.
S2.2: the variance and the entropy that calculate population after last iteration, and population " precocity " level index, according to the fitness function of setting, calculates the fitness value of each individuality in population;
After neuralward network inputs one group of training sample vector, neural network has corresponding actual output
according to desired output with actual export the fitness value obtaining each individuality, get inverse by the square error of RBF neural and a very little constant sum and obtain.A kth individual fitness value is as follows:
Wherein c be greater than and close to zero constant, object is to prevent denominator from being 0, getting c=10 here
-4; The input amendment number of RBF neural is 400; The output layer nodes of RBF neural is 2;
for the desired output of RBF neural;
for the actual output of RBF neural.
S2.3: use roulette selection method and optimized individual conversation strategy to carry out selection operation to population according to the fitness value of all individualities;
In roulette wheel selection method, the probability that each individuality is selected and its fitness value are proportional, and optimized individual store method a kind ofly individuality the highest for fitness value in population is not carried out pairing and intersect, and directly copies to the system of selection in the next generation.
S2.4: judge population variance and entropy, carries out cross and variation operation according to formula;
Obtain population variance and entropy and population " precocity " level index by previous step, try to achieve according to following formula the cross and variation probability that population adapts to current Population status most:
Wherein, P
c1, P
c2be respectively minimum and maximum genetic algorithm crossover probability, get 0.9 and 0.6, P
m1, P
m2be respectively minimum and maximum genetic algorithm mutation probability, get 0.1 and 0.001; k
1, k
2be a coefficient for cross and variation rate change, value is between (0,1); D (t), S (t) are for when the population variance of former generation and entropy, and a, b be its judgment threshold, are used for the Evolving State of judgement population, and value is 7 and 2.5; G has represented that d is the threshold value of non-evolutionary generation, and value is 5 to the current algebraically on behalf of only not evolving continuously since last time evolves; Δ is the evaluation index of population " precocity " degree.
According to adaptive cross and variation probability, cross and variation operation is carried out to population at individual
S2.5: after genetic manipulation, produces population of new generation.
So far just complete the training process of one group of training sample, then iteration repeats above step, until reach end condition, end condition is set to optimum individual fitness value and is less than 10 here
-8.
S2.6: the structure optimizing RBF neural, obtains the RBF neural after the training of optimized parameter and structure.
Once the correction of the structure of RBF neural is carried out in training after stopping, and after completing iteration, if certain connection weights of satisfactory solution are approximately zero, (precision is set to 10 here
-6), so delete this connection; If the weights of certain hidden layer node are all approximately zero, so delete this node and parameter thereof, thus reduce node in hidden layer, the population that regeneration is new re-starts repetitive exercise, finally obtains optimum RBF neural structure.
Under it should be noted that the prerequisite that only performance of network does not reduce after structure optimization, could using the network after optimization as the network finally obtained; Otherwise, still using the network of the network before optimization as final acquisition.
S3: the RBF neural input of the textural characteristics of test sample book trained, carries out Images Classification test.
After the texture feature vector value of input sample to be tested image, RBF neural can obtain an Output rusults, when judging which kind of image Output rusults belong to, employing is Euclidean distance determining method, which kind of which is nearestly just judged to be for output vector and four object vectors (00,01,10,11), prerequisite is that distance needs to reach certain precision, if arbitrary distance is all greater than 10
-4then be judged to be other images.
Adopt and carry out test experiments based on the standard genetic optimization sorting technique of RBF neural and the sorting technique based on innovatory algorithm of proposition, concrete experimental result as shown in Table 3 and Table 4.As can be seen from the table, compared to the sorting technique that standard genetic is optimized, the sorting technique based on the self-adapted genetic algorithm improved can be classified to unmanned plane Aerial Images better, obtains the correct classification rate of more succinct network structure and Geng Gao.
The classification results of table 2 standard genetic optimization method
Table 3 classification results of the present invention
The classification results of table 4 two kinds of methods compares
Claims (2)
1., based on the neural network image sorting technique of improving expert inquiry method, it is characterized in that: the method comprises the following steps,
S1: adopt the feature extraction algorithm based on gray level co-occurrence matrixes, extracts the textural characteristics of sample image, obtain training sample with the texture feature vector of test sample book;
S2: using the input of the textural characteristics of training sample as RBF neural, adopts the network learning method Training RBF Neural Network based on genetic optimization, generates the RBF neural trained; Training RBF Neural Network step is as follows:
S2.1: determine population scale, maximum iteration time correlation parameter, carries out real coding to the parameter of RBF neural, produces initial population at random;
S2.2: the variance and the entropy that calculate population, and population " precocity " level index, according to the fitness function of setting, calculates the fitness value of each individuality in population;
S2.3: use roulette selection strategy to carry out selection operation to population according to the fitness value of individuality;
S2.4: judge population variance and entropy, carries out cross and variation operation according to formula;
S2.5: after genetic manipulation, produces population of new generation, returns step 2.2 until reach the condition of convergence of setting;
S2.6: the structure optimizing RBF neural, obtains the RBF neural sorter after the training of optimized parameter and structure
S3: the RBF neural input of the textural characteristics of test sample book trained, carries out Images Classification test.
2. the neural network image sorting technique based on improving expert inquiry method according to claim 1, is characterized in that: the method comprises the following steps,
Choose this 4 class unmanned plane Aerial Images of sand ground, meadow, water and forest to test; This 4 class geomorphologic map picture is more representative, is the main study subject of example; Other landforms directly can not separate class in the method, and are directly classified as class of should not landing, and do not add concern, therefore without disaggregated classification, are so also to accelerate algorithm process speed;
The general more complicated of unmanned plane during flying environment, the image photographed also is that various atural object is interlaced, be difficult to obtain only containing the Aerial Images of single landforms, for obtaining qualified training test sample book, dividing processing need be carried out to the original image of unmanned plane shooting, thus obtain the unmanned plane Aerial Images storehouse only containing single landforms, should not choose at the mixed edge of area and classification of handing over of all kinds of atural object, to ensure that data have typicalness, thus can classify accurately;
Adopt the unmanned plane Aerial Images of above-mentioned 4 classes after cutting to implement, concrete implementation step is as follows:
S1: image texture characteristic extracts:
Obtain unmanned plane Aerial Images to be sorted, each class image selects 50 width as training sample, chooses 30 width images in addition as test sample book; Adopt the feature extraction algorithm based on gray level co-occurrence matrixes, extract the textural characteristics of sample image, obtain 400 training sample texture feature vectors and 120 test sample book texture feature vectors, table 1 is the texture feature vector after four width example images extract;
The textural characteristics value of table 1 unmanned plane Aerial Images
S2: using the input of the texture feature vector of training sample as RBF neural, adopt the network learning method Training RBF Neural Network based on genetic optimization, generate the RBF neural sorter trained, Training RBF Neural Network step is as follows, here to input the proper vector of one group of (400) training sample image:
S2.1: determine population scale 100, maximum iteration time 2000, the input layer number of RBF neural is 4; Output layer nodes is set to 2, namely represents 4 class unmanned plane Aerial Images to be sorted respectively with 00 ~ 11, and 00 represents sand ground, 01 represents meadow, 10 and represent waters, 11 and represent forest; Initial node in hidden layer is set to 8;
Carry out real coding to the parameter of RBF neural, produce initial population at random, table 2 show in more detail individual coded format, is initially random number;
The coded format of table 2 individuality
Wherein c
ifor the center of i-th hidden layer node of RBF neural, i=1,2 ..., 8; σ
ifor the width of i-th hidden layer node of RBF neural; w
ijfor i-th hidden layer node of RBF neural is to the connection weights between a jth output layer node, j=1,2;
S2.2: the variance and the entropy that calculate population after last iteration, and population " precocity " level index, according to the fitness function of setting, calculates the fitness value of each individuality in population;
After neuralward network inputs one group of training sample vector, neural network has corresponding actual output
according to desired output with actual export the fitness value obtaining each individuality, get inverse by the square error of RBF neural and a very little constant sum and obtain; A kth individual fitness value is as follows:
Wherein c be greater than and close to zero constant, object is to prevent denominator from being 0, getting c=10 here
-4; The input amendment number of RBF neural is 400; The output layer nodes of RBF neural is 2;
for the desired output of RBF neural;
for the actual output of RBF neural;
S2.3: use roulette selection method and optimized individual conversation strategy to carry out selection operation to population according to the fitness value of all individualities;
In roulette wheel selection method, the probability that each individuality is selected and its fitness value are proportional, and optimized individual store method a kind ofly individuality the highest for fitness value in population is not carried out pairing and intersect, and directly copies to the system of selection in the next generation;
S2.4: judge population variance and entropy, carries out cross and variation operation according to formula;
Obtain population variance and entropy and population " precocity " level index by previous step, try to achieve according to following formula the cross and variation probability that population adapts to current Population status most:
Wherein, P
c1, P
c2be respectively minimum and maximum genetic algorithm crossover probability, get 0.9 and 0.6, P
m1, P
m2be respectively minimum and maximum genetic algorithm mutation probability, get 0.1 and 0.001; k
1, k
2be a coefficient for cross and variation rate change, value is between (0,1); D (t), S (t) are for when the population variance of former generation and entropy, and a, b be its judgment threshold, are used for the Evolving State of judgement population, and value is 7 and 2.5; G has represented that d is the threshold value of non-evolutionary generation, and value is 5 to the current algebraically on behalf of only not evolving continuously since last time evolves; Δ is the evaluation index of population " precocity " degree;
According to adaptive cross and variation probability, cross and variation operation is carried out to population at individual
S2.5: after genetic manipulation, produces population of new generation;
So far just complete the training process of one group of training sample, then iteration repeats above step, until reach end condition, end condition is set to optimum individual fitness value and is less than 10 here
-8;
S2.6: the structure optimizing RBF neural, obtains the RBF neural after the training of optimized parameter and structure;
Once the correction of the structure of RBF neural is carried out in training after stopping, and after completing iteration, if certain connection weights of satisfactory solution are approximately zero, (precision is set to 10 here
-6), so delete this connection; If the weights of certain hidden layer node are all approximately zero, so delete this node and parameter thereof, thus reduce node in hidden layer, the population that regeneration is new re-starts repetitive exercise, finally obtains optimum RBF neural structure;
Under it should be noted that the prerequisite that only performance of network does not reduce after structure optimization, could using the network after optimization as the network finally obtained; Otherwise, still using the network of the network before optimization as final acquisition;
S3: the RBF neural input of the textural characteristics of test sample book trained, carries out Images Classification test;
After the texture feature vector value of input sample to be tested image, RBF neural can obtain an Output rusults, when judging which kind of image Output rusults belong to, employing is Euclidean distance determining method, which kind of which is nearestly just judged to be for output vector and four object vectors (00,01,10,11), prerequisite is that distance needs to reach certain precision, if arbitrary distance is all greater than 10
-4then be judged to be other images;
Adopt and carry out test experiments based on the standard genetic optimization sorting technique of RBF neural and the sorting technique based on innovatory algorithm of proposition, concrete experimental result as shown in Table 3 and Table 4; As can be seen from the table, compared to the sorting technique that standard genetic is optimized, the sorting technique based on the self-adapted genetic algorithm improved can be classified to unmanned plane Aerial Images better, obtains the correct classification rate of more succinct network structure and Geng Gao.
The classification results of table 2 standard genetic optimization method
Table 3 classification results of the present invention
The classification results of table 4 two kinds of methods compares
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510846339.8A CN105488528B (en) | 2015-11-26 | 2015-11-26 | Neural network image classification method based on improving expert inquiry method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510846339.8A CN105488528B (en) | 2015-11-26 | 2015-11-26 | Neural network image classification method based on improving expert inquiry method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105488528A true CN105488528A (en) | 2016-04-13 |
| CN105488528B CN105488528B (en) | 2019-06-07 |
Family
ID=55675499
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510846339.8A Expired - Fee Related CN105488528B (en) | 2015-11-26 | 2015-11-26 | Neural network image classification method based on improving expert inquiry method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105488528B (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105938563A (en) * | 2016-04-14 | 2016-09-14 | 北京工业大学 | Weld surface defect identification method based on image texture |
| CN106372660A (en) * | 2016-08-30 | 2017-02-01 | 西安电子科技大学 | Spaceflight product assembly quality problem classification method based on big data analysis |
| CN106898011A (en) * | 2017-01-06 | 2017-06-27 | 广东工业大学 | A kind of method that convolutional neural networks convolution nuclear volume is determined based on rim detection |
| CN107016371A (en) * | 2017-04-09 | 2017-08-04 | 北京工业大学 | UAV Landing Geomorphological Classification method based on improved depth confidence network |
| CN107479368A (en) * | 2017-06-30 | 2017-12-15 | 北京百度网讯科技有限公司 | A kind of method and system of the training unmanned aerial vehicle (UAV) control model based on artificial intelligence |
| CN108615048A (en) * | 2018-04-04 | 2018-10-02 | 浙江工业大学 | It is evolved based on disturbance and fights the defence method of sexual assault to Image Classifier |
| CN108777584A (en) * | 2018-07-06 | 2018-11-09 | 中国石油大学(华东) | A kind of fast Optimization of polarization code decoding parameter |
| CN108873901A (en) * | 2018-06-27 | 2018-11-23 | 深圳市创艺工业技术有限公司 | A kind of Unmanned Systems |
| CN108875639A (en) * | 2018-06-20 | 2018-11-23 | 甘肃万维信息技术有限责任公司 | A kind of optimization and recognition methods based on genetic algorithm recognition of face |
| CN109241817A (en) * | 2018-07-02 | 2019-01-18 | 广东工业大学 | A kind of crops image-recognizing method of unmanned plane shooting |
| CN109765820A (en) * | 2019-01-14 | 2019-05-17 | 南栖仙策(南京)科技有限公司 | A kind of training system for automatic Pilot control strategy |
| CN109919950A (en) * | 2019-03-12 | 2019-06-21 | 哈尔滨工程大学 | A kind of unmanned boat waterborne target image-recognizing method based on genetic neural network |
| CN111105029A (en) * | 2018-10-29 | 2020-05-05 | 北京地平线机器人技术研发有限公司 | Neural network generation method and device and electronic equipment |
| US10748042B2 (en) | 2018-06-22 | 2020-08-18 | Cnh Industrial Canada, Ltd. | Measuring crop residue from imagery using a machine-learned convolutional neural network |
| CN112163068A (en) * | 2020-09-25 | 2021-01-01 | 山东电力研究院 | Information prediction method and system based on autonomous evolution learner |
| CN113537399A (en) * | 2021-08-11 | 2021-10-22 | 西安电子科技大学 | Polarized SAR image classification method and system of multi-target evolutionary graph convolution neural network |
| CN115473672A (en) * | 2022-08-03 | 2022-12-13 | 广西电网有限责任公司电力科学研究院 | Leak-proof hole detection method based on online interactive WEB dynamic defense |
| CN117095308A (en) * | 2023-10-19 | 2023-11-21 | 中国石油大学(华东) | Genetic algorithm-based full-polarization SAR image sea surface oil spill detection method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040044633A1 (en) * | 2002-08-29 | 2004-03-04 | Chen Thomas W. | System and method for solving an optimization problem using a neural-network-based genetic algorithm technique |
| CN102567742A (en) * | 2010-12-15 | 2012-07-11 | 中国科学院电子学研究所 | Automatic classification method of support vector machine based on selection of self-adapting kernel function |
| CN104050505A (en) * | 2013-03-11 | 2014-09-17 | 江南大学 | Multilayer-perceptron training method based on bee colony algorithm with learning factor |
| CN104732249A (en) * | 2015-03-25 | 2015-06-24 | 武汉大学 | Deep learning image classification method based on popular learning and chaotic particle swarms |
-
2015
- 2015-11-26 CN CN201510846339.8A patent/CN105488528B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040044633A1 (en) * | 2002-08-29 | 2004-03-04 | Chen Thomas W. | System and method for solving an optimization problem using a neural-network-based genetic algorithm technique |
| CN102567742A (en) * | 2010-12-15 | 2012-07-11 | 中国科学院电子学研究所 | Automatic classification method of support vector machine based on selection of self-adapting kernel function |
| CN104050505A (en) * | 2013-03-11 | 2014-09-17 | 江南大学 | Multilayer-perceptron training method based on bee colony algorithm with learning factor |
| CN104732249A (en) * | 2015-03-25 | 2015-06-24 | 武汉大学 | Deep learning image classification method based on popular learning and chaotic particle swarms |
Non-Patent Citations (2)
| Title |
|---|
| 何同弟: "高光谱图像的分类技术研究", 《中国博士学位论文全文数据库•信息科技辑》 * |
| 高鹏毅: "BP神经网络分类器优化技术研究", 《中国博士学位论文全文数据库•信息科技辑》 * |
Cited By (26)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105938563A (en) * | 2016-04-14 | 2016-09-14 | 北京工业大学 | Weld surface defect identification method based on image texture |
| CN106372660A (en) * | 2016-08-30 | 2017-02-01 | 西安电子科技大学 | Spaceflight product assembly quality problem classification method based on big data analysis |
| CN106898011A (en) * | 2017-01-06 | 2017-06-27 | 广东工业大学 | A kind of method that convolutional neural networks convolution nuclear volume is determined based on rim detection |
| CN106898011B (en) * | 2017-01-06 | 2019-10-29 | 广东工业大学 | A method of determining convolutional neural networks convolution nuclear volume based on edge detection |
| CN107016371A (en) * | 2017-04-09 | 2017-08-04 | 北京工业大学 | UAV Landing Geomorphological Classification method based on improved depth confidence network |
| CN107479368A (en) * | 2017-06-30 | 2017-12-15 | 北京百度网讯科技有限公司 | A kind of method and system of the training unmanned aerial vehicle (UAV) control model based on artificial intelligence |
| US11150655B2 (en) | 2017-06-30 | 2021-10-19 | Beijing Baidu Netcom Science And Technology Co., Ltd. | Method and system for training unmanned aerial vehicle control model based on artificial intelligence |
| CN108615048A (en) * | 2018-04-04 | 2018-10-02 | 浙江工业大学 | It is evolved based on disturbance and fights the defence method of sexual assault to Image Classifier |
| CN108615048B (en) * | 2018-04-04 | 2020-06-23 | 浙江工业大学 | Defense method for image classifier adversity attack based on disturbance evolution |
| CN108875639A (en) * | 2018-06-20 | 2018-11-23 | 甘肃万维信息技术有限责任公司 | A kind of optimization and recognition methods based on genetic algorithm recognition of face |
| US10748042B2 (en) | 2018-06-22 | 2020-08-18 | Cnh Industrial Canada, Ltd. | Measuring crop residue from imagery using a machine-learned convolutional neural network |
| CN108873901A (en) * | 2018-06-27 | 2018-11-23 | 深圳市创艺工业技术有限公司 | A kind of Unmanned Systems |
| CN109241817A (en) * | 2018-07-02 | 2019-01-18 | 广东工业大学 | A kind of crops image-recognizing method of unmanned plane shooting |
| CN109241817B (en) * | 2018-07-02 | 2022-02-22 | 广东工业大学 | Crop image recognition method shot by unmanned aerial vehicle |
| CN108777584A (en) * | 2018-07-06 | 2018-11-09 | 中国石油大学(华东) | A kind of fast Optimization of polarization code decoding parameter |
| CN111105029A (en) * | 2018-10-29 | 2020-05-05 | 北京地平线机器人技术研发有限公司 | Neural network generation method and device and electronic equipment |
| CN111105029B (en) * | 2018-10-29 | 2024-04-16 | 北京地平线机器人技术研发有限公司 | Neural network generation method, generation device and electronic equipment |
| CN109765820A (en) * | 2019-01-14 | 2019-05-17 | 南栖仙策(南京)科技有限公司 | A kind of training system for automatic Pilot control strategy |
| CN109919950A (en) * | 2019-03-12 | 2019-06-21 | 哈尔滨工程大学 | A kind of unmanned boat waterborne target image-recognizing method based on genetic neural network |
| CN112163068A (en) * | 2020-09-25 | 2021-01-01 | 山东电力研究院 | Information prediction method and system based on autonomous evolution learner |
| CN112163068B (en) * | 2020-09-25 | 2022-11-01 | 国网山东省电力公司电力科学研究院 | Information prediction method and system based on autonomous evolution learner |
| CN113537399A (en) * | 2021-08-11 | 2021-10-22 | 西安电子科技大学 | Polarized SAR image classification method and system of multi-target evolutionary graph convolution neural network |
| CN115473672A (en) * | 2022-08-03 | 2022-12-13 | 广西电网有限责任公司电力科学研究院 | Leak-proof hole detection method based on online interactive WEB dynamic defense |
| CN115473672B (en) * | 2022-08-03 | 2024-03-29 | 广西电网有限责任公司电力科学研究院 | Leak-proof detection method based on online interactive WEB dynamic defense |
| CN117095308A (en) * | 2023-10-19 | 2023-11-21 | 中国石油大学(华东) | Genetic algorithm-based full-polarization SAR image sea surface oil spill detection method |
| CN117095308B (en) * | 2023-10-19 | 2024-01-05 | 中国石油大学(华东) | Genetic algorithm-based full-polarization SAR image sea surface oil spill detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105488528B (en) | 2019-06-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105488528A (en) | Improved adaptive genetic algorithm based neural network image classification method | |
| CN110298663B (en) | Fraud transaction detection method based on sequence wide and deep learning | |
| CN103927531B (en) | It is a kind of based on local binary and the face identification method of particle group optimizing BP neural network | |
| CN111967343A (en) | Detection method based on simple neural network and extreme gradient lifting model fusion | |
| CN113378632A (en) | Unsupervised domain pedestrian re-identification algorithm based on pseudo label optimization | |
| CN105787501B (en) | Power transmission line corridor region automatically selects the vegetation classification method of feature | |
| CN108090658A (en) | Arc fault diagnostic method based on time domain charactreristic parameter fusion | |
| CN106600059A (en) | Intelligent power grid short-term load predication method based on improved RBF neural network | |
| CN106529574B (en) | Image classification method based on sparse autocoder and support vector machines | |
| CN106446942A (en) | Crop disease identification method based on incremental learning | |
| CN111612125B (en) | Novel HTM time pool method and system for online learning | |
| CN103955702A (en) | SAR image terrain classification method based on depth RBF network | |
| CN106096627A (en) | The Polarimetric SAR Image semisupervised classification method that considering feature optimizes | |
| CN105913450A (en) | Tire rubber carbon black dispersity evaluation method and system based on neural network image processing | |
| CN113269252A (en) | Power electronic circuit fault diagnosis method for optimizing ELM based on improved sparrow search algorithm | |
| CN111311702B (en) | Image generation and identification module and method based on BlockGAN | |
| CN110287985B (en) | Depth neural network image identification method based on variable topology structure with variation particle swarm optimization | |
| CN112912902A (en) | Artificial neural network and method for training artificial neural network by using epigenetic neurogenesis | |
| CN110442143A (en) | A kind of unmanned plane situation data clustering method based on combination multiple target dove group's optimization | |
| CN107016371A (en) | UAV Landing Geomorphological Classification method based on improved depth confidence network | |
| Zhang et al. | Evolving neural network classifiers and feature subset using artificial fish swarm | |
| CN112766603A (en) | Traffic flow prediction method, system, computer device and storage medium | |
| CN107766828A (en) | UAV Landing Geomorphological Classification method based on wavelet convolution neutral net | |
| CN110503113A (en) | A kind of saliency object detection method restored based on low-rank matrix | |
| CN108876144B (en) | Transformer substation pre-addressing method based on deep learning algorithm |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190607 Termination date: 20211126 |