技术领域technical field
本发明涉及基于稀疏异构分组的高光谱遥感影像的分类器构建方法。The invention relates to a classifier construction method for hyperspectral remote sensing images based on sparse heterogeneous grouping.
背景技术Background technique
高光谱遥感影像是通过高光谱传感器，在电磁波谱的紫外、可见光、近红外和中红外区域，以数十至数百个连续且细分的光谱波段对目标区域同时成像产生的。利用高光谱影像可以在更深层次上反应地物的特征，对于识别较难区分的地物具有很重要的作用，目前已经广泛的应用于地表分类、农业监测、环境管理等领域。Hyperspectral remote sensing images are generated by simultaneous imaging of target areas with tens to hundreds of continuous and subdivided spectral bands in the ultraviolet, visible, near-infrared, and mid-infrared regions of the electromagnetic spectrum through hyperspectral sensors. The use of hyperspectral images can reflect the characteristics of ground objects at a deeper level, and plays an important role in identifying difficult-to-distinguish ground objects. It has been widely used in land surface classification, agricultural monitoring, environmental management and other fields.
要利用高光谱遥感影像需要对每个像元进行分类，获得每个像元对应的土地利用类型。构建分类器可以对高光谱影像进行自动分类可对影像中的每一个像元赋予对应的类目，因此构建分类器对于高光谱影像应用十分重要。由于高光谱遥感影像包含数百个波段，数据维度较高，直接构造分类模型会引起过渡拟合现象，所以目前采用的技术主要是先利用主成分分析、基于决策能力的属性选取、基于遗传算法的染色体选取等技术进行高光谱波段的选取，然后根据选取到的波段利用神经网、支持向量机、决策树等算法构造分类器，进而利用该分类器进行自动化分类。此类方法具有以下两个局限性，一是，此类方法需要大量的训练样本，而某些地区可获得的训练样本数量非常少，难以保证选取到有价值波段；二是，较难保证选取到的波段具有足够的差异性来全面的反应待分类的目标，使得获得的分类模型与分类目标之间存在较大偏差。两种局限性均会导致分类精度降低，使得高光谱遥感影像分类质量下降。To use hyperspectral remote sensing images, it is necessary to classify each pixel and obtain the land use type corresponding to each pixel. Building a classifier can automatically classify hyperspectral images and assign a corresponding category to each pixel in the image, so building a classifier is very important for hyperspectral image applications. Since the hyperspectral remote sensing image contains hundreds of bands, the data dimension is high, and the direct construction of the classification model will cause transition fitting phenomenon, so the current technology mainly uses principal component analysis, attribute selection based on decision-making ability, and genetic algorithm-based The chromosome selection and other technologies are used to select hyperspectral bands, and then according to the selected bands, a classifier is constructed using neural network, support vector machine, decision tree and other algorithms, and then the classifier is used for automatic classification. This type of method has the following two limitations. First, this type of method requires a large number of training samples, and the number of training samples available in some areas is very small, so it is difficult to ensure that valuable bands are selected; second, it is difficult to ensure the selection of The obtained bands have sufficient differences to fully reflect the target to be classified, so that there is a large deviation between the obtained classification model and the classification target. Both limitations will lead to a decrease in classification accuracy, resulting in a decline in the classification quality of hyperspectral remote sensing images.
发明内容Contents of the invention
本发明是为了解决现有技术需要大量的训练样本及分类精度降低的问题，而提出的一种基于稀疏异构分组的高光谱遥感影像的分类器构建方法。The present invention proposes a method for constructing a classifier based on sparse heterogeneous grouping of hyperspectral remote sensing images in order to solve the problems of requiring a large number of training samples and reducing classification accuracy in the prior art.
一种基于稀疏异构分组的高光谱遥感影像的分类器构建方法按以下步骤实现：A classifier construction method for hyperspectral remote sensing images based on sparse heterogeneous grouping is implemented in the following steps:
步骤一：输入一个待分类的高光谱遥感影像；Step 1: Input a hyperspectral remote sensing image to be classified;
步骤二：输入高光谱遥感影像分类的类目个数M，输入有分类标签的训练样本集合LC，在一个高光谱遥感影像中随机选取获得无分类标签的训练样本集合LUC，整合有分类标签的训练样本集合和无分类标签的训练样本集合构造样本集L；Step 2: Input the number M of categories of hyperspectral remote sensing image classification, input the training sample set LC with classification labels, randomly select a training sample set LUC without classification labels from a hyperspectral remote sensing image, and integrate the training sample set LUC with classification labels The training sample set and the training sample set without classification labels construct the sample set L;
步骤三：高光谱遥感影像包含波段集合Band，对每一个波段构造稀疏描述矢量，构成稀疏描述矢量集合V：Step 3: The hyperspectral remote sensing image contains a band set Band, and a sparse description vector is constructed for each band to form a sparse description vector set V:
V＝(V1,V2,…,Vi,…,VBN)V=(V1 ,V2 ,...,Vi ,...,VBN )
V1为Band1的描述矢量，V2为Band2的描述矢量，Vi为Bandi的描述矢量，VBN为BandBN的描述矢量；V1 is the description vector of Band1 , V2 is the description vector of Band2 , Vi is the description vector of Bandi , VBN is the description vector of BandBN ;
Band＝(Band1,Band2,…,Bandi,…,BandBN)Band＝(Band1 ,Band2 ,…,Bandi ,…,BandBN )
Band1为第1个波段，Band2为第2个波段，Bandi为第i个波段，BandBN为第BN个波段；Band1 is the first band, Band2 is the second band, Bandi is the i-th band, and BandBN is the BN-th band;
步骤四：根据稀疏描述矢量集合V，将高光谱遥感影像的所有波段分为M个异构组，构成异构组列表GL；Step 4: According to the sparse description vector set V, divide all the bands of the hyperspectral remote sensing image into M heterogeneous groups to form a heterogeneous group list GL;
步骤五：在异构组列表GL中，每个组取出组内的第一个和最后一个波段，构成待训练波段列表TBList；Step 5: In the heterogeneous group list GL, each group takes out the first and last bands in the group to form the band list TBList to be trained;
步骤六：根据TBList和有分类标签的训练样本集合LS构造训练样本子集，通过支持向量机算法学习获得高光谱遥感影像的分类器。Step 6: Construct a training sample subset according to TBList and the training sample set LS with classification labels, and learn a classifier for hyperspectral remote sensing images through support vector machine algorithm.
发明效果：Invention effect:
本发明提供一种基于稀疏异构分组的高光谱遥感影像的分类器构建方法，利用本方法可以利用较少样本选取到一组有足够的差异性的高光谱遥感影像的波段，并利用这些波段构建一个分类模型。通过该方法可以实现较高精度的高光谱遥感影像分类。可以广泛的应用于地表分类、农业监测、环境管理等领域，尤其是对于训练样本获取困难、地物类型混杂难于确定样本的地区具有较好的应用价值。The invention provides a method for constructing a hyperspectral remote sensing image classifier based on sparse heterogeneous grouping. Using this method, a group of hyperspectral remote sensing image bands with sufficient differences can be selected with fewer samples, and these bands can be used Build a classification model. This method can achieve high-precision hyperspectral remote sensing image classification. It can be widely used in surface classification, agricultural monitoring, environmental management and other fields, especially in areas where it is difficult to obtain training samples and the types of ground objects are mixed and it is difficult to determine samples.
附图说明Description of drawings
图1为本发明流程图；Fig. 1 is a flowchart of the present invention;
图2为构造样本集L的流程图；Fig. 2 is the flowchart of constructing sample set L;
图3为获得稀疏描述矢量集合V的流程图；Fig. 3 is the flowchart of obtaining sparse description vector set V;
图4为获得异构组列表GL的流程图；FIG. 4 is a flowchart of obtaining a heterogeneous group list GL;
图5为获得待训练波段列表TBList的流程图；Fig. 5 is the flowchart of obtaining the band list TBList to be trained;
图6为获得高光谱遥感影像的分类器的流程图。Fig. 6 is a flowchart of a classifier for obtaining hyperspectral remote sensing images.
具体实施方式detailed description
具体实施方式一：如图1所示，一种基于稀疏异构分组的高光谱遥感影像的分类器构建方法包括以下步骤：Specific embodiment one: as shown in Figure 1, a classifier construction method based on sparse heterogeneous grouping of hyperspectral remote sensing images includes the following steps:
步骤一：输入一个待分类的高光谱遥感影像；Step 1: Input a hyperspectral remote sensing image to be classified;
步骤二：如图2所示，输入高光谱遥感影像分类的类目个数M，输入有分类标签的训练样本集合LC，在一个高光谱遥感影像中随机选取获得无分类标签的训练样本集合LUC，整合有分类标签的训练样本集合和无分类标签的训练样本集合构造样本集L；Step 2: As shown in Figure 2, input the number M of hyperspectral remote sensing image classification categories, input the training sample set LC with classification labels, and randomly select a training sample set LUC without classification labels from a hyperspectral remote sensing image , integrate the training sample set with classification label and the training sample set without classification label to construct sample set L;
步骤三：高光谱遥感影像包含波段集合Band，对每一个波段构造稀疏描述矢量，构成稀疏描述矢量集合V：Step 3: The hyperspectral remote sensing image contains a band set Band, and a sparse description vector is constructed for each band to form a sparse description vector set V:
V＝(V1,V2,…,Vi,…,VBN)V=(V1 ,V2 ,...,Vi ,...,VBN )
V1为Band1的描述矢量，V2为Band2的描述矢量，Vi为Bandi的描述矢量，VBN为BandBN的描述矢量；V1 is the description vector of Band1 , V2 is the description vector of Band2 , Vi is the description vector of Bandi , VBN is the description vector of BandBN ;
Band＝(Band1,Band2,…,Bandi,…,BandBN)Band＝(Band1 ,Band2 ,…,Bandi ,…,BandBN )
Band1为第1个波段，Band2为第2个波段，Bandi为第i个波段，BandBN为第BN个波段；Band1 is the first band, Band2 is the second band, Bandi is the i-th band, and BandBN is the BN-th band;
步骤四：根据稀疏描述矢量集合V，将高光谱遥感影像的所有波段分为M个异构组，构成异构组列表GL；Step 4: According to the sparse description vector set V, divide all the bands of the hyperspectral remote sensing image into M heterogeneous groups to form a heterogeneous group list GL;
步骤五：在异构组列表GL中，每个组取出组内的第一个和最后一个波段，构成待训练波段列表TBList；Step 5: In the heterogeneous group list GL, each group takes out the first and last bands in the group to form the band list TBList to be trained;
步骤六：根据TBList和有分类标签的训练样本集合LS构造训练样本子集，通过支持向量机算法学习获得高光谱遥感影像的分类器。Step 6: Construct a training sample subset according to TBList and the training sample set LS with classification labels, and learn a classifier for hyperspectral remote sensing images through support vector machine algorithm.
具体实施方式二：本实施方式与具体实施方式一不同的是：所述步骤一中的高光谱遥感影像的宽度为Width，高度为Height，包含BN个波段，所有波段的构成波段的集合Band。Embodiment 2: This embodiment differs from Embodiment 1 in that: the hyperspectral remote sensing image in step 1 has a width of Width and a height of Height, and includes BN bands, and all bands form a set Band of bands.
其它步骤及参数与具体实施方式一相同。Other steps and parameters are the same as those in Embodiment 1.
具体实施方式三：本实施方式与具体实施方式一或二不同的是：所述步骤二中LC为一个包含N个样本的集合；Specific embodiment 3: The difference between this embodiment and specific embodiment 1 or 2 is that: in the step 2, LC is a set containing N samples;
LC＝(LC1,LC2,…LCI,…LCN)LC＝(LC1 ,LC2 ,…LCI ,…LCN )
在LC中LC1是第1个样本，LC2是第2个样本，LCI是第I个样本，1≤I≤N，LCN是第N个样本；In LC, LC1 is the first sample, LC2 is the second sample, LCI is the I-th sample, 1≤I≤N, LCN is the N-th sample;
对于每个样本LCI＝(X,Y,C)，其中，X为该样本在高光谱遥感影像中的X坐标，Y为该样本在高光谱遥感影像中的Y坐标；C为类目编号，取值范围为1到M。For each sample LCI = (X, Y, C), where X is the X coordinate of the sample in the hyperspectral remote sensing image, Y is the Y coordinate of the sample in the hyperspectral remote sensing image; C is the category number , the value ranges from 1 to M.
其它步骤及参数与具体实施方式一或二相同。Other steps and parameters are the same as those in Embodiment 1 or Embodiment 2.
具体实施方式四：本实施方式与具体实施方式一至三之一不同的是：所述步骤二中LUC包含U个样本；Embodiment 4: This embodiment differs from Embodiment 1 to Embodiment 3 in that the LUC in the step 2 includes U samples;
LUC＝(LUC1,LUC2,…LUCJ,…LUCU)LUC＝(LUC1 ,LUC2 ,...LUCJ ,...LUCU )
在LUC中LUC1是第1个样本，LUC2是第2个样本，LUCJ是第J个类目，1≤J≤N,LUCU是第U个类目；LUCJ＝(UX,UY,UC)；In LUC, LUC1 is the first sample, LUC2 is the second sample, LUCJ is the Jth category, 1≤J≤N, LUCU is the Uth category; LUCJ = (UX, UY ,UC);
其中，UX＝Random(1～Width)为1到Width之间的随机整数，为该样本在高光谱遥感影像中的X坐标；UY＝Random(1～Height)为1到Height之间的随机整数，为该样本在高光谱遥感影像中的Y坐标；UC＝-1表示对应的样本所属类别未知。Among them, UX=Random(1~Width) is a random integer between 1 and Width, which is the X coordinate of the sample in the hyperspectral remote sensing image; UY=Random(1~Height) is a random integer between 1 and Height , is the Y coordinate of the sample in the hyperspectral remote sensing image; UC=-1 means that the category of the corresponding sample is unknown.
其它步骤及参数与具体实施方式一至三之一相同。Other steps and parameters are the same as those in Embodiments 1 to 3.
具体实施方式五：本实施方式与具体实施方式一至四之一不同的是：所述步骤二中构造样本集L具体为：Specific implementation mode five: the difference between this implementation mode and one of the specific implementation modes one to four is: the construction of the sample set L in the step two is specifically:
L＝LC∪LUCL＝LC∪LUC
样本集L为分类标签的训练样本集合LC与无分类标签的训练样本集和LUC的并集，L表示为：The sample set L is the union of the training sample set LC with classification labels and the training sample set without classification labels and LUC, and L is expressed as:
L＝(L1,L2,…,LU+N)L＝(L1 ,L2 ,…,LU+N )
在L中共包含U+N个样本，L1为第1个样本，L2为第2个样本，LU+N为第U+N个样本；There are U+N samples in L, L1 is the first sample, L2 is the second sample, LU+N is the U+N sample;
其中任意一个元素Lk＝(SX，SY，SC)，SX为其横坐标，SY为其纵坐标，SC为其对应类目；若该样本来自LC，则SX＝X，SY＝Y，SC＝C；若该样本来自于LUC，则SX＝UX，SY＝UY，SC＝UC。Any one of the elements Lk = (SX, SY, SC), SX is its abscissa, SY is its ordinate, and SC is its corresponding category; if the sample comes from LC, then SX=X, SY=Y, SC =C; if the sample comes from LUC, then SX=UX, SY=UY, SC=UC.
其它步骤及参数与具体实施方式一至四之一相同。Other steps and parameters are the same as in one of the specific embodiments 1 to 4.
具体实施方式六：本实施方式与具体实施方式一至五之一不同的是：如图3所示，所述步骤三中Bandi的描述矢量Vi的构造过程为：Specific embodiment six: the difference between this embodiment and one of specific embodiments one to five is: as shown in Figure 3, the construction process of the description vector Vi of Bandi in the step three is:
步骤三一：初始化Bandi的描述矢量Vi；Step 31: Initialize the description vector Vi of Band i;
Vi＝zeros(M)Vi =zeros(M)
其中，zeros(M)表示生成一个M个元素的矢量，该矢量所有元素值为0；Among them, zeros(M) means to generate a vector of M elements, and all elements of the vector are 0;
步骤三二：计算Bandi中的最大值Max和最小值Min；Step 32: Calculate the maximum value Max and the minimum value Min in Bandi ;
步骤三三：描述矢量计数器Counter＝1，即将描述矢量计数器Counter置为1；Step 33: describe the vector counter Counter=1, that is, set the description vector counter Counter to 1;
步骤三四：初始化计数矢量C1和计数矢量C2；Step three and four: Initialize counting vector C1 and counting vector C2;
C1＝zeros(10)，其中zeros(10)表示初始化一个10个元素，且值全为0的矢量；C1=zeros(10), where zeros(10) means to initialize a vector with 10 elements and all values are 0;
C2＝zeros(10)，其中zeros(10)表示初始化一个10个元素，且值全为0的矢量；C2=zeros(10), where zeros(10) means to initialize a vector with 10 elements and all values are 0;
步骤三五：分段计数器SCounter＝1，即将分段计数器SCounter置为1；Step 35: segment counter SCounter=1, that is, set the segment counter SCounter to 1;
步骤三六：构造值域区间[Lower,Upper]；其中，[Lower,Upper]通过如下公式计算：Step 36: Construct the range interval [Lower,Upper]; where, [Lower,Upper] is calculated by the following formula:
Lower＝Min+(SCounter-1)×(Max-Min)/10Lower＝Min+(SCounter-1)×(Max-Min)/10
Upper＝Min+SCounter×(Max-Min)/10Upper＝Min+SCounter×(Max-Min)/10
步骤三七：对于样本集L，每一个样本根据其SX和SY坐标取出第i个波段对应的值value，统计数量落入区间[Lower,Upper]的样本个数计入C1和C2之中；Step 37: For the sample set L, take out the value corresponding to the i-th band for each sample according to its SX and SY coordinates, and count the number of samples whose statistics fall into the interval [Lower, Upper] into C1 and C2;
C1[SCounter]＝在样本集L中，样本的value在区间[Lower,Upper]内，且所属类别SC等于Counter的样本个数；C1[SCounter]=In the sample set L, the value of the sample is in the interval [Lower,Upper], and the category SC is equal to the number of samples of Counter;
C2[SCounter]＝在样本集L中，样本的value在区间[Lower,Upper]内，且所属类别SC等于-1的样本个数；C2[SCounter]=In the sample set L, the value of the sample is in the interval [Lower,Upper], and the number of samples whose category SC is equal to -1;
步骤三八：SCounter＝SCounter+1，分段计数器SCounter的数值增加1；Step 38: SCounter=SCounter+1, the value of the segment counter SCounter increases by 1;
步骤三九：若SCounter>10，转到步骤三十，否则转至步骤三六；Step 39: If SCounter>10, go to step 30, otherwise go to step 36;
步骤三十：根据C1和C2计算有标记样本对于第Counter个类目的稀疏程度；计算公式为：Step 30: According to C1 and C2, calculate the degree of sparseness of the labeled samples for the category of Counter; the calculation formula is:
其中C1[t]为C1中的第t个元素，C2[t]为C2中的第t个元素；Where C1[t] is the tth element in C1, and C2[t] is the tth element in C2;
步骤三十一：将Vi的第Counter个元素赋值为Sparse；Step 31: Assign Sparse to the Counter element of Vi ;
Vi[Counter]＝SparseVi [Counter] = Sparse
步骤三十二：Counter＝Counter+1，描述矢量计数器Counter的数值增加1；Step 32: Counter=Counter+1, the value describing the vector counter Counter is increased by 1;
步骤三十三：若Counter>M，则转到步骤三十四，否则转到步骤三四；Step 33: If Counter>M, go to step 34, otherwise go to step 34;
步骤三十四：计算Vi的过程结束。Step 34: The process of calculating Vi ends.
其它步骤及参数与具体实施方式一至五之一相同。Other steps and parameters are the same as one of the specific embodiments 1 to 5.
具体实施方式七：本实施方式与具体实施方式一至六之一不同的是：所述步骤四中GL具体为：Embodiment 7: The difference between this embodiment and one of Embodiments 1 to 6 is that the GL in step 4 is specifically:
GL＝(GL1,GL2,…,GLi,..,GLM)GL＝(GL1 ,GL2 ,...,GLi ,..,GLM )
其中，GL1为第1个类目对应的分组，GL2为第2个类目对应的分组，GLi为第i个类目对应的分组，1≤i≤M，GLM为第M个类目对应的分组。Among them, GL1 is the group corresponding to the first category, GL2 is the group corresponding to the second category, GLi is the group corresponding to the i-th category, 1≤i≤M, GLM is the M-th The grouping corresponding to the category.
其它步骤及参数与具体实施方式一至六之一相同。Other steps and parameters are the same as one of the specific embodiments 1 to 6.
具体实施方式八：本实施方式与具体实施方式一至七之一不同的是：如图4所示，所述GLi的构造过程为：Embodiment 8: This embodiment is different from Embodiment 1 to Embodiment 7 in that: as shown in FIG. 4 , the construction process of the GLi is as follows:
步骤a：初始化GLi，其内容为空，GLi＝()；Step a: Initialize GLi , its content is empty, GLi =();
步骤b：描述矢量列表V，根据每一个描述矢量的第i个元素的数值，找到数值最大的矢量位置j，即第j个波段；Step b: describe the vector list V, according to the value of the i-th element of each description vector, find the vector position j with the largest value, that is, the j-th band;
步骤c：将第j个波段Bandj并入GLi中，GLi＝GLi∪Bandj；Step c: Merge the jth band Bandj into GLi , GLi=GLi∪Bandj ;
步骤d：将Vj的所有元素设置为-1；Step d: set all elements of Vj to -1;
步骤e：若GLi中的元素个数<(BN/M)，则转到步骤b，否则转到步骤f；BN/M为BN除以M的商。Step e: If the number of elements in GLi <(BN/M), go to step b, otherwise go to step f; BN/M is the quotient of BN divided by M.
步骤f：构造GLi的过程结束。Step f: The process of constructing GLi ends.
其它步骤及参数与具体实施方式一至七之一相同。Other steps and parameters are the same as one of the specific embodiments 1 to 7.
具体实施方式九：本实施方式与具体实施方式一至八之一不同的是：如图5所示，所述步骤五中构成待训练波段列表TBList的具体过程为：Specific embodiment nine: the difference between this embodiment and one of the specific embodiments one to eight is: as shown in Figure 5, the specific process of forming the band list TBList to be trained in the step five is:
步骤五一：初始化TBList，其内容为空，TBList＝()；Step 51: Initialize TBList, its content is empty, TBList=();
步骤五二：GCounter＝1，将群组计数器GCounter置为1；Step 52: GCounter=1, set the group counter GCounter to 1;
步骤五三：CGroup＝GL[GCounter]，取出GL中的第GCounter个元素放入当前群组变量CGroup中；Step five and three: CGroup=GL[GCounter], take out the GCounter element in GL and put it into the current group variable CGroup;
步骤五四：TBList＝TBList∪CGroup[1]，将当前群组变量CGroup的第一个元素加入到TBList中；Step five and four: TBList=TBList∪CGroup[1], the first element of the current group variable CGroup is added in the TBList;
步骤五五：TBList＝TBList∪CGroup[last],，将当前群组变量CGroup的最后一个元素加入到TBList中，其中last代表CGroup的最后一个元素在集合中的位置；Step five and five: TBList=TBList∪CGroup[last], add the last element of the current group variable CGroup to the TBList, where last represents the position of the last element of the CGroup in the collection;
步骤五六：GCounter＝GCounter+1，群组计数器GCounter的数值增加1；Step five and six: GCounter=GCounter+1, the value of the group counter GCounter is increased by 1;
步骤五七：若GCounter<M，则转至步骤五三，否则转至步骤五八；Step five and seven: if GCounter<M, then go to step five and three, otherwise go to step five and eight;
步骤五八：构造TBList过程结束。Step 58: The process of constructing TBList ends.
其它步骤及参数与具体实施方式一至八之一相同。Other steps and parameters are the same as those in Embodiments 1 to 8.
具体实施方式十：本实施方式与具体实施方式一至九之一不同的是：如图6所示，所述步骤六中获得高光谱遥感影像的分类器的具体过程为：Embodiment 10: The difference between this embodiment and Embodiment 1 to Embodiment 9 is that: as shown in FIG. 6 , the specific process of obtaining the classifier of the hyperspectral remote sensing image in the step 6 is:
步骤六一：SubCounter＝1，将样本构造计数器SubCounter置为1；Step 61: SubCounter=1, set the sample construction counter SubCounter to 1;
步骤六二：SubSamples＝()，将训练样本子集SubSamples置为空集合；Step six two: SubSamples=(), set the training sample subset SubSamples as an empty set;
步骤六三：[X,Y,C]＝LC[SubCounter]，取出LC中第SubCounter个样本的X，Y坐标以及其所属类别C；Step six three: [X, Y, C]=LC[SubCounter], take out the X, Y coordinates of the SubCounter sample in LC and its category C;
步骤六四：取出TBList所有波段在X，Y位置的数值和所属类别C构成一个样本放入样本变量Sample中；Step 64: Take out the values of all bands in TBList at the X, Y positions and the category C to form a sample and put it into the sample variable Sample;
步骤六五：SubSamples＝SubSamples∪Sample，将样本变量Sample所存储的内容加入到训练样本子集SubSamples中；Step 65: SubSamples=SubSamples∪Sample, add the content stored in the sample variable Sample to the training sample subset SubSamples;
步骤六六：SubCounter＝SubCounter+1，样本构造计数器SubCounter的数值增加1；Step six six: SubCounter=SubCounter+1, the value of the sample construction counter SubCounter is increased by 1;
步骤六七：若SubCounter<＝N，则转到步骤六三，否则转至步骤六八；Step 67: if SubCounter<=N, then go to step 63, otherwise go to step 68;
步骤六八：利用支持向量机算法学习SubSamples中的所有样本获得一个分类器；Step six and eight: use the support vector machine algorithm to learn all samples in SubSamples to obtain a classifier;
步骤六九：输出分类器，分类器对高光谱遥感影像进行自动分类。Step 69: Outputting a classifier, which automatically classifies hyperspectral remote sensing images.
其它步骤及参数与具体实施方式一至九之一相同。Other steps and parameters are the same as one of the specific embodiments 1 to 9.
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610472840.7ACN106096663B (en) | 2016-06-24 | 2016-06-24 | A kind of classifier construction method of the target in hyperspectral remotely sensed image based on the grouping of sparse isomery |
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610472840.7ACN106096663B (en) | 2016-06-24 | 2016-06-24 | A kind of classifier construction method of the target in hyperspectral remotely sensed image based on the grouping of sparse isomery |
| Publication Number | Publication Date |
|---|---|
| CN106096663Atrue CN106096663A (en) | 2016-11-09 |
| CN106096663B CN106096663B (en) | 2019-03-26 |
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610472840.7AActiveCN106096663B (en) | 2016-06-24 | 2016-06-24 | A kind of classifier construction method of the target in hyperspectral remotely sensed image based on the grouping of sparse isomery |
| Country | Link |
|---|---|
| CN (1) | CN106096663B (en) |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107071858A (en)* | 2017-03-16 | 2017-08-18 | 许昌学院 | A kind of subdivision remote sensing image method for parallel processing under Hadoop |
| CN111666313A (en)* | 2020-05-25 | 2020-09-15 | 中科星图股份有限公司 | Correlation construction and multi-user data matching method based on multi-source heterogeneous remote sensing data |
| CN112257603A (en)* | 2020-10-23 | 2021-01-22 | 深圳大学 | Hyperspectral image classification method and related equipment |
| CN115346009A (en)* | 2022-05-18 | 2022-11-15 | 上海航遥信息技术有限公司 | Geographic entity semantic modeling method based on hyperspectral data and inclined three-dimensional data |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090092299A1 (en)* | 2007-10-03 | 2009-04-09 | Siemens Medical Solutions Usa, Inc. | System and Method for Joint Classification Using Feature Space Cluster Labels |
| CN102096825A (en)* | 2011-03-23 | 2011-06-15 | 西安电子科技大学 | Graph-based semi-supervised high-spectral remote sensing image classification method |
| CN104751191A (en)* | 2015-04-23 | 2015-07-01 | 重庆大学 | Sparse self-adaptive semi-supervised manifold learning hyperspectral image classification method |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090092299A1 (en)* | 2007-10-03 | 2009-04-09 | Siemens Medical Solutions Usa, Inc. | System and Method for Joint Classification Using Feature Space Cluster Labels |
| CN102096825A (en)* | 2011-03-23 | 2011-06-15 | 西安电子科技大学 | Graph-based semi-supervised high-spectral remote sensing image classification method |
| CN104751191A (en)* | 2015-04-23 | 2015-07-01 | 重庆大学 | Sparse self-adaptive semi-supervised manifold learning hyperspectral image classification method |
| Title |
|---|
| CE ZHANG ET.AL: "A novel multi-parameter support vector machine for image classification", 《INTERNATIONAL JOURNAL OF REMOTE SENSING》* |
| 王立国 等: "引入负相似的高光谱图像半监督分类", 《信号处理》* |
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107071858A (en)* | 2017-03-16 | 2017-08-18 | 许昌学院 | A kind of subdivision remote sensing image method for parallel processing under Hadoop |
| CN111666313A (en)* | 2020-05-25 | 2020-09-15 | 中科星图股份有限公司 | Correlation construction and multi-user data matching method based on multi-source heterogeneous remote sensing data |
| CN111666313B (en)* | 2020-05-25 | 2023-02-07 | 中科星图股份有限公司 | Correlation construction and multi-user data matching method based on multi-source heterogeneous remote sensing data |
| CN112257603A (en)* | 2020-10-23 | 2021-01-22 | 深圳大学 | Hyperspectral image classification method and related equipment |
| CN112257603B (en)* | 2020-10-23 | 2022-06-17 | 深圳大学 | Hyperspectral image classification method and related equipment |
| CN115346009A (en)* | 2022-05-18 | 2022-11-15 | 上海航遥信息技术有限公司 | Geographic entity semantic modeling method based on hyperspectral data and inclined three-dimensional data |
| CN115346009B (en)* | 2022-05-18 | 2023-04-28 | 上海航遥信息技术有限公司 | Geographic entity semantic modeling method based on hyperspectral data and oblique three-dimensional data |
| Publication number | Publication date |
|---|---|
| CN106096663B (en) | 2019-03-26 |
| Publication | Publication Date | Title |
|---|---|---|
| CN105825511B (en) | A kind of picture background clarity detection method based on deep learning | |
| CN110309867B (en) | A mixed gas identification method based on convolutional neural network | |
| CN104899255B (en) | Suitable for the construction method of the image data base of training depth convolutional neural networks | |
| CN110119687A (en) | Detection method based on the road surface slight crack defect that image procossing and convolutional neural networks combine | |
| CN106355192A (en) | Support vector machine method based on chaos and grey wolf optimization | |
| Yang et al. | Detecting interchanges in road networks using a graph convolutional network approach | |
| CN107871136A (en) | Image Recognition Method Based on Convolutional Neural Network with Sparsity Random Pooling | |
| CN106096663A (en) | A kind of grader construction method of the target in hyperspectral remotely sensed image being grouped based on sparse isomery | |
| CN104484681A (en) | Hyperspectral remote sensing image classification method based on space information and ensemble learning | |
| CN108830243A (en) | Hyperspectral image classification method based on capsule network | |
| CN106485325B (en) | Two phase flow multivariate information fusion method based on complex network and deep learning and application | |
| CN109376753A (en) | A Densely Connected 3D Spatial Spectral Separation Convolutional Deep Network and Its Construction Method | |
| CN105678309A (en) | An image multi-label annotation algorithm based on multi-instance bag feature learning | |
| CN107203761B (en) | Road width estimation method based on high-resolution satellite image | |
| EP2889366A1 (en) | Culture medium information registration system, colony detection device, program, and health management system | |
| CN105160351A (en) | Semi-monitoring high-spectral classification method based on anchor point sparse graph | |
| CN110263735A (en) | A method of tree species classification being carried out to artificial forest high-spectral data using Three dimensional convolution neural network | |
| CN110032654A (en) | A kind of supermarket's commodity input method and system based on artificial intelligence | |
| Guo et al. | Automatic and accurate calculation of rice seed setting rate based on image segmentation and deep learning | |
| CN112418079B (en) | Industrial building video monitoring and installation method based on combination of map analysis and optimal path searching algorithm of security level | |
| CN116246171B (en) | A spatial-spectral multi-scale hyperspectral remote sensing image target detection method and device | |
| CN117015812A (en) | System for clustering data points | |
| CN107895136A (en) | A kind of colliery area recognizing method and system | |
| CN116704241A (en) | A hyperspectral remote sensing image classification method with full-channel 3D convolutional neural network | |
| CN106408009A (en) | Depth belief network-based neighborhood weighted averaging hyperspectral image classification method |
| 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 |