阅读说明：本技术 具有包裹类型预检的特征选择方法 (Feature selection approach with package type preliminary examination ) 是由 刘小英 于 2018-05-16 设计创作，主要内容包括：本发明公开了具有包裹类型预检的特征选择方法,涉及包裹类别特征选择,包括包裹类别样本图像,还包括以下步骤：S1,进行图像去噪处理后,对所述包裹类别样本图像进行图像信息统计,获得基本统计特征集；S2,获取基于图像自身属性的特征点集；S3,获取区域定义的特征点集；S4,依据S1-S3获取的特征集,计算每个特征集子项的方差,并选取方差最小的6个特征,再将6个特征采用特征压缩方法进行降维,形成3个描述特征作为图像的特征,作为包裹类别的特征参数。本发明通过选择合适的特征,能够有效提高包裹类别特征选择的稳定性、精确性及实时性；能够综合考虑不同条件下的特征有效性,提高图像特征的抗干扰能力。(The invention discloses the feature selection approach with package type preliminary examination, it is related to wrapping up category feature selection, including wrapping up classification sample image, it is further comprising the steps of: S1, after carrying out image denoising processing, image information statistics is carried out to the package classification sample image, obtains basic statistics feature set；S2 obtains the feature point set based on image self attributes；S3 obtains the feature point set that region defines；S4 calculates the variance of each feature set subitem according to the feature set that S1-S3 is obtained, and choose the smallest 6 features of variance, 6 features are subjected to dimensionality reduction using Feature Compression method again, form feature of 3 Expressive Features as image, the characteristic parameter as package classification.The present invention can effectively improve stability, accuracy and the real-time of package category feature selection by selecting suitable feature；The characteristic validity under different condition can be comprehensively considered, improve the anti-interference ability of characteristics of image.)

1. have package type preliminary examination feature selection approach, including package classification sample image, which is characterized in that further include with Lower step:

S1 after carrying out image denoising processing, carries out image information statistics to the package classification sample image, obtains basic statistics Feature set；

S2 obtains the feature point set based on image self attributes；

S3 obtains the feature point set that region defines；

S4 calculates the variance of each feature set subitem, and choose the smallest 6 spies of variance according to the feature set that S1-S3 is obtained Sign, then 6 features are subjected to dimensionality reduction using Feature Compression method, feature of 3 Expressive Features as image is formed, as package The characteristic parameter of classification.

2. the feature selection approach with package type preliminary examination according to claim 1, which is characterized in that the region is fixed The feature point set of justice, including LOG operator, Forstner operator, SIFT operator.

3. the feature selection approach with package type preliminary examination according to claim 1, which is characterized in that described based on certainly The feature point set of body attribute, including marginal point, angle point, crosspoint.

4. the feature selection approach with package type preliminary examination according to claim 1, which is characterized in that the basic system Count feature, including brightness, center of gravity, grey level histogram.

5. the feature selection approach with package type preliminary examination according to claim 1, which is characterized in that the feature pressure For contracting method using in first 6 of minimum variance ranking, ranking is that the feature of odd number is multiplied with the feature of its latter position, is formed new Feature.

Technical field

The present invention relates to package category feature selections, and in particular to has the feature selection approach of package type preliminary examination.

Background technique

Express firm when collecting express mail should field test view internals, but to test view means tradition single for express delivery detection, therefore, benefit The probability that dangerous goods are detected can be effectively improved by carrying out package detection with image recognition technology.Package classification image at present It is more that there are distracters, it is difficult to the problem of selecting good identification feature.

Summary of the invention

It is more that there are distracters the technical problem to be solved by the present invention is to wrapping up classification image at present, it is difficult to which selection is good to be known The problem of other feature, and it is an object of the present invention to provide the feature selection approach with package type preliminary examination, solves the above problems.

The present invention is achieved through the following technical solutions:

Feature selection approach with package type preliminary examination, including package classification sample image, further comprising the steps of:

S1 after carrying out image denoising processing, carries out image information statistics to the package classification sample image, obtains basic Statistical nature collection；

S2 obtains the feature point set based on image self attributes；

S3 obtains the feature point set that region defines；

S4 calculates the variance of each feature set subitem, and it is 6 the smallest to choose variance according to the feature set that S1-S3 is obtained Feature, then 6 features are subjected to dimensionality reduction using Feature Compression method, feature of 3 Expressive Features as image is formed, as packet Wrap up in the characteristic parameter of classification.

Further, the feature point set that the region defines, including LOG operator, Forstner operator, SIFT operator.

Further, the feature point set based on self attributes, including marginal point, angle point, crosspoint.

Further, the basic statistics feature, including brightness, center of gravity, grey level histogram.

Further, the Feature Compression method is using in first 6 of minimum variance ranking, ranking be odd number feature with The feature of its latter position is multiplied, and forms new feature.

Compared with prior art, the present invention having the following advantages and benefits:

1, the present invention has the feature selection approach of package type preliminary examination, by selecting suitable feature, can effectively mention Stability, accuracy and the real-time of height package category feature selection；

2, the present invention has the feature selection approach of package type preliminary examination, and the feature that can be comprehensively considered under different condition has Effect property, improves the anti-interference ability of characteristics of image.

Specific embodiment

To make the objectives, technical solutions, and advantages of the present invention clearer, below with reference to embodiment, the present invention is made Further to be described in detail, exemplary embodiment of the invention and its explanation for explaining only the invention, are not intended as to this The restriction of invention.