基于BoF模型的多特征融合果蔬图像分类方法

doi:10.16180/j.cnki.issn1007-7820.2020.07.009

摘要/Abstract

摘要：

针对传统BoF模型无法有效利用图像颜色及纹理来更好地表述果蔬特征的问题,文中提出了一种在BoF模型中进行多特征融合的果蔬图像分类算法。该算法首先提取并融合图像的颜色矩和SURF特征形成SURFC特征描述子;然后分别对CLBP及SURFC特征进行K-均值聚类以生成特征词典,并使用特征词典对所有特征量化编码;最后使用SVM对编码结果进行训练得到分类器并识别。实验结果表明,BoF模型融合颜色和纹理特征后,在果蔬图像分类效果上明显优于单一特征或者其他特征融合的BoF模型,识别率最高可达到94%,更适合果蔬图像分类。

关键词: BoF模型, SURF, 果蔬识别, 特征融合, CLBP, SVM

Abstract:

In view of the problem that traditional BoF model cannot effectively use image color and texture to better express fruit and vegetable characteristics, this paper proposed a multi-feature fusion algorithm for fruit and vegetable image classification in BoF model. Firstly, the algorithm extracted and fused the color moments and SURF features of the images to form SURFC feature descriptors. Secondly, K-means clustering was performed on CLBP and SURFC features to generate a feature dictionary, and all features were quantized by using the feature dictionary. Finally, SVM was used to train the coding result to get the classifier and recognize it. Results of the experiment showed that the BoF model with fused color and texture features was significantly better than the BoF model with single feature or other feature fusion. The recognition rate was up to 94%, which was more suitable for fruit and vegetable image classification.

Key words: BoF model, SURF, fruits and vegetables recognition, feature fusion, CLBP, SVM

中图分类号:

TP391

张泽晨,巨志勇. 基于BoF模型的多特征融合果蔬图像分类方法[J]. 电子科技, 2020, 33(7): 41-45.

ZHANG Zechen,JU Zhiyong. Multi-feature Fusion Fruit and Vegetable Image Classification Based on Bag of Feature Model[J]. Electronic Science and Technology, 2020, 33(7): 41-45.

图/表 9

图1

图2

图3

表1

表2

表3

表4

图4

图5

参考文献 17

[1]	Arivazhagan S, Shebiah R N, Nidhyanandhan S S, et al. Fruit recognition using color and texture features[J]. Journal of Emerging Trends in Computing and Information Sciences, 2010,1(2):90-94.
[2]	Zhang Y, Wu L. Classification of fruits using computer vision and a multiclass support vector machine[J]. Sensors, 2012,12(9):12489-12505.
[3]	Rocha A, Hauagge D C, Wainer J, et al. Automatic fruit and vegetable classification from images[J]. Computers and Electronics in Agriculture, 2010,70(1):96-104.
[4]	Tao H, Zhao L, Xi J, et al. Fruits and vegetables recognition based on color and texture features[J]. Transactions of the Chinese Society of Agricultural Engineering, 2014,30(16):305-311.
[5]	Guo Z, Zhang L, Zhang D. A completed modeling of local binary pattern operator for texture classification[J]. IEEE Transactions on Image Processing, 2010,19(6):1657-1663.
[6]	白鑫, 卫琳. 基于双级特征提取与度量的图像检索算法[J]. 包装工程, 2018,39(21):198-205.
	Bai Xin, Wei Lin. Image retrieval algorithm based on two-level feature extraction and measurement[J]. Packaging Engineering, 2018,39(21):198-205.
[7]	生海迪, 段会川, 孔超. 基于语义短语的空间金字塔词袋模型图像分类方法[J]. 小型微型计算机系统, 2015,36(4):877-881.
	Sheng Haidi, Duan Huichuan, Kong Chao. Spatial pyramid bag-of-words model for image classification based on semantic phrases[J]. Journal of Chinese Computer Systems, 2015,36(4):877-881.
[8]	Boulkenafet Z, Komulainen J, Hadid A. Face antispoofing using speeded-up robust features and fisher vector encoding[J]. IEEE Signal Processing Letters, 2016,24(2):141-145.
[9]	周柱, 甘屹, 孙福佳. 一种融合Gabor+SIFT特征的人脸识别算法[J]. 电子科技, 2019,32(4):1-5.
	Zhou Zhu, Gan Yi, Sun Fujia. A face recognition algorithm based on Gabor+SIFT feature fusion[J]. Electronic Science and Technology, 2019,32(4):1-5.
[10]	袁非牛, 夏雪, 李钢, 等. 面向烟雾识别与纹理分类的Gabor网络[J]. 中国图象图形学报, 2019,24(2):269-281.
	Yuan Feiniu, Xia Xue, Li Gang, et al. Gabor network for smoke recognition and texture classification[J]. Journal of Image and Graphics, 2019,24(2):269-281.
[11]	汤勃, 孔建益, 伍世虔. 机器视觉表面缺陷检测综述[J]. 中国图象图形学报, 2017,22(12):1640-1663.
	Tang Bo, Kong Jianyi, Wu Shiqian. Overview of machine vision surface defect detection[J]. Journal of Image and Graphics, 2017,22(12):1640-1663.
[12]	朱善玮, 李玉惠. 基于SVM和VAR/LBP的车脸识别[J]. 电子科技, 2018,31(7):7-10.
	Zhu Shanwei, Li Yuhui. Vehicle face recognition based on SVM and VAR/LBP[J]. Electronic Science and Technology, 2018,31(7):7-10.
[13]	李志欣, 李艳红, 张灿龙. 一种多特征融合的场景分类方法[J]. 小型微型计算机系统, 2018,39(5):1085-1091.
	Li Zhixin, Li Yanhong, Zhang Canlong. A multi-feature fusion scene classification method[J]. Journal of Chinese Computer Systems, 2018,39(5):1085-1091.
[14]	宋彦, 蒋兵, 戴礼荣. 基于多视觉码本的图像表示[J]. 模式识别与人工智能, 2013,26(10):909-915.
	Song Yan, Jiang Bing, Dai Lirong. Image representation based on multiple visual codebooks[J]. Pattem Recognition and Aitificial Intelligence, 2013,26(10):909-915.
[15]	Boureau Y L, Bach F, Lecun Y, et al. Learning mid-level features for recognition[C].San Francisco:IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2010.
[16]	Bahri A, Zouaki H. A SURF-Color moments for images retrieval based on bag of features[J]. European Journal of Computer Science and Information Technology, 2013,1(1):11-22.
[17]	郭礼华, 罗材. 食材数据库统计与对比实验性能分析[J]. 中国图象图形学报, 2017,22(8):1079-1088.
	Guo Lihua, Luo Cai. Food database statistics and comparative experimental performance analysis[J]. Journal of Image and Graphics, 2017,22(8):2079-1088.

Approachs	人工拍摄图片
Approachs	k=100	k=300	k=500	k=100	k=300	k=500
HSV	64.25	79.36	81.75	39.15	50.47	53.04
ColMom	62.99	71.49	72.01	36.22	41.45	43.52
LBP	49.83	62.95	70.13	32.78	40.23	41.45
Gabor	52.46	55.49	62.15	40.23	48.29	50.15
HOG	62.26	62.48	64.26	42.56	47.29	53.87
SURF	71.09	74.78	76.76	49.82	53.11	53.35

Approachs	人工拍摄图片
Approachs	k=100	k=300	k=500	k=100	k=300	k=500
SURF+HSV	77.24	77.24	78.40	53.26	55.05	56.76
SURF+ColMom	87.01	89.95	89.68	65.64	68.65	68.71
SURF+LBP	76.76	78.74	80.11	53.97	55.76	56.12
SURF+Gabor	69.72	71.23	71.58	50.25	51.13	52.16
SURF+HOG	72.56	74.08	73.92	54.62	55.58	56.19
SURFC	89.21	90.98	91.71	69.79	70.53	72.59

Approachs	人工拍摄图片
Approachs	k=100	k=300	k=500	k=100	k=300	k=500
SURFC+LBP	88.45	89.95	90.37	67.93	68.51	71.58
SURFC+Gabor	80.56	82.13	83.45	59.95	62.89	68.78
SURFC+HOG	75.46	78.86	79.37	58.15	60.58	63.49

Approachs	人工拍摄图片			网络下载图片
Approachs	k=100	k=300	k=500	k=100	k=300	k=500
SURFC+CLBP_S	91.39	92.82	93.64	74.09	76.02	77.74
SURFC+CLBP_M	91.23	93.44	92.89	73.22	75.88	76.95
SURFC+CLBP_M/C	90.98	92.89	93.57	72.73	76.02	75.45
SURFC/CLBP_S_M/C	91.35	91.93	91.87	71.58	76.38	77.67
SURFC+CLBP_SM	91.05	91.87	92.62	73.22	77.38	77.95
SURFC+CLBP_SMC	91.73	93.23	94.48	72.51	76.52	77.74