Field Broccoli Head Recognition Technology Based on Laws and Gabor Filter

被引：0
作者：
Zhao X. [1 ,2 ]
Xu G. [1 ]
Chen J. [1 ]
Yu G. [1 ]
Dai L. [1 ,2 ]
机构：
[1] Faculty of Mechanical Engineering, Zhejiang Sci-Tech University, Hangzhou
[2] Key Laboratory of Transplanting Equipment and Technology of Zhejiang Province, Hangzhou
来源：
Nongye Jixie Xuebao/Transactions of the Chinese Society for Agricultural Machinery | 2023年 / 54卷 / 04期
关键词：
broccoli; Gabor filtering; image recognition; Laws filtering; texture enhancement;
D O I：
10.6041/j.issn.1000-1298.2023.04.032
中图分类号：
学科分类号：
摘要：
Correctly identifying the field location of broccoli is the basis for realizing automatic harvesting of broccoli. Because the flower ball color is similar to the plant stem, broccoli cannot be identified only by color features. The algorithm firstly strengthened the boundary texture of the image through pretreatment and Laws filter, in which the filter kernel function of Laws adopted E5 xEs. Then Gabor filter was applied to the texture enhanced image, and Gabor transform which was a short-time window Fourier transform proposed to meet the locality of two dimensional images in spatial and frequency domain, with window function of Gaussian function. Through Gabor filter, each pixel had a 1 X 8 dimensions texture feature vector, which was generated by eight different Gabor filtering kernel functions that were determined by the wavelengths of one sinusoidal modulation wave and the directions of eight different kernel functions. The texture feature vector was zero-mean normalization to speed up the convergence of clustering process, and K-means clustering segmentation and open operation were performed to obtain the potential region of broccoli heads. Meanwhile, the image was segmented based on color features. Through converting RGB (red, green, blue) image into HSV (Hue, Saturation, Value) image, the Hue component of the image was threshold to filter out ground pixels. Finally, the results of texture feature recognition and color feature recognition were fused to realize the recognition of field broccoli heads. A total of 792 images were used for the experiment. The experimental results showed that this method could accurately identify the broccoli field images. The precision rate was 96. 96%, the recall rate was 94. 41%, and the Fl score was 95. 67%. Through the algorithm recognition of three sets of different shooting environment data sets, the Fl score of the three sets of data sets was always maintained at more than 94%, which had good shooting environment adaptability and laid a foundation for automatic harvesting of broccoli by agricultural robots. © 2023 Chinese Society of Agricultural Machinery. All rights reserved.
引用
收藏
页码：313 / 322
页数：9
相关论文
共 32 条
  • [1] CHENG Xiangyun, SONG Xin, Review on vision system of fruit and vegetable picking robot, Journal of Zhejiang Agricultural Science, 60, 3, pp. 490-493, (2019)
  • [2] ZHOU Jun, CHENG Jiayu, Moving obstacle detection based on machine vision for agricultural mobile robot, Transactions of the Chinese Society for Agricultural Machinery, 42, 8, pp. 154-158, (2011)
  • [3] JIMENEZ A R, CERES R, PONS J L., A survey of computer vision methods for locating fruit on trees, Transactions of the ASAE, 43, 6, pp. 1911-1920, (2000)
  • [4] GAO Fangfang, WU Zhenchao, SUO Rui, Et al., Apple detection and counting using real-time video based on deep learning and object tracking, Transactions of the CSAE, 37, 21, pp. 217-224, (2021)
  • [5] BI Song, GAO Feng, CHEN Junwen, Et al., Detection method of citrus based on deep convolution neural network [J], Transactions of the Chinese Society for Agricultural Machinery, 50, 5, pp. 181-186, (2019)
  • [6] LIU Fang, LIU Yukun, LIN Sen, Et al., Fast recognition method of tomato fruit in complex environment based on improved YOLO [J], Transactions of the Chinese Society for Agricultural Machinery, 51, 6, pp. 229-237, (2020)
  • [7] ZHAO Dean, WU Rendi, LIU Xiaoyang, Et al., Robot apple picking location in complex background based on YOLO deep convolution neural network[J], Transactions of the CSAE, 35, 3, pp. 164-173, (2019)
  • [8] WANG Dandan, SONG Huaibo, HE Dongjian, Research progress of apple picking robot vision system, Transactions of the CSAE, 33, 10, pp. 59-69, (2017)
  • [9] CHEN Yan, WANG Jiasheng, ZENG Zeqin, Et al., Visual pre positioning method of litchi picking robot in large field of view, Transactions of the CSAE, 35, 23, pp. 48-54, (2019)
  • [10] MU Longtao, GAO Zongbin, CUI Yongjie, Et al., Kiwifruit detection of far-view and occluded fruit based on improved AlexNet [J], Transactions of the Chinese Society for Agricultural Machinery, 50, 10, pp. 24-34, (2019)
1234