第4期毛亮等：基于稀疏场水平集的荔枝图像分割算法349

从表中可知，本文算法在计算速度上接近标准模糊C 均值聚类算法，优于水平集方法。因此，与水平集方法相比，本文算法有着更好的分割性能和实时性。

4结 论

该文采用HSV彩色空间中色调H分量的旋转分量为分割特征，通过模糊聚类算法自动获取合适的初始演化曲线轮廓，然后利用水平集方法对目标区域轮廓进行精确的提取。结果表明，该方法很好地保持了目标区域的完整性。对成熟荔枝分割的正确率达到了84.1%。由于农业复杂的环境等因素，该文算法仍存在一定的问题，对遮挡和光照严重不均匀的情况，鲁棒性不高；对荔枝图像中存在多串多果，分割时无法确定主次串等，这些问题都需要在进一步研究。

[参 考 文 献]

[1] D. M. Bulanon, T. F. Burks, V. Alchanatis. Fruit Visibility

Analysis for Robotic Citrus Harvesting [J]. Transactions of the ASABE, 2009, 52(1): 277－283.

[2] Bulanon D. M., and T. Kataoka. Fruit detection system and

an end effector for robotic harvesting of Fuji apples [J]. Agric

Eng Int: CIGR Journal, 2010, 12(1): 203－210.

[3] 宋淑然，洪添胜，王卫星，等. 基于马氏距离的荔枝图像

分割设计方法 [J]. 沈阳农业大学学报，2005，36(6)：655

－658.

Song Shuran, Hong Tiansheng, Wang Weixing, et al.

Computer program based on the mahalanobis distance for image division of lichee pictures [J]. Journal of Shenyang Agricultural University, 2005,36(6)：655－658.(in Chinese with English abstract)

[4] 田锐，郭艳玲. 基于机器视觉的葡萄自动识别技术 [J]. 东

北林业大学学报，2008，36(11)：95－97.

Tian Rui, Guo Yanling. Automatic identification techniques of grape based on computer vision [J].Northeast Forestry University, Journal of Northeast Forestry University. 2008, 36(11): 95－97. (in Chinese with English abstract)

[5] Osher S, Sethian J A. Fronts propagating with curvature

dependent speed: algorithms based on the Hamilton-Jacobi

formulations[J]. Journal of Computational Physics, 1988, 79(1): 12－49.

[6] Adalstein D, Sethian J. A fast level set method for

propagating interfaces[J]. Journal of Computational Physics,

1994, 118(2): 269－277.

[7] Whitaker R. A level-set approach to 3D reconstruction from

range data. International Journal of Computer Vision [J], 1998, 29(3): 203－231.

[8] Chan T, Vese L. Active contours without edges [J]. IEEE

Transactions on Image Processing, 2001, 10(2): 266－277 [9] Wang Xiaofeng, Huang Deshuang, Xua Huan. An efficient

local Chan–Vese model for image segmentation [J]. Pattern

Recognition, 2010, 43(3): 603－618.

[10] Shi Y, Karl W. A fast level set method without solving pdes

[C]//In Proceedings of ICASSP, 2005, 5: 97－100

[11] Shi Yang-gang, Karl W C. Real-time tracking using level set

[C]//IEEE Computer Society Conference on Computer

Vision and Pattern Recognition. Washington, DC, USA: IEEE Computer Society, 2005, 2: 34－41.

[12] Li Chunming, Xu Chenyang, Gui Changfeng, et al. Level set

evolution without re-initialization: a new variational formulation

[C]// IEEE International Conference on Computer Vision and

Pattern Recognition. San Diego: IEEE, 2005, 1: 430－436. [13] Li B N, Chui C K, S.H. Ong, et al. Integrating FCM and level

sets for liver tumor segmentation [C]//In Proceedings of the13th International Conference on Biomedical Engineering,

IFMBE Proceedings, 2009, 23: 202－205.

[14] Li Bingnan, Chee Kong Chui, Stephen Chang, et al.

Integrating spatial fuzzy clustering with level set methods for

automated medical image segmentation[J]. Computers in Biology and Medicine, 2011, 41(1): 1－10.

[15] Smith A R. Color gamut transformation pairs [J]. Computer

Graphics, 1978, 12(3): 12－19.

[16] 鲍吉斌，汪源源，马煌，等. 基于H分量旋转的色盲矫正

方法 [J]. 生物医学工程进展，2008，3(29)：125－130.

Bao Jibin, Wang Yuanyuan, Ma Huang, et al. Color adaptation for color vision deficiency by rotating H component[J].

Progress in Biomedical Engineering, 2008, 3(29): 125－130.

(in Chinese with English abstract)

Litchi image segmentation algorithm based on sparse field level set

Mao Liang, Xue Yueju※, Kong deyun, Liu Guoying, Huang Ke, Lu Qifu, Wang Kai

(College of Engineering South China Agricultural University, Key Lab of Key Technology on Agricultural Machine and Equipment Ministry

of Education, Guangzhou 510642, China)

Abstract: In order to provide picking robot with complete contour of litchi, hue component of HSV color space was selected, and the rotation of hue was used as the feature for image segmentation. Then, fuzzy clustering algorithm was utilized to obtain the appropriate initial evolution curve contour automatically, the sparse field level set method was adopted to extract the target region precisely, and the segmentation regions were labeled. The label image could restore the original image of the segmented regions. Finally, the results show that the proposed algorithm can not only overcome the impact of random noise interference, but also maintain the integrity of the segmentation area, and the correct segmentation rate is up to 84.1%.

Key words: algorithms,image segmentation, fuzzy clustering, litchi, sparse field, level set