| 摘要: |
| 为实现自然条件下棉花叶片的精准分割，提出一种粒子群（Particle swarm optimization，PSO）优化算法和K-means聚类算法混合的棉花叶片图像分割方法。本算法将棉花叶片图像在RGB颜色空间模式下采用二维卷积滤波进行去噪预处理，并将预处理后的彩色图像从RGB转换到目标与背景差异性最大的Q分量、超G分量、a*分量；随后在K均值聚类的一维数据空间中，利用PSO算法向全局像素解的子空间搜寻，通过迭代搜寻得到全局最优解，确定最佳聚类中心点，改善K均值聚类的收敛效果；最后，对像素进行聚类划分，从而得到棉花叶片分割结果。按照不同天气条件和不同背景采集了1 200幅棉花叶片样本图像，对本研究算法进行测试。试验结果表明：该算法对于晴天、阴天和雨天图像中目标（棉花叶片）分割准确率分别达到92.39%、93.55%、88.09%，总体平均分割精度为91.34%，并与传统K均值算法比较，总体平均分割精度提高了5.41%。分割结果表明，本研究算法能够对3种天气条件（晴天、阴天、雨天）与4种复杂背景（白地膜、黑地膜、秸秆、土壤）特征混合的棉花叶片图像实现准确分割，为棉花叶片的特征提取与病虫害识别等后续处理提供支持。 |
| 关键词: 棉花叶片 复杂背景 天气条件 K均值聚类 粒子群优化（PSO） 图像分割 |
| DOI：10.11841/j.issn.1007-4333.2018.02.12 |
| 投稿时间：2017-03-03 |
| 基金项目:国家自然科学基金项目（31501229） |
|
| Image segmentation method for cotton leaf under complex background and weather conditions |
|
LI Kai1,2, ZHANG Jianhua2, FENG Quan1, KONG Fantao2, HAN Shuqing2, WU Jianzhai2
|
| (1.Mechanical and Electrical Engineering College, Gansu Agricultural University, Lanzhou 730070, China;2.Institute of Agricultural Information/Key Laboratory of Agricultural Information Service Technology of Ministry of Agriculture, Chinese Academy of Agricultural Sciences, Beijing 100081, China) |
| Abstract: |
| In order to realize the accurate segmentation of cotton leaf under natural conditions,a new image segmentation method based on particle swarm optimization (PSO) algorithm and K-means clustering algorithm was proposed.A two-dimensional convolution filter was used to denoise cotton leaf images in RGB color space,and the denoised cotton images were converted from RGB components into Q,super G and a* components with larger difference between cotton and the background.The subspace of solutions to the global pixel was calculated in one dimension data space of K means clustering by using the PSO algorithm later.The global optimal solution was acquired by iterative search.The optimal clustering center was determined.The convergence effect of K means clustering was improved by this method.Finally,pixels were classified into different clusters and the leaf area was segmented.Considering the effects of different weather conditions and different backgrounds on imaging,1 200 images of cotton leaves under various imaging conditions were captured.The segmentation performance of the proposed algorithm was investigated based on these images.The experimental results showed that the cotton image segmentation accuracies of the algorithm under sunny,cloudy and raining conditions were 92.39%,93.55% and 88.09%,respectively.The overall average segmentation accuracy was 91.34%.Compared with the traditional K means algorithm,the overall average segmentation accuracy of the proposed algorithm was improved by 5.41%.The results showed that the proposed algorithm accurately segmented the cotton leaf images,which combined with 3 weather conditions (sunny day,cloudy day and rainy day) and 4 complex background features (white mulch film,black mulch film,straw,soil).The proposed algorithm could be applied to feature extraction and plant diseases and pests identification. |
| Key words: cotton leaf complex background weather condition K means clustering particle swarm optimization image segmentation |
