| 叶面雾滴沉积分布直接检测算法研究 |
| |
| 引用本文： | 李睿,张伟,王訢丞,符海霸,于丽娟,张平.叶面雾滴沉积分布直接检测算法研究[J].中国农机化学报,2020(3):74-78. |
| |
| 作者姓名： | 李睿 张伟 王訢丞 符海霸 于丽娟 张平 |
| |
| 作者单位： | 黑龙江八一农垦大学电气与信息学院;黑龙江八一农垦大学动物科技学院 |
| |
| 基金项目： | 海南省自然科学基金面上项目(519MS097);黑龙江省大学生创新创业训练计划项目(201910223016);黑龙江八一农垦大学校内培育课题(XZR2017-16)。 |
| |
| 摘 要： | 提出一种基于阈值边缘提取算法和HSV颜色模型的二次分割叶片雾滴图像识别算法。通过模拟喷雾试验得到三种不同雾滴密度的叶片样本,保留叶片自身轮廓信息的同时分割叶面雾滴,计算叶面积与雾滴覆盖率关系。结果表明应用Otsu阈值边缘提取与HSV空间混合算法相对于传统的k-means聚类分割算法、动态阈值分割算法,更适用于叶面雾滴分布的识别与检测,三种覆盖密度叶片的分割准确率分别为:95.16%、94.23%、93.76%,平均准确率为94.38%;雾滴覆盖率检测相对误差分别为:2.82%、4.11%、7.59%,平均相对误差分别为4.84%。基于阈值边缘提取与HSV空间提取的混合算法可分割叶面雾滴图像并检测完整叶面上雾滴覆盖率,识别结果能够满足识别精度的要求。
|
| 关 键 词： | 非接触 叶面 图像分割 沉积分布 检测算法 覆盖率 |
Study on direct identification algorithm of droplet deposition on leaf surface |
| |
| Li Rui,Zhang Wei,Wang Xincheng,Fu Haiba,Yu Lijuan,Zhang Ping.Study on direct identification algorithm of droplet deposition on leaf surface[J].Chinese Agricultural Mechanization,2020(3):74-78. |
| |
| Authors: | Li Rui Zhang Wei Wang Xincheng Fu Haiba Yu Lijuan Zhang Ping |
| |
| Institution: | (College of Electrical and Information Science,Heilongjiang BaYi Agricultural University,Daqing,163319,China;College of Animal Science and Technology,Heilongjiang BaYi Agricultural University,Daqing,163319,China) |
| |
| Abstract: | In order to detect the droplet distribution on the surface of crop leaves, the method of non-contact droplet recognition and segmentation based on machine vision was explored. An image recognition algorithm based on threshold edge extraction algorithm and HSV color model is proposed. Three kinds of leaf samples with different droplet densities were obtained by simulated spray test. The relationship between leaf area and droplet coverage was calculated by dividing fog droplets on the leaves while retaining the contour information of the leaves themselves. The results show that Otsu threshold edge extraction and HSV spatial hybrid algorithm are more suitable than k-means clustering segmentation algorithm and dynamic threshold segmentation algorithm for the identification and detection of leaf fog droplet distribution. The segmentation accuracy of the three coverage density leaves was 95.16%,94.23% and 93.76% respectively, with an average accuracy of 94.38%. The relative errors of fog drop coverage rate are 2.82%,4.11% and 7.59% respectively, and the average relative errors are 4.84%. The mixed algorithm based on threshold edge extraction and HSV spatial extraction can segment the fog-drop images on the leaves and detect the fog-drop coverage on the intact leaves, so the recognition results can meet the requirements of recognition accuracy. |
| |
| Keywords: | non-contact foliage image segmentation sedimentary distribution detection algorithm coverage |
| 本文献已被 CNKI 维普 等数据库收录！ |
|
