杜鹃花各生长期识别与监测研究
摘要：
针对传统BoF算法缺乏空间信息的问题,文中提出一种改进式BoF算法,并将其应用于杜鹃花各生长期识别与病虫害监测问题。该算法在基于LAB的颜色特征中融入有序的空间信息,形成了新的空间颜色聚合特征来代替传统颜色直方图,有效解决了颜色特征变化尺度小的问题。该算法提取SURF特征代替原有的SIFT特征,通过一种多类特征学习算法融合颜色特征和SURF特征实现图像分类,并通过进一步分析叶片特征来快速识别杜鹃花植株的生长期与病害。经过仿真得知,基于LAB的颜色聚合向量的改进式BoF模型识别率达到了90.6%,较传统颜色直方图的图像分类方法图像检索速度增加3倍,更容易实现特征融合。
关键词: 改进式BoF算法, 空间颜色聚合特征, SURF, LAB, 多类特征学习, 叶片特征, 特征融合
Abstract:
In view of the problem that the traditional BoF algorithm lacks spatial information, an improved BoF algorithm is proposed and applied to the identification of various growth stages and the monitoring of pest and diseases of Rhododendron in this paper. The ordered spatial information is integrated into the LAB-based color features by this algorithm to form a new spatial color aggregation feature to replace the traditional color histogram, which effectively solved the problem of the small scale of color feature change. Besides, SURF features are extracted to replace the original SIFT features. The image classification is realized by a multi-class features, and the leaf features are further analyzed to quickly identify the growth period and disease of Rhododendron plants. Simulation results show that the recognition rate of the improved BoF model of the LAB-based color aggregation vector is 90.6%. Compared with the image classification method of the traditional histogram, the image retrieval speed is increased by 3 times, making it easier to implement the feature fusion.
Key words: improved BoF algorithm, spatial color aggregation feature, SURF, LAB, multi-class feature learning, leaf features, feature fusion
中图分类号:
TP391.4相关知识
农业害虫自动识别与监测技术（资料）.doc
杜鹃花属4种植物花粉形态特点研究
中国茶叶：人工智能识别茶树病虫害的应用与展望
兰花病毒病的识别与综合防治策略
天津海岸带入侵植物互花米草的遥感监测与分析
基于ResNet50的植物病害识别研究与系统应用实现
基于图像的植物病害识别与分类
江苏首台病虫智能监测系统在南通通州试运行 实现小麦赤霉病自动监测
基于图像处理的花卉识别技术的研究与实现
昆虫的分类与形态学的研究方法与应用技术.pptx
网址: 杜鹃花各生长期识别与监测研究 https://m.huajiangbk.com/newsview131176.html
| 上一篇: 峡谷晚报：EDG夺冠入选2021 |
下一篇: 无人机遥感监测作物病虫害胁迫方法 |
