针对传统的基于机器视觉的茶叶嫩芽检测方法存在手工特征提取鲁棒性较差以及准确率较低等问题,首次将基于深度学习的目标检测算法YOLO应用到复杂背景下的茶叶嫩芽图像的检测,并从多尺度检测方面对YOLO网络架构进行了改进,用大尺度和中尺度检测代替了原来的多尺度检测.在预处理阶段,通过结合超绿特征以及OSTU算法对复杂背景下的茶叶嫩芽图像进行了图像分割,使得茶叶嫩芽区域更加明显.实验结果表明,通过与其他算法对比,基于深度学习的目标检测算法对复杂背景下的茶叶嫩芽具有较高的检测精度,为复杂背景下茶叶嫩芽的智能化采摘设备的研究提供了基础.
The traditional tea bud detection method based on machine vision has the problems of poor robustness and low accuracy in manual feature extraction. The target detection algorithm YOLO based on deep learning is applied to the detection of tea tender image under complex background for the first time, and the architecture of YOLO network is improved from the aspect of multi-scale detection, using large-scale and mesoscale detection instead of the original multi-scale detection. In the preprocessing stage, the image of tea shoots under complex background is segmented by combining the super green feature and the OSTU algorithm, which makes the tea tender bud area more obvious. The experimental results show that, by comparing with other algorithms, the target detection algorithm based on deep learning has a high detection precision for the tea tender shoots under complex background, which provides a basis for the research of intelligent picking equipment for tea tender shoots under complex background.
[1] 汤一平,王伟羊,朱威,等.基于机器视觉的茶陇识别与采茶机导航方法[J].农业机械学报,2016,47(1):45-50.DOI:10.6041/j.issn.1000-1298.2016.01.007.
[2] 韩余,肖宏儒,秦广明,等.国内外采茶机械发展状况研究[J].中国农机化学报,2014,35(2):20-24.DOI:10.13733/j.jcam.issn.2095-5553.2014.02.006.
[3] 汤一平,韩旺明,胡安国,等.基于机器视觉的乘用式智能采茶机设计与试验[J].农业机械学报,2016,47(7):15-20.DOI:10.6041/j.issn.1000-1298.2016.07.003.
[4] 王财盛.基于机器视觉的采茶机割刀控制方法研究与设计[D].杭州:浙江工业大学,2017.
[5] 吴雪梅,张富贵,吕敬堂.基于图像颜色信息的茶叶嫩叶识别方法研究[J].茶叶科学,2013,33(6):584-589.DOI:10.3969/j.issn.1000-369X.2013.06.018.
[6] 吴雪梅,唐仙,张富贵,等.基于K-means聚类法的茶叶嫩芽识别研究[J].中国农机化学报,2015,36(5):161-164.DOI:10.13733/j.jcam.issn.2095-5553.2015.05.040.
[7] 唐仙,吴雪梅,张富贵,等.基于阈值分割法的茶叶嫩芽识别研究[J].农业装备技术,2013,39(6):10-14.DOI:10.3969/j.issn.1671-6337.2013.06.004.
[8] GIRSHICK R, DONAHUE J, DARRELL T, et al. Region-based convolutional networks for accurate object detection and segmentation[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015,38(1):142-158.DOI:10.1109/TPAMI.2015.2437384.
[9] UIJLINGS J R, SANDE K E, GEVERS T, et al. Selective search for object recognition[J].International Journal of Computer Vision, 2013, 104(2):154-171.DOI:10.1007/s11263-013-0620-5.
[10] HE K, ZHANG X, REN S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 37(9):1904-1916. DOI:10.1007/978-3-319-10578-9_23.
[11] 曹诗雨,刘跃虎,李辛昭.基于Fast R-CNN 的车辆目标检测[J].中国图象图形学报,2017,22(5):671-677.DOI:10.11834/jig.160600.
[12] REN S, HE K, GIRSHICK R, et al. Faster R-CNN:Towards real-time object detection with region proposal networks[J].IEEE Transactions on Pattern Analysis & Machine Intelligence, 2015, 39(6):1137-1149. DOI:10.1109/TPAMI.2016.2577031.
[13] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once:unified, real-time object detection[Z].The IEEE Conference on Computer Vision and Pattern Recognition, Las Vegas, 2016. DOI:10.1109/CVPR.2016.91.
[14] LIU W, ANGUELOV D, ERHAN D, et al. SSD:Single shot multibox detector[M] // Computer Vision - ECCV 2016, Springer International Publishing, 2016:21-37. DOI:10.1007/978-3-319-46448-0_2.
[15] 赵旭江.基于卷积神经网络的遥感图像目标检测与识别[D].合肥:中国科学技术大学,2018.
