关键词: 小麦茎蘖数； 冠层图像； 快速诊断, 群体； 智能手机； 无人机

Abstract:  In order to improve the diagnostic efficiency of wheat population,the feasibility of using image recognition instead of manual sampling counting was studied.The number of stems in the four stages of seedling,pre-wintering,reviving and jointing at 105 wheat fields in Yangxin County,Shandong Province were diagnosed from 2016 to 2017 by smart phone,UAV and manual sampling counting,respectively.The results showed that,the correlations of smart phone image recognition and manual sampling counting in the four growth stages were pre-wintering stage(^Ｒ2=0.900，P＜0.001 0)＞jointing stage (Ｒ²=0. 240,P＜0.001 0)＞reviving stage (Ｒ²=0.130，P＜0.001 0)＞seedling stage (Ｒ²=0.010,P＜0.290 0);The correlations of UAV image recognition and manual sampling counting in the three growth stages were pre-wintering stage(Ｒ²=0.760,P＜0.001 0)reviving stage(Ｒ²=0.320,P＜0.010 0)＞seedling stage (Ｒ²=0.005，P＜0.880 0).In terms of diagnostic efficiency,manual sampling counting unit took about 100.0 min/ha,smart phone image recognition unit took about 5.0 min/ha,while UAV image recognition took about 1.5 min/ha.The results show that smart phone image recognition can be used for wheat population size diagnosis in pre-wintering to jointing stage,and the diagnosis accuracy is highest in the pre-wintering  stage.For large-area wheat fields，the method of UAV image recognition can be used for wheat population size diagnosis in pre-winterring and reviving stage．

Key words: Wheat stems number, Canopy image, Ｒapid diagnosis, Population, Smart phone, UAV