植物远程快速识别的方法

Remote rapid recognition method for plants

针对野外植物种类多、自动识别难问题,提出了一种采用云计算和移动终端结合的架构,移动终端对植物拍照后上传至云服务器,云服务器基于已建立的植物数据库并结合位置信息快速识别植物的方法。为克服野外网络信号差导致图片上传时间长的缺点,以及减少网络带宽占用,系统将最近频繁查询条目和邻近区域条目推送至移动终端形成数据库缓存,移动终端可在本地完成比对。采用植物叶子构成小规模实验,测试表明,植物识别平均准确率为93.18%,识别时间小于1.5 s。该方法采用离线训练计算颜色特征值及形状特征值,并结合数据局部性原理,实现了植物快速比对,可满足野外植物识别需求。

To make wild plant recognition more accessible, a novel method combining cloud computing with mobile terminal was proposed. The mobile terminal uploaded a plant photo to the cloud server for recognition based on the established plant library and speeded up the processing with the location information. To solve weak signal problem in the wild field, the system pushed the recent searching items and items in neighboring area to the library buffer therefore reduced the network bandwidth requirement, so that the mobile terminal could finish matching locally. To illustrate the availability, a small scale set of plants was tested and the result shows that the recognizing accuracy is 93.18% and recognition time is 1.5 seconds. The method used off-line training to calculate the color characteristic value and shape characteristic value, and combined with the principle of data locality, it can recognize the plant quickly, which can satisfy the need for field plant recognition.