人工智能目标识别技术对高原鼠兔定位监测的研究

THE RESEARCH OF USING A RTIFICIAL INTELLIGENCE TARGET RECOGNITION TECHNOLOGY TO LOCATION MONITORING OCHOTONA CURZONIAE

啮齿动物是青藏高原草地生态系统的主要成员之一,鼠害爆发是造成草原退化、载畜量下降的主要根源之一,严重破坏草地生态,然而高原属免监测往往采用人工调查,费时费力,难以满足对害鼠长期、连续地监测需求。本研究利用地面定位观测站手段,构建了高原鼠免智能监测体系,并基于图像识别算法智能化提取样地高原鼠免密度和有效洞口,进行区域鼠害程度分级评价。研究表明,青海省门源县自动站样地监测高原鼠免有效洞口数均值为1536个/hm^(2),鼠免密度均值为741只/hm^(2),危害程度评价为“极度危害”等级,经人工调查验证,有效洞口数绝对误差为7个,有效洞口数相对误差为6.73%,表明人工智能目标识别技术用于高原鼠免定位监测是可行的,节省了人力成本,为高原鼠免监测提供了新技术。

Rodents are one of themainmembers of the grassland ecosystem in the Qinghai-Tibet Plateau.Rat infestation outbreaks are one of themain causes of grassland degradation and livestock capacity decline,which seriously damages the grassland ecology.however,themanual survey method is difficult tomeet the long-term and continuous monitoring requirements of Ochotona curzoniae,In this research,the method of ground positioning observation stations,was used to construct an intelligent monitoring system for Ochotona curzoniae.The number of effective holes and rodent density indicators of Ochotona curzoniae were intelligently extracted based on the image recognitionalgorithm,and the regional rodent infestation degree was graded and evaluated.The results showed that,Theaverage number of effective openings is 1536 perhectare,and the average density of Ochotona curzoniae is 741 per hectare,frommonitored in the sample area ofmenyuan County,Qinghai Province,the degree of damage was evaluated as"extremely harmful".as manual investigation and verification,the absolute error of the effective number of openings is 7,and the relative error of the effective number of openings is 6.73%.This indicates that the artificial intelligence target recognition technology is feasible for Ochotona curzoniae locationalmonitoring,It's saving labor costs,and providinga new idea for Ochotona curzoniae monitoring.