The Analysis Of Hyperspectral On Ground And The Design Of Inspecting Software For Rice Insect Pest

Hyperspectral technique on ground is a fast and lossless technique. It can resolve the shortage of wasting time and strength when we take the technique of eyeballing and checking with hands, and can reduce labor intensity of checking people; this technique can combine with secondary planet, and private the information of insect pest in small measure space; this technique checking insect pest is very practicability relative to vision-checking technique of using computer.This article take the insect pest checking of Sogatella furcifera as my object, and dealt with the measurement and analysis the spectral of the level of single leaf and canopy during main harmful time using ASD Field Spec instrument at the field trial rice paddy cultivating between agricultural laboratory of Jiangsu University and plant-protecting department of Zhenjiang in the year of 2006 and 2007. The special checking software was designed to check the number of rice Sogatella furcifera.Through the spectral measurement of the level of single leaf and canopy, spectral reflectivity would take place relevant change under the level of single leaf and canopy when rice were harmed with Sogatella furcifera. Under these two levels, there shape of wave was similar, and the change rule of spectral reflectivity was the same. Through the correlative analyse, it got the sensitive wave band under the level of single leaf: 514-602nm, 697-1339nm, 1501-1749nm and 2101-2299nm, and the diagnostic wave point: 562nm, 774nm and 1691nm; it got the sensitive wave band under the level of canopy: 525-571nm, 711-1339nm and 1501-1749nm, and the diagnostic wave point: 558nm, 779nm and 1685nm.Through the spectral quantitative analyse of rice insect pest of Sogatella furcifera, the wave points of 774nm and 1691nm can diagnose the number of Sogatella furcifera better under the level of single leaf as establishing model using single variable, and the correlative coefficient of predicting model was separately 0.91 and 0.85, and the standard deviation of predicting model was separately 325.4 and 367.3, and relative error of average was separately 16.4% and 18.5%; 12 regression variable were screen out by stepwise regression method as establishing model using many variables, and the correlative coefficient of predicting model was 0.948, and the standard deviation of predicting model was 87.9, and relative error of average was 13.6%. Under the level of canopy, the wave points of 779nm and the RVI made up of 558nm and 779nm can diagnose the number of Sogatella furcifera better as establishing model using single variable, and the correlative coefficient of predicting model was separately 0.921 and 0.806, and the standard deviation of predicting model was separately 306.3 and 334.7, and relative error of average was separately 15.2% and 17.5%; 7 regression variable were screen out by stepwise regression method as establishing model using many variables, and the correlative coefficient of predicting model was 0.982, and the standard deviation of predicting model was 79.3, and relative error of average was 12.7%.The aim of the research is to provide essential testing evidence and technological basis for further development of forecasting rice insect pest of Sogatella furcifera and precise spraying technology.