Monitorin Model Research Of Greenhouse Muskmelon Water Content Based On Multi-Information Technology Fusion

The production management of the modern greenhouse is mainly depended on the surrounding environment information of the plant,and there was a lag and also could not achieve precise irrigation.So real-time,fast,accurate and effective monitoring method is one of the keys to solve the precision management of muskmelon cultivation.According to the water condition of the muskmelon plant,the water utilization will be effectively improved to improve the quality of plants and fruits.Accurate monitoring means to be more efficient,which is an urgent need to promote the development and expansion of the muskmelon industry.This study using two varieties muskmelon named"Wanglu"and"Arus"as the experimental material,with muskmelon plant water status monitoring models been established for the purpose,through different crop tests under different water treatments,studied the two varieties muskmelon of different development stages of the muskmelon water status monitoring models.The simulation results are verified by independent experimental data.The specific conclusions are as follows:1.In 50⁶0%of seedling stage,55⁷0%of extensor period and 55⁶5%of the fruits period?relative water content?,two varieties of muskmelon were taller than T1 and T2 in height by 19.59%and 14.93%;The dry and fresh weight was 64.29%,45.71%,24.34%,55.32%and 31.67%,19.53%higher than other treatments;The water content of the leaves was 22.35%,8.55%and 2.30%,higher than other treatments;the contents of the stems,chlorophyll and photosynthesis were better than other treatments,and the plant growth was the highest.2.The NDVI_610nm,805nm and RI_641nm,998nm of the first derivative of the canopy blade were constructed,and the determination coefficients were0.528 and 0.491 respectively.The NDVI_680nm,734nm,RI_680nm,750nm and the first derivative of the first derivative of the upper blade of the muskmelon were constructed in the first derivative of the first derivative of the first derivative of the upper blade of the muskmelon.The model of RI_603nm,758nm03nm,758nm was constructed by the first derivative of the first derivative,and the coefficient was 0.746,0.743 and 0.707 respectively.Using independent test data to test the model,the r² of model test and the relative mean square error were 0.667,0.660,0.600 and 1.409%,1.629%and 1.923%respectively.3.Of the largest contribution to the first five indicators by Radom forest at different development stages,simplified prediction model,the growth period of the prediction model can remain more than 0.75 of the prediction accuracy,the seedling stage was 0.776,and tendril period was0.944,fruit growth of 0.900.Through higher contribution index monitoring leaf water content of the neural network model is set up and use the independent test data on the model test,model test precision relative root means square error of 1.62%,0.79%and 0.66%respectively.4.Select the three groups of characteristic bands with the most sensitive moisture response of the plant canopy,610nm and 805nm,680nm,743nm and 750nm,1040nm vegetation index NDVI,RI and three growth periods for the visible and near-infrared images obtained from the visible and near-infrared images as input of multi-information fusion.The correlation coefficients of the measured values and predicted values were0.710,0.857 and 0.939 respectively.MRE was 1.27%,0.74%and 0.07%respectively;RRMSE was 1.45%,0.97%and 1.58%respectively.The results show that the irrigation threshold of each growth stage of the two varieties muskmelon of greenhouses was determined;Combining spectrum monitoring and phenotypic monitoring could improve precision of the monitor model that based on the classification of matrix water content and determined their own water conditions,guided irrigation provides the basis for later production,to achieve fine management of different growth stages of moisture in order to provide the basis for guiding irrigation for later production and realized the fine management of water content in different growth stages.This paper provides a basis for the establishment of the melon irrigation management decision model with multi-information as input quantity.