Rapid Detection Method Of Multidimensional Information Fusion For Nitrogen, Phosphorus, Potassium Content Of Tomato Leaves Based On Interaction

Due to irrational use of chemicals in agricultural facilities, the environment is polluted and the economic profits are seriously affected. In recent decades, nondestructive testing technology has become one of the important mean of nutrition diagnosis for guiding gardening fertilization. So, it is urgent required that the nutrients of crop growth were accurately monitored and diagnosed to achieve precise management. This study combined the advantages of hyperspectral with polarization spectral technology, which can be analyzed the change of color, texture, shape, microstructure and other characteristics on plant nutrient deficiency. Based on the study of the interaction of nitrogen, phosphorus and potassium in tomato, information fusion was utilized to analyze the N, P, K content of tomato leaves. Moreover, according to the N, P, K content in leaves, crop fertilization levels were inverted to provide nutrients stress information. The main contents of this study were as follows:Tomato plant’s sample were prepared with independent and interactive deficiencies of N, P, K. N, P, K element contents were tested adopting Kjeldahl mettiod, spectrophotometry, flame photometric analysis respectively.The changes of reflectance on characteristic wavelengths were taken into consideration to assess the nutritional status of crops. After the first order differential transformation, sensitive bands were selected by using uninformative variables elimination and Successive Projections Algorithm. Two common sensitive wavelengths were identified of 566.29 nm, 693.71 nm, and N, P, K specific sensitive wavelengths corresponding sequence is: 724.66 nm, 474.85 nm, 762.24 nm respectively. Texture features were extracted from images under sensitive wavelength of the first principal component based on GLCM. Though correlation analysis to find features showed high correlation with N, P, K, that were: nitrogen MEAN、HOM、ENT、COR; phosphorus HOM、CON、DIS、COR; potassium HOM、ENT、DIS、ASM. The polarization degree and polarization angle features were extracted by Stocks method. The genetic and partial least squares method was used to select characteristics that the common sensitive wavelength with N,P,K degree of polarization features, that were 549.01 nm, and N,P,K specific sensitive wavelengths that were 398.43, 549.01 nm; 740.54, 857.91nm;400.27, 695.66 nm respectively. The N,P,K angle of polarization features were extracted with common sensitive wavelength, that were 745.35 nm, and N,P,K specific sensitive wavelengths that were 724.49, 856.61nm;567.01, 838.91nm;367.58, 863.51 nm respectively.GA-BPANN and LS-SVM methods were using to establish the Multi information fusion models of N, P, K contents. The results showing that, the effect of LS-SVM model is better. The determination coefficients of N, P, K elements were 0.9196, 0.8396 and 0.8892. The root mean square errors were 0.1525%, 0.2491mg/g and 0.1962mg/g. Furthermore, LS-SVM has the best performance towards the interaction samples prediction. Multi-factor quadratic polynomial regression could be used to modify the model and the determination coefficients of N, P, K elements were 0.9230, 0.8468, 0.8954 respectively, SEE of N, P, K elements were 0.2161, 0.3617, 0.5180 respectively after modification.Through the interaction of gray correlation analysis and correlation analysis, it was concluded that P N, K N, K P and N K were Synergistic effect, K P and P K were antagonistic effect. The relationship of nitrogen, phosphorus and potassium content in each were analyzed by fuzzy clustering.Influence of N, P, K contacts in Leaves and fertilizer state were explored to estimate level of fertilization. Then fuzzy pattern recognition model were established and verified and verified to classify validate interaction samples, which N, P, K contacts were predicted. The results showing that, flowering stage, initial fruiting stage, mid-fruiting stage and picking stage of identification rate were 92.11% and 97.37% and 89.47%, 92.11% respectively.The multi-information fusion detection on tomato nitrogen, phosphorus and potassium nutrition diagnosis was feasible and conducive to high accuracy and fast detection, thus provides basis to methods about crop nutrients for the development of fast and accurate diagnostic instrument with important academic value and application prospect.