Non-destructive Detection Of Potato Early Blight Disease Based On Hyperspectral Imaging Technology

As the fourth largest food crop after wheat, rice and corn of the world, the potato occupies an important position in agricultural production in our country or even the world. However, potato early blight is hindering the development of potato industry in China, as a very common plant disease in the field. The traditional detection methods of crop disease are difficult to recognize the disease degree and predict the physical and chemical index of potato with rapidity and accuracy. So to explore an accurate and fast method of disease detection is of great significance to the development of potato industry.The detection of potato early blight was studied in this paper based on hyperspectral imaging technology. The spectral information of all samples’ region of interests was extracted by ENVI. This research conducted the early and classification recognition of potato early disease of leaves, and established the prediction model of chlorophyll content(SPAD value) under the early blight duress, combining with different chemical metrology methods. And the rapid detection device of potato early blight based on the LED with particular wavelength was designed and developed. The main conclusions are as follows:(1) Early recognition models were established by the BP neural network(BPNN) and least squares support vector machines(LS-SVM) algorithm, obtaining the recognition rate of 100%, respectively. And to extract the characteristic wavelengths using the methods of successive projections algorithm(SPA) and x-Loading Weights(x-LW), they were 720, 766 nm and 710, 753, 769 nm, respectively. Meanwhile, classification recognition models were established by the LS-SVM algorithm under the different methods of spectral pretreatment. The LS-SVM model with Multiplicative scatter correction(MSC) pretreatment is optimal, with the recognition rate of 97.54% and 93.55% for calibration set and prediction set. The results showed that it is feasible for the early detection and classification detection of potato early blight using hyperspectral imaging.(2) The PLSR models of SPAD value were established under the stress of potato early blight by the hyperspectral imaging technology, with different pretreatment methods of full spectrum data. The optimal pretreatment method was the Normalize pretreatment by comparison. Afterwards, the characteristic wavelengths were extracted using the method of SPA after the Normalize pretreatment of full spectrum data, and they were 452, 536, 710, 728, 960 nm, respectively. And the prediction models were builded for PLSR, MLR, BPNN and LS-SVM. The results showed that the optimal model was the Normalize-SPA-LS-SVM, with the correlation coefficient of prediction(Rp) of 0.9499 and the root mean square error of prediction(RMSEP) of 2.866.(3) The fast and non-destructive testing device of potato early blight was designed and developed based on the LED with particular wavelength. The hardware design included the optical system and photoelectric detection circuit. The software design had the main program design, A/D conversion program design, liquid crystal display program design, alarm circuit program design and so on. With the hardware circuit being tested, through continuous modification and adjustment, it could realize the rapid detection of potato early blight. However, the rate of overall recognition of device was low, and precision needed to be improved, and the function also needed to be perfect.