Study On Monitoring Methods Of Fruit Pests Based On Infrared Sensor And Image Recognition

Now existing pest monitoring methods include acoustic measurement, piezoelectric measurement, infrared measurement and machine vision recognition technology.Traditional monitoring technique for pests in orchard environment have shortages such as weak effectiveness, inaccurate count and poor universal. In view of this, pest detection technology in the future development trend will undoubtedly be an integrated detection method. Comprehensive utilization of the existing testing technology will form a multiple information fusion method to detect pests and provide reliable scientific decision for comprehensive prevention and control of fruit pests to reduce the loss.In this paper, based on the previous studies, the improvement of pest trapping device is provided, which provides a reliable platform for the realization of pest monitoring methods,and realizes the functional requirements of pest counting system. Infrared measurement and machine vision recognition technology was integrated to identify pest species and count pest populations, and pest information was obtained from two above aspects. The linear regression equation was used to establish the fusion formula of the pest count result, and the accuracy of the fusion result was verified by comparing with the manual counting. In this paper, common orchard pests such as Grapholitha molesta, Dichocrocis punctiferalis,Adoxophyes orana were used as target pests, and the results of infrared sensors and machine vision were obtained by randomly randomizing the target pests in the laboratory.The infrared sensor counting unit mainly uses the photoelectric effect to carry on the examination to the pest quantity. Infrared sensor circuit mainly consists of infrared emission tube, infrared receiver tube, operational amplifier, controller, communication module and so on. The unit uses a differential amplifying circuit, extracts a detection signal to be compared with a reference signal, and amplifies the signal to extract a useful signal from the noise. Image processing part: Using MATLAB software for feature extraction, the color feature, texture feature and shape feature of the pest image were used to compose the 20-D dataset. The ant colony algorithm, simulated annealing algorithm and principal component analysis were used to compare the characteristics of the insect pests, and the support vector machine was used to identify and count the pests. Fusion count results were obtained by a formula operation which was derived from the linear regression analysis of SPSS. The result of infrared sensor and machine vision were the input of the formula. Field experiments were carried out in peach orchards to monitor the activity of Grapholitha molesta and to test the pest count method and the effect of equipment operation.