An Automatic Identification System of Agricultural Pests Based on Machine Vision

摘要

An automatic identification system of agricultural pests based on machine vision, image processing and pattern identification was presented. First the hardware system was completed and pests' image which posture helped processing and identifying was obtained. Second image preprocessing was made and total 12 features including color, area, perimeter, fractal dimension and ridgelet edge moment were extracted. Third feature selection was implemented to select 7 features. Finally BP neural network classifier was designed to classify and identify pests. For usual 8 species of pests recognition rate was up to 85.7%. Results show that the system is automatic and high recognition rate and meet the demand of real application.