Insects Image Intelligent Identification Technology Research

Under the support of Major projects of The Chinese Academy of Agricultural Science and Beijing Entry-Exit Inspection and Quarantine Bureau, this paper mainly researches in insect images' overall outline feature and other insects' wing-context characteristic extraction and recognition methods. This paper puts forward a tiny predecessors inertia method for image and a kind of extraction grayscale image. The corrective numerical calculation method and achieves nearly 30% of the symmetric axis inclination accuracy than the original precise, and provides image processing and analysis of further ready to more accurate preparation work. It puts forward several insect image contour perimeter algorithm improved methods, including diagonal improvement of numerical accuracy, curve approximate re-description, edge pixels eliminate etc, at the expense of small complexity conditions, the more precise contour of the circumference that insects image reasonable values. Also it puts forward a kind of insect wing-context image grid identification method, this method for images' segmentation, for different regional different weights set order to achieve the best result. Against insects wing-context image recognition process line position relatively constant particularity, and put forward a indisputable theory value range search comparison algorithm, which turns between time narrowed nearly n times for the selected area (n wing-context total number), thus greatly enhancing the line sequential search comparative processing speed. It comes with new insect image characteristic parameters, including islands number, and the island normalized center distance, islands area accounts for insects'image of the total area of proportion and so on, which greatly improved the accuracy of insect image recognition, about increased by nearly eight percent.