Pest grain grade judgment method and device based on visual saliency

Abstract

The invention provides a pest grain grade judgment method and device based on visual saliency. The method comprises the steps of obtaining a current to-be-detected image; inputting the current to-be-detected image into a pre-trained grain pest saliency detection and quantity estimation model, and outputting a visual saliency map and a pest quantity of the current to-be-detected image, the grain pest saliency detection and quantity estimation model comprising two branches, the saliency detection branch comprising a color space enhancement module, a backbone network module, a cascaded hole convolution module and a feature aggregation module, and the number estimation branch being additionally provided with an adaptive average pooling layer and a full connection layer behind the backbone network; and converting the number of pests into a standard pest density, and determining the pest grain grade of the current to-be-detected image. In this way, the saliency detection and quantity estimation precision of grain pests and other small targets is improved, whether pests exist in grain or not and the quantity of the pests can be rapidly determined, and then the pest grain grade of the to-be-detected grain can be judged.

Description

technical field [0001] The invention relates to the technical field of image processing, and more particularly, to a method and device for judging insect food grades based on visual salience. Background technique [0002] Grain is the main source of food for human beings and most domestic animals, and the post-harvest loss of grain due to gnawing, excretion, reproduction and other activities of stored grain pests accounts for about 10% of the total grain output every year. In order to take appropriate measures to reduce the loss of grain storage, it is necessary to detect pests in grain early in the process of grain storage and production and processing, and to detect the density of grain pests and determine the level of insect grain. [0003] The national standard "Technical Specifications for Grain and Oil Storage" pointed out that the main pests in grain include ten kinds of corn weevil, rice weevil, grain beetle, large grain robber, wheat moth, and Indian grain moth. Ac...

AI Technical Summary

Problems solved by technology

However, due to the large variety of grain insects, small size, and complex morphological structure, and the influence of stored grains due to varieties, grades, impurities, and incompleteness, most of the existing methods are designed for a certain certain grain insects in a specific experimental environment. However, it is difficult to be extended to the practical application of grain depots with complex environments, so it is necessary to improve the robustness and adaptability of the grain insect visual detection method

[0006]Because the target of stored grain pests is small and the background of grain images is complex and diverse, the detection accuracy of existing methods for small-scale targets is low, which is not conducive to the identified pests

Images