A Multi-Strategy Vision Accurate Inspection Algorithm for Storage Pests

The paper proposes a multi-strategy vision accurate inspection algorithm for storage pests, that is singular value decomposition (SVD) adaptive adjust noise variance to reduce spectral reflectance and shadows influence so as to significantly enhance storage pests' information. And on this basis the paper uses wavelet transform to enhance images for achieve the goal in both sketching out approximately contours and showing details of storage pest. Secondly, the paper uses OTSU image binary method to detect storage pests (tiny targets) to improve the vision detection accuracy. Finally, according to the area and perimeter parameters of the connected domain in the storage pest image, the regional location and quantity detection of storage pests are obtained. Blob algorithm can extract these connected domains. The experimental results show that the algorithm can not only accurately detect the number of storage pests, but also mark the location of storage pests, which is of great significance to improve the real-time detection technology of storage pests and realize food security management.