Research On Machine Vision Recognition Of Three-dimensional Features Of American White Moth

Accurate and timely identification of pests is very important in agricultural development.The pest identification method based on deep learning is robust,but at present,the two-dimensional image is mainly used for sample expansion,and the diversity of training samples is limited,which affects the recognition accuracy.There is not much research on insect 3D posture information at home and abroad,and there are fewer pest identification and classification for deep learning,and there are few researches on the sample database based on3 D pest data.Therefore,it is particularly important to study the extraction and recognition technology of insect three-dimensional posture features,which can realize the amplification of moth pest protection feature samples and provide training sample guarantee for deep learning models.Based on the completion of the three-dimensional reconstruction of moth pests with interspecies characteristic differences,this paper studies the three-dimensional posture information acquisition method of the key parts of moth pests,and provides an information source for the subsequent construction of the three-dimensional deformation calculation method of moth pest preservation characteristic sample amplification.Taking the adult of the Hyphantria cunea as the research object,we first find the key parts that can represent the three-dimensional posture information of the moth pest adult,and then establish a three-dimensional posture information extraction method of the moth pest based on the principle of stereo vision.The Hyphantria cunea is divided into wings and trunk,and three-dimensional posture feature extraction is performed respectively.The specific method takes 90° as the difference,take 4 photos around each Hyphantria cunea,preprocess the image,extract the edge points,the wings are first edge-fitted to find the intersection point,the key points are coarsely located,and then the key points are accurately positioned and integrated After matching,the three-dimensional coordinates of each key point of the Hyphantria cunea wings are finally obtained,and the three-dimensional posture information is calculated;the head and tail of the trunk are first positioned,and then the equation fitted by the belly edge is calculated to represent its three-dimensional posture information.Use the laser measurement method to obtain the reference value and compare it with the experimental measurement value according to a certain method to verify the reliability and accuracy of this method.Experimental results show that this method can quickly and accurately obtain three-dimensional posture information of moth pests.Finally,based on this method,a software system for three-dimensional gesture information recognition of Hyphantria cuneawas designed.In this paper,a set of three-dimensional posture information extraction methods of moth pests based on the principle of stereo vision is established to quantify the three-dimensional posture information of moth pests.In the field of insect three-dimensional research,avoiding the construction of complex three-dimensional models of insect bodies provides new ideas for the identification and classification of multi-posture pests in the future,and promotes the application of machine vision and deep learning in pest identification.