Development Of Magnolia Plant Identification System Based On Deep Learning

Plants are the mainstay of all life on earth.To better protect plants,we must have a more detailed understanding of the plant species.Traditional plant classification methods have higher requirements for equipment and identification personnel.Nowadays,computer technology has developed rapidly.With the help of image processing,deep learning and other technologies,plant classification can be performed more quickly and accurately.Leaf is an important organ that is ubiquitous and easy to observe in higher plants.It is usually selected as a feature carrier in plant classification based on computer technologies such as deep learning.In this paper,the Magnolia leaf data set is constructed.Based on the data set and the YOLOv3 algorithm,the Magnolia leaf detection and recognition algorithm is improved.Firstly,the image data is enhanced by image normalization,horizontal transformation,scaling,rotation,brightness adjustment and other methods.Secondly,according to the characteristics of the leaves of Magnolia plants,the algorithm of YOLOv3 is improved,including increasing the number of anchor boxes to 12,and using k-means++ clustering algorithm to obtain the best width and height of anchor boxes,and obtaining the identification algorithm of Magnolia plants,YOLOv3?T,which includes 4 scales of output detection map.Thirdly,this paper optimizes the loss function of the network model,weights the width and height of the function,uses classification network pre-training,and uses a multi-scale training method during the training phase.Finally,the sample images of 4859 Magnolia leaf test sets were tested,and the results showed that the recognition rate of Magnolia plant identification network designed in this paper can reach 96.61%,compared with 78.39% of the recognition rate of YOLOv3 algorithm,which is greatly improved.Compared with the existing plant leaf identification methods,the Magnolia plant detection and identification model designed in this paper is more targeted to the Magnolia plants,which can greatly improve the accuracy of identification and effectively improve the problems of slow identification and low efficiency of traditional methods.Based on the YOLOv3?T algorithm designed in this paper and Py Qt5,the Magnolia plant identification system was designed and developed,which can realize the rapid and effective detection and identification of Magnolia plants without the intervention of professionals,it can realize the identification of Magnolia plants,so that it can be more close to the daily application without being subject to the traditional instruments and equipment.