Design Of Detection System For Wheat Leaf Unfoldment Based On Mask R-CNN

Wheat is a cereal crop widely cultivated around the world,and its caryopsis is one of the staple foods of human beings.According to statistics,wheat provides more than 20% of the world’s total protein and heat to the human body,occupying an extremely important position in human production and life.Leaf unfolding marks the beginning of a new growth stage for wheat.Through its detection can help producers choose more high-quality wheat varieties,thereby improving wheat yield and quality.In the traditional wheat seed selection process,the observation of leaf spread is mainly completed manually.Manual observation consumes a large amount of manpower and time,while the accuracy of observation is relatively low,which is easy to cause significant errors.With the development of computer vision,object detection and instance segmentation technologies have brought about a new round of transformation,which can more efficiently and accurately complete related observations and records than traditional manual methods.Using a target detection and instance segmentation model to detect wheat leaves,it is possible to automatically determine the current unfolding situation of wheat leaves,thereby grasping the growth status of wheat in a timely manner.This article aims to improve the Mask R-CNN algorithm for target detection and instance segmentation of unfolded leaves and their creases in wheat leaf images,enabling users to easily and intuitively observe the unfolded leaves,while further understanding the unfolding situation of the leaves through the leaf creases.In addition,the visualization system can detect images uploaded by users or capture camera images for real-time automatic detection.The main research content of this article is as follows:(1)Create a wheat leaf dataset WLD.Use a camera to collect wheat leaf data,and perform related operations to generate expanded data to obtain the original dataset.Classify and label the expanded leaves in the original data set along the contour,and then extract the leaf subgraph using image operations,marking the creased parts in the subgraph to obtain the expanded leaf data set.(2)Design a wheat leaf unfolding instance segmentation algorithm based on improved Mask R-CNN.The Mask R-CNN algorithm is used to train the marking data of the unfolded blade and the marking data of the crease in the blade,and then the overall algorithm is improved by replacing the backbone network to achieve optimization of the detection effect.(3)Develop a wheat leaf unfolding detection system based on Flask+Vue.Using the Flask backend framework,Vue front-end framework,and My SQL database development system,users can upload wheat leaf images to view the image and confidence of the unfolded leaves and their creases after detection and segmentation.They can also query and delete historical detection records.In addition,the batch image detection function can achieve batch detection of multiple images,and the real-time video detection function can automatically detect the images collected by the camera according to frequency,and inform users of the detection situation through email and SMS.