Water Shorting Identification Of Fast-growing Broadleaf Sapling Based On Computer Vision

Water shorting identification of plants promotes the development of intelligent automatic irrigation systems.This study combines image segmentation algorithm with pattern recognition algorithms to analyze whether sapling is lack of water.This paper uses Chinese ash,poplar,willow,ginkgo as the research object.Segmentation algorithm is used to derive a single leaf.Then calculate color feature value,shape feature value,texture feature value of broadleaf sapling leaf image.Combining the soil moisture characteristics,establish plant water shorting identification system by the depth of the BP neural network and SVM learning algorithm training model.This paper compares the segmentation algorithm with classification algorithm respectively,finally establish a system of water shorting identification by Mean-shift segmentation algorithm and the radial basis kernel function SVM classifier.Recognition rate can reach 90.73%.Main work process is as follows:(1)Cultivate eucalyptus,willow,poplar,ginkgo.Establish environmental database.Group each species of sapling,then collect samples of normal growth status and water shorting status every day.At last,establish a corresponding database according to illumination,temperature,soil moisture data.(2)Research image segmentation algorithm.Using a digital camera to collect leaf pictures whose mesophyll and grain full,then choose algorithms for image gray processing,image smoothing,image intensification and edge detection.Choose the most suitable segmentation algorithm for this paper after comparison.(3)Extract the leaf features and select features.This paper identifies whether broadleaf sapling is water shorting depending on characteristic parameters and recognition classifier.Choose color feathers of the mean,variance;Texture feature of the texture energy,contrast,correlation,entropy;Shape feathers of rectangular degree,elongation,circular degree and density.Extract six valuable characteristics by data variance calculation,Then form 7D feature vectors combining with the feature of soil moisture.(4)Study and put forward the BP neural network and SVM method to establish a system of leaf water shorting identification.Input 7D feather vectors to build a BP neural network classification model and a SVM classification model.Choose the most optimal classifier by the recognition rate and training time.