LOW ALTITUDE REMOTE SENSED (LARS) IMAGERY ANALYSIS USING CONVOLUTIONAL NEURAL NETWORK FOR VEGETATION MAPPING OF A TOMATO CROP

Remote sensing from an Unmanned Aerial Vehicle (UAV), also called as Near Earth remote sensing or Low Altitude Remote sensing is evolving as an interesting option to compliment satellite based remote sensing. This development opens options for agriculture applications. Vegetation mapping is one of the key applications. Vegetation of a tomato crop is carried out in this work. Tomato is chosen since it is a commercial crop. Vegetation mapping is carried out by segmentation performed using a Deep convolutional neural network (CNN). We propose SEG-1DCNN a novel method for vegetation mapping. One dimension deep convolutional neural network architecture is used for classification. The classified gray scale image is thresholded to binary using OTSU thresholding. The performance of SEG-1DCNN is compared with SEG-SVM, a method based on Support Vector Machine (SVM) algorithm. The performance of vegetation mapping is analyzed using parameters based on the confusion matrix. The performance of SEG-1DCNN is comparable with SEG-SVM.