Greenhouse Cucumber Image Processing by Neural Network to use in the Cucumber Harvester Robot
Subject Areas :
Keywords: detection, image processing, Neural network, greenhouse cucumber,
Abstract :
Greenhouse Cucumber Image Processing by Neural Network to use in the Cucumber Harvester Robot Moein Zand Abstract In this study, a method for detection of greenhouse cucumber fruits in greenhouse pictures by neural network approach was proposed. For this purpose, first, after histogram equalization in the desired picture ithasmultipliedwitha windowcontaining apictureof acucumberfruit that caused to create bigger coefficients in areas suspected to be cucumbers with exploitation of this local maximums, the initial points as windows likely to be cucumbers are obtained. Then,in order todetectthe exactlocationofcucumbers, these areas and their surrounding areas are referred to a neural network, the network with a number of pictures included cucumber images, no cucumber images and their favorable response has been Trained. Areas obtained from the network detect exact location of the cucumbers in the image. Provided method, was simulated in MATLAB software and the neural network toolbox. A neural network learning algorithm and a structure for neural network was proposed. Then proposed method utilizes the network structure, has been final simulation and on 120 images taken by a digital camera from a greenhouse were tested.The proposedmethodhadlittleerror (4/7% error of un-detection and 3/13% error of wrong detection) and could by high rate and at the appropriate time (the average for each image 5/12 seconds) to detect cucumber fruits in images. Keywords: Greenhouse Cucumber, Image Processing, Detection, Neural Network
_||_