Tomato recognition based on feature analysis and fast development of recognition software

Abstract

In order to guide manipulator to pick tomato,an automatic tomato recognition based on shape and texture features analysis and development method of recognition software are researched.First,child images of tomato and disturbed objects are extracted re-spectively from tomato images captured under outdoor scene.Three shape parameters and five texture parameters are chosen to do stati-stical analyses.The algorithm of tomato recognition is verified quickly under the enviroment of Matlab.And then,OpenCV function library is used to reproduce the recognition process under VC++ development environment to improve real-time of tomato recognition.Fast development of software is realized.By processing of 674 tomato images,the tests show that right recognition ratio of harvesting mature tomatoes can reach more than 90 percent when growth state of tomatoes is seperated each other.By processing of different size of tomato images,the tests show that the implementation efficiency of the codes programmed with OpenCV is risen considerably in com-parison with the codes programmed with Matlab.Morever,the software can capture real-time image and process synchronously binocular image,and may provide a reference for development of fruit-picking robotic vision software.