A method of fruit picking robot target identification based on machine vision

Abstract

In order to solve the problem of fruit picking robot target identification,the paper put forward a novel method based on machine vision.Firstly,the fruits were segmented from background by the algorithm based on normalized color difference.Secondly,the fruit edges were extracted by the single pixel tracking algorithm and during tracking the corners were deleted by linear projection method.The crossed edges were broken off at the corner.It improved the accuracy of recognition in the case of occlusion.Lastly,circles were detected by least square method to identify fruits.The experimental results showed that it take 83.548 ms to process a medium size image averagely.Its accuracy is larger than 95%.The algorithm is fast and robust.It meets the real time requirement of picking robot.