Detecting mango fruits by using randomized hough transform and backpropagation neural network

• Main Authors: Azim Zaliha, Abd Aziz, Mohd Nordin, Abdul Rahman, Mohd Rizon, Mohamed Juhari, Yahaya, Ibrahim
• Format: Conference or Workshop Item
• Language: English
• Published: 2014
• Subjects: SB Plant culture; T Technology (General)
• Online Access: http://eprints.unisza.edu.my/376/1/FH03-FIK-15-02445.jpg; http://eprints.unisza.edu.my/376/

A new method for mango detection is presented in this paper. This method is based on preprocessing operators on image which includes converting to gray image, finding edges, calculating distances to edges, opening morphology and converting to binary color image. To take advantage of oval shaped mango fruit, we apply Randomized Hough Transform method to detect potential places for mango fruit in input images. By using Back propagation Neural Network, we recognize mango fruits from these potential places. The dataset used to implementing this paper is 50 RGB images captured of mango fruits on trees. As shown in experimental results, in the case of clear fruit in input images, the detection rates up to 96.26% while it decreases in the case of partially covering or overlapping. However, this method can be applied to detect other fruits in varied sizes and colors.