Loose Fruit Recognition System With Implementation Of SURF Feature Extraction Method

Abstract

Oil extraction rate (OER) is one of the important part in increasing the global oil production. Loose fruit (palm oil) has the highest free fatty acid (FFA) compare to the fruit in the bunches. If 1 % of loose fruit are left uncollected during the harvest estimated 80,000 tons of crude palm oil is lost. In this paper, we propose a system which can recognized loose fruit which can help the time when collecting loose fruit. This system introduces the Speed Up Robust Features (SURF) method for the feature extraction. From the evaluation of the results, we observe that the accuracy of recognition system when implementing SURF method depend on the minimum match number of key-points matching which need to be at the optimum number. This will lead to reducing the false match also keep the recognition successful. From the evaluation results, we identified that the SURF method was compatible with 7 minimum features match when recognizing the loose fruit under this experimental condition.