iman zabbah; zahra heydaran; behnaz behzadiyan; roya Rezaie
Abstract
Iran is the largest saffron producer in the world. Saffron is of great economic importance and due to increasing activities of export competitors, it requires support, expansion, and development of exports. The growing trend of saffron export share on one hand and income generation and entrepreneurship ...
Read More
Iran is the largest saffron producer in the world. Saffron is of great economic importance and due to increasing activities of export competitors, it requires support, expansion, and development of exports. The growing trend of saffron export share on one hand and income generation and entrepreneurship for rural residents on the other hand double the necessity of attention to this product. Competition in this supply chain, despite powerful competitors, will be possible through mechanizing processes from cultivation to harvesting and grading to packaging. Therefore, evaluating the physical characteristics of saffron for proper grading is considered essential. In this study, the physical characteristics for the purpose of saffron grading using various artificial intelligence methods including Multilayer Perceptron Neural Networks (MLP), Learning Vector Quantization (LVQ), Self-Organizing Maps (SOM), Fuzzy Neural Networks (FNN), and Adaptive Neuro-Fuzzy Inference System (ANFIS) have been investigated. The database under study relates to 113 saffron samples categorized into 4 classes of Saffron Negin (top quality), Grade 2 Saffron (good), Grade 3 Saffron (normal), and Grade 4 Saffron (poor) collected by the researchers. The analysis results show that saffron grading considering uncertainties in the ANFIS model is superior to other methods, with a classification accuracy of 97.5% and a test sample detection error rate of 0.3484.
Abstract
This study was conducted to achieve effective and low-cost technology for non-destructive grading of unshelled almonds in real time. A laboratory prototype of an intelligent online impact-acoustic system composed of a feeding unit, acoustical recognition unit, and pneumatic separator with an electronic ...
Read More
This study was conducted to achieve effective and low-cost technology for non-destructive grading of unshelled almonds in real time. A laboratory prototype of an intelligent online impact-acoustic system composed of a feeding unit, acoustical recognition unit, and pneumatic separator with an electronic controller unit was constructed and tested. To evaluate system operation according to almond variety and class (hard, semi-soft, and soft), the effect of an acoustic signal generated by dropping the nuts onto a steel plate was captured by microphone and the amplitude, phase, and power spectral density were extracted from analysis of the sound signal in the time and frequency domains by means of fast Fourier transform. A multilayer perceptron neural network with a LM training function were used in all experiments. The classification accuracy using validation data was about 96.2% in the offline mode, but accuracy decreased to 88% in the online mode. This decrease in accuracy was probably the result of a difference in size and mass of the almond samples in the hard and semi-soft classes.