A methodology for fresh tomato maturity detection using computer vision

Recent advancements in computer vision have provided opportunities for new applications in agriculture. Accurate yield estimation of fruit and vegetable crops is very important for better harvesting and marketing planning and logistics. This paper proposes a method for detecting the maturity levels (green, orange, and red) of fresh market tomatoes (Roma and Pear varieties) by combining the feature color value with the backpropagation neural network (BPNN) classification technique. A maturity detection device based on computer vision technology was designed specifically to acquire the tomato images in the lab. The tomato images were processed and the targets of the tomatoes were obtained based on the image processing technology. After that, the maximum inscribed circle of the tomato's surface was identified as the color feature extraction area. The color feature extraction area was divided into five concentric circles (sub-domains). The average hue values of each sub-region were extracted as the feature color values and used to describe the maturity level of the samples. After that, the five feature color values were imported to the BPNN as input values to detect the maturity of the tomato samples. Analysis of the results shows that the average accuracy for detecting the three maturity levels of tomato samples using this method is 99.31%; and the standard deviation is 1.2%.