Study on Spinach Chlorophyll Detection Method Using Computer Vision and Artificial Olfactory Sensor
Article
Figures
Metrics
Preview PDF
Reference
Related
Cited by
Materials
Abstract:
Developing a rapid and nondestructive detection method for evaluating the quality of post-harvest spinach is essential for spinach producers and retailers. Typically, chlorophyll content was used as a reference index for spinach freshness, and the samples used in this study were stored at 4 ℃ for 12 d before being assessed by computer vision (for sample images) and electronic nose (for sample odors). Feature variables extracted from image and odor information were analyzed by principal component analysis (PCA) method. A chlorophyll content prediction model was established with back propagation artificial neural network（BPANN）. The principal components (PCs) were used as the input parameters for the prediction model, and the chlorophyll content was used as the output. Experiments showed that the optimal prediction model of BPANN based on computer vision was obtained with four PCs. The root-mean-square error (RMSE) was 0.1978 mg/g and 0.2147 mg/g, and the correlation coefficient (R) was 0.8457 and 0.7995 for training and prediction sets, respectively. The results of the prediction model using BPANN based on the electronic nose showed that the RMSE was 0.3119 mg/g and 0.3032 mg/g, and R was 0.7013 and 0.7493 for training and prediction sets, respectively. The results of the BPANN model based on the fusion technique showed that the RMSE was 0.1759 mg/g and 0.2121 mg/g, while R was 0.8888 and 0.8736 for training and prediction sets, respectively. The prediction accuracy of the fusion technique was improved compared with that of either of the single detection methods. Overall, the results showed that it is feasible to predict the chlorophyll content of spinach during storage using computer vision and electronic nose, and the fusion technology is helpful in improving the prediction accuracy.
