Fast estimation of optical properties of pear using a single snapshot technique combined with a least-squares support vector regression model based on spatial frequency domain imaging.

Spatial frequency domain imaging has great potential in agricultural produce quality control due to its advantage of wide-field mapping of absorption (µa) and reduced scattering (µs') parameters. However, it is not widely adopted in real applications due to the large time cost during image acquisition and inversion calculation processes. In this study, a single snapshot technique was used to obtain ac and dc components (Rd_ac, Rd_dc) of diffuse reflectance of turbid media (phantoms and pears). The validation results for the snapshot method indicate that at the spatial frequency of 1000/3 m-1, it achieved the optimal demodulation, by comparison with the results obtained by the commonly used time-domain amplitude demodulation method. Diffusion approximation, artificial neural network, least-squares support vector machine regression (LSSVR), and LSSVR combined with a genetic algorithm (LSSVR+GA) were then used to predict µa and µs' from the obtained Rd_ac, Rd_dc at the fx of 1000/3 m-1. Validation results indicated that the LSSVR method took the least time to calculate µa and µs' with high performance. The proposed imaging system and algorithm were implemented for the inspection of a pear bruise. Results indicated that the bruise, which is not obviously distinguishable in original gray maps, can show obvious contrast in calculated µa and µs' maps, especially in µa maps. Further, the contrast becomes more obvious with the passage of time. In summary, this study developed a low-cost spatial frequency imaging system and matching software that could realize fast detection of optical properties for a pear with the proposed snapshot and LSSVR algorithms.

CCD-Based Skinning Injury Recognition on Potato Tubers (Solanum tuberosum L.): A Comparison between Visible and Biospeckle Imaging.

Skinning injury on potato tubers is a kind of superficial wound that is generally inflicted by mechanical forces during harvest and postharvest handling operations. Though skinning injury is pervasive and obstructive, its detection is very limited. This study attempted to identify injured skin using two CCD (Charge Coupled Device) sensor-based machine vision technologies, i.e., visible imaging and biospeckle imaging. The identification of skinning injury was realized via exploiting features extracted from varied ROIs (Region of Interests). The features extracted from visible images were pixel-wise color and texture features, while region-wise BA (Biospeckle Activity) was calculated from biospeckle imaging. In addition, the calculation of BA using varied numbers of speckle patterns were compared. Finally, extracted features were implemented into classifiers of LS-SVM (Least Square Support Vector Machine) and BLR (Binary Logistic Regression), respectively. Results showed that color features performed better than texture features in classifying sound skin and injured skin, especially for injured skin stored no less than 1 day, with the average classification accuracy of 90%. Image capturing and processing efficiency can be speeded up in biospeckle imaging, with captured 512 frames reduced to 125 frames. Classification results obtained based on the feature of BA were acceptable for early skinning injury stored within 1 day, with the accuracy of 88.10%. It is concluded that skinning injury can be recognized by visible and biospeckle imaging during different stages. Visible imaging has the aptitude in recognizing stale skinning injury, while fresh injury can be discriminated by biospeckle imaging.