Thermal imaging for detecting fungal infection in stored wheat

The feasibility of using an infrared thermal imaging system to identify the fungal infection in stored wheat was studied. Thermal images of bulk wheat grains infected by Aspergillus glaucus group, Aspergillus niger van Tieghem and Penicillium spp. were obtained using an un-cooled focal planar array type infrared thermal camera after heating grain for 180 s with a plate heater placed 10 mm above the grain and maintained at 90 °C, and then cooling in ambient air for 30 s. In total, twelve temperature features were derived from heated and cooled wheat and four-way and pair-wise classification models were developed by linear and quadratic discriminant analyses (LDA and QDA). Leave-one-out and bootstrapping methods were used to validate the developed classification models. Pair-wise LDA and QDA classification models gave a maximum accuracy of 100% for healthy samples and more than 97% and 96% for infected samples, respectively. Four-way LDA and QDA classification models yielded relatively low classification accuracies for fungus-infected samples due to the non-significant changes in the temperature features between grain samples infected with different species of fungi.