Skip to main content
ARS Home » Research » Publications at this Location » Publication #199851

Title: Textural Analysis of Hyperspectral Images for Improving Detection Accuracy

item Park, Bosoon
item Lawrence, Kurt
item Windham, William
item Kise, Michio
item THAI, CHI - UGA

Submitted to: ASABE Annual International Meeting
Publication Type: Abstract Only
Publication Acceptance Date: 3/14/2006
Publication Date: 7/11/2006
Citation: Park, B., Lawrence, K.C., Windham, W.R., Kise, M., Thai, C.N. 2006. Textural Analysis of Hyperspectral Images for Improving Detection Accuracy [abstract]. ASABE Annual International Meeting.

Interpretive Summary:

Technical Abstract: Detection of fecal contamination is crucial for food safety to protect consumers from food pathogens. Previous studies demonstrated a hyperspectral imaging system has a potential for poultry fecal contaminant detection by measuring reflectance intensity. The simple image ratio with optimal thresholding of reflectance values at 565 and 517-nm images was effective for identifying fecal contaminants on the surface of broiler carcasses. This simple algorithm performed well with high fecal detection accuracy. However, differentiating false positives from contaminants was always challenging. In addition to image processing in the spatial domain, further processing such as textural analysis can help reduce false positives. In this research, textural analysis of hyperspectral images was conducted to improve detection accuracy by minimizing false positives. Specifically, textural analysis with co-occurrence matrix of hyperspectral images was performed to identify “true” contamination regardless of fecal sources (duodenum, ceca, colon) and diets (corn, wheat, milo with soybean mixture). In addition, co-occurrence matrix textural features including average, variance, entropy, contrast, correlation, moment of poultry hyperspectral images were investigated for selecting optimal features to represent contamination. The goal of this study was to examine textural analysis algorithms to differentiate between contaminants and false positives of hyperspectral images for poultry food safety inspection.