This paper presents a novel approach to identify chile pods in two dimensional images of chile harvests in order to quantitatively evaluate harvesting and cleaning systems. Chilies vary greatly in shape, color, and texture throughout the season and within varieties and thus pose a unique challenge compared to the identification of other more uniform objects. Since chilies do not have distinct point features typically exploited in object recognition algorithms, a normalized shape distribution function was used to describe each object's general shape. 486 harvest object images were collected throughout the season and then placed in categories. Representative shape functions were derived from chilies in a given category and the other chilies, stems, and leaves were compared to each class. Sample shape distributions were found to provide a robust signature to discriminate among chile pods and harvest trash objects.