Detection of the Flower from Field Image using Morphological Technique

Abstract

Proposed work focus on designing and implementing an algorithm for detecting and counting red flowers in the field. It uses image analysis technique to identify and estimate number of red flowers in a field with varying light, growth and flower sizes. The algorithm can be used for yield estimation. Identifying the flowers can help the farmer, by providing useful data viz. the number of flowers in a row, bloomed flowers after previous investigation. The proposed technique will be useful to cater the realistic agricultural problems comprising coping with changing light conditions, shadowing, and occlusion. Proposed approach detects flowers by using a fixed threshold, partitioning over the HSV color space, and morphological cues. The threshold classifies the images into dull and brighter images. After that portioning is done respectively on the hue, saturation and volume, then categorized as per flowers size and destiny. The images were gathered from different cameras taken at various angles, distances and periods of the day. Different parameters of images were calculated. For the view facing the flowers than any other view results analysis shows improved performance. Afternoon captured images gives good precision and recall values. No difference is noticed between the images of same location captured using two different cameras. Comparative analysis results in better precision and recall while observing images captured in the noon time and from the front.