Fruit detection in the orchard under natural lighting condition is still a challenging task. This paper presents a study of image processing using local features to detect individual pear in the field under natural lighting condition. An RGB-Near infrared (NIR) camera was used to take high contrast images using reflectance difference between the fruit and the leaf. A two-stage approach was investigated to detect fruits. In the first stage, the Haar-like feature was applied to detect interesting regions. The second stage, the Histogram of Oriented Gradients (HOG) was used to detect fruits. We demonstrate that we can obtain higher detection accuracy than single stage approach.