Method For Complete-Surface Image Information Acquisistion Of Apples

Quality detection and classification of fruits was an important way to improve the added value of fruit commodities. Information leakage and repetition of collected fruit surface images during color and defect detection would influence subsequent classification.We used apple as research objects and focused on system construction and image mosaic for the whole surface image information acquisition of spherical fruit.The main research contents were as follows:(1) An image acquisition platform was build up which consisted of basis visible light machine vision system, electric rotary machine and stepping motor controller which could control the turning of fruits accurately(2) The petal element model mosaic method for surface image information acquisition of apple was established. The petal elements of each apple image were extracted to compose target image based on the average radius of apple and the number of petal elements. The test of238images from17groups of samples showed the average redundant information was2.91%for apple images with mean degree of circularity0.875～0.942.(3) The adaptability of SIFT operator to surface image of apple was tested and verified. The test results of112apple images indicated that SIFT operator was available for the feature points detection of apple images which has been transformed such as geometry scaling, brightness change, pepper noise and small-angle rotation.(4) Splicing of apple’s surface images based on SIFT description operator was realized. The research that using strip regions of center of apple image, RANSAC fake matching point elimination and optimal splicing points to realize image mosaic of surface images of apples. The splicing experiments of270images with20degrees intervals showed that this method could overcome the influence of transverse direction image distortionof apples to mosaic and remove the repeated region image of apple surface.(5) The method for acquiring rotating angle of apple based on stem side images was presented. The detection results of19980images using feature points extraction based on SIFT operator, center extraction of minimum spanning circle and removing of fake matching points in feature points showed:this algorithm was most appropriate for Red Delicious Apples; the correlation coefficients between test value and actual value were0.9846; this algorithm could be used for the angle calculation of apples on consecutive positions.