Nondestructive Detection Of Eggshell Surface Dirt Based On Dynamic Image Processing

With the improvement of people material living standards, pursuiting of high quality and safe products has become a trend. Now the research of egg-quality detection cannot meet the on-line detection requirements. In this paper, an automatic detection system for dynamic eggshell surface dirt based on machine vision is developed. It can quickly detect the eggshell external defect such as surface dirt spots and blood spots. The research work lay a solid foundation for future study of egg-quality detection. The main contents are as follows:1.Egg-quality inspection hardware system is integrated, which contains the egg mechanical conveyor system, machine vision system and image segmentation processing software system. The best match parameters between mechanical conveying system and machine vision image acquisition system are determined. Three groups of egg images are acquired by machine vision system. They can reflect three different surface aspects for the same single egg image and ensure to acquire the whole information on the eggshell surface.2.In the paper, the single image was segmented from dynamic eggs and then was stored orderly in order to extract the whole surface features of single eggshell. The paper proposed a new image processing algorithm for the segmented egg image. Firstly, the 3×3 square window median filter is used to denoise. Then, the application effects of classical operators including Roberts operator, Sobel operator, Prewitt operator and Log operator were compared. Sobel is the best one, so it is used to detect egg edge. Besides,the quadtree-based splitting and merging algorithm is choosed to segment multi-regional images. At last, a new image is got by image fusion technique based on wavelet transform, in order to extract characteristic parameters using regional marker method.3.Visual C++ and MATLAB is succeeded to mix programming. The image processing software is designed under the Visual C++. MATLAB image processing software is called under the Visual C++ condition conveniently. The image treatment of various algorithms are realized with a message map, such as single partition, image grayscale, image enhancement, image smoothing, image sementation, region labering and egg dirt spots, blood spots area, perimeter eggs, egg diameter and short diameter and egg shape index and so on. Then, on-line egg-quality comprehensive detection is completed.