基于DSP实时图像分割算法的鸡蛋蛋壳破损检测

Detection of eggshell defects using the segmentation algorithm of real-time image based on DSP

为了克服声学振动法易受到蛋壳结构和外界噪声等因素的影响,提高判别准确率,提出基于DSP实时图像采集处理的鸡蛋蛋壳破损检测方法。采用白色LED环形同轴光源,达到鸡蛋表面均匀受光。将DSP实时采集的鸡蛋图片灰度化处理后,应用外接矩形法描绘鸡蛋近似椭圆投影的外切矩形。在该矩形区域内,删除鸡蛋投影边缘,利用改进型标记法判断鸡蛋是否破损。实验结果表明,此算法对于蛋壳裂缝长度大于8mm的鸡蛋识别率达到96%,对裂缝长度小于6mm的鸡蛋难以识别;但当蛋壳斑点面积大于15.04mm2且与蛋壳颜色差异性大时,好壳蛋被误判为损壳蛋。

vulnerable for the shell structure and the external noise and other factors,proposed DSP-based non-destructive testing method to achieve image and recognize the eggs with crack real-timely.A LED front lighting was used to generate images of a single egg.Regarding the projection of egg as an elliptic,and drawing the outer oval-cut rectangle.In this area,removing projection edge and detecting crack by improved labeling.The experimental results show that the rate of testing precision of cracked eggs can reach 96% when the crack length more than 8mm,and when the crack length is less than 6mm the eggs are difficult to identify.But when the dot-area is more than 15.04 mm2 and its color has large differences,the intact eggs would be judged into cracked eggs.