Eggshell crack detection based on computer vision and acoustic response by means of back-propagation artificial neural network

An experimental system utilizing acoustic response system (ARS) and computer vision system (CVS) for eggshell crack detection was implemented. Firstly, the acoustic response signals were captured and analyzed, and six parameters (f1, f2, f3, f4: the dominant response frequency; CS: the mean value of coefficient of skewness; CE: the mean value of coefficient of excess) were analyzed. The ARS including a back-propagation artificial neural network model with a structure of 6 input nodes, 15 hidden nodes, and one output node was built to detect eggshell cracks. Secondly, the eggshell images were captured and processed by the computer vision system, and five geometrical characteristic parameters of crack and noise regions on the eggshell images were acquired. The CVS including a back-propagation artificial neural network model with a structure of 5 input nodes, 10 hidden nodes, and one output node was built to detect eggshell cracks. Finally, the quality of eggs, with or without cracks, was evaluated based on detection results from both CVS and ARS. This method allows the fusion of information obtained from CVS to ARS. The results showed that the detection accuracy of cracked eggs were 68 and 92%, respectively, by CVS and ARS. However, the accuracy equaled to 98% by the information infusion of two techniques. The result was superior to only one technique, and the method based on the information fusion of computer vision and acoustic response was applicable for detecting egg cracks. This research provides a new technology detection of cracked egg.