Egg crack detection using acoustic response signals based on AI

[BALCI29]

 

Contact Information

Country: TURKEY

Year Submitted: 2018         

University: AFYON KOCATEPE UNIVERSITY

List of Team Members (with year of graduation): Zekeriya BALCI (2016)

Faculty Advisers: Asst. Prof İsmail YABANOVA

Main Contact Email Address: iyabanova@gmail.com  balcizekeriya29@gmail.com

Project Information

Title: Egg crack detection using acoustic response signals based on AI.

Description: In this project, the detection of egg shell cracks was carried out by means of artificial neural network and support vector machines using sound signals that generated a mechanical effect that would not damage the egg shell surface.

Products: 

• Sofware 
• Labview
• Real Time
• Crio-Device Drivers
• FPGA-Target
• Analytics and Machine Learning Toolkit

• Hardware
• cRIO-9074
• NI-9375
• NI-9215

The Challenge: 

With this study, the detection of cracks in the egg shell was realized with artificial intelligence. In addition, micro-cracks that can not be seen and can not be detected by image processing methods can be classified correctly. It is possible to switch between dynamically desired artificial intelligence classifier (ANN-SVM) with the developed user interface program.

The Solution:

The aim of this work is to perform the crack detection process in the egg shell with the maximum possible accuracy. The experimental setup of the study is given in Figure 1. System consists of egg support unit, driver and amplifier circuit, cRIO, power supply, computer and software environment.

 Figure 1 : Experimental Setup

When the Egg Crack Detection User Interface program shown in Fig. 1 is executed, cRIO system sends a trigger signal to the card which drives the mechanical effect unit. The driver card energizes the mechanical impact unit and applies an impact to the egg shell as a physical impact so as not to damage the egg. The acoustic signals generated by the mechanical effect are amplified by the amplifier circuit. The analog signal is then converted to a digital signal via cRIO and the response signal received from the egg is sent to the software environment for classification.

The Egg Crack Detection User Interface software converts the egg signal into an input data format suitable for artificial intelligence functions. Based on the selected artificial intelligence method (SVM or ANN), the input data set from the egg data is applied and class prediction for the egg is performed. Artificial intelligence training and test results are given in Table 1

Table 1 : Artificial intelligence training and test results

Artificial Intelligence Method	Training	Testing
ANN	0,99	1
SVM	1	1

In Figure 2-5, images related to detection of intact and cracked eggs in real time with SVM and ANN are given.

Figure 2 : SVM pretiction result for crack egg.

Figure 3 : SVM pretiction result for intact egg.

Figure 4 : ANN pretiction result for crack egg.

Figure 5 : ANN pretiction result for intact egg.

Figure 6 : Crack egg with micro-crack that is not visible under normal condition

Figure 7: Micro-crack is visible if pressure applied.

Link to Video

https://youtu.be/8gEoVAkY678

Attach Poster