Real-Time System Research On Eggshell Crack Detection Based On DSP

Compared with the intact eggs in storage, the broken ones are more likely to go bad. According to national food safety regulations, it's necessary to remove broken eggs before the eggs are processed. In this thesis, on the basis of summarizing the worldwide research fruits of eggshell breakage identification, the image processing technology was used to segment the egg image real-timely and a simulated online monitoring experimental platform was set up to recognize broken eggs. The main content and conclusions of this research were as follows:(1) A simulated test platform for online detection was established. The test platform included a SEED-VPM642 experimental box (image processing system), an image acquisition system and a result display module. White LED light was chosen as the experimental light source, its intensity was adjusted by control circuit which was based on single chip computer 8051F330 PWM. The experiment result showed that it can provide a constant light intensity. A speed-adjustable device for egg rotation was designed, and coaxial illumination was adopted concerning on the light and camera, so that the system can inspect entire surface of the shell and acquire dynamic images.(2) According to the characteristics of the original image, two schemes including outer oval-cut rectangle and superposition algorithm were proposed to segment image in real time. The experiment proved that both methods can segment images in real time, but the stability and accuracy of latter was superior the former, so superposition algorithm was selected to segment images.(3) An innovative method of connected region labeling was proposed where the search area was extended to 5×5 regions, and a threshold T was setted to eliminate mendacious cracks. The optimal threshold T was selected through experiments, and it was found that when T was equaled 40, the best recognition can be obtained.(4) Some factors, such as narrowing the target area, reducing redundancy and simplifying the algorithm steps, etc., were taken into considerations, to optimize the code gradually. Finally, the average time to process an image was 0.7s.(5) TMS320DM642 chip doesn't have standard serial port, so the system expanded RS232 serial circuit to receive and send data. Three images were captured from every egg. After comprehensive treatment, the final result was displayed synchronously in the result display module by data communications.(6) Experimental verification of the study revealed that the identification accuracy rate was proportional to the length of breakage. When the length was more than 8mm, the identification accuracy rate of cracked eggs reached 96%.65 white shell eggs were randomly selected to test the system, and the identification accuracy rate of intact eggs and cracked eggs were 86.57% and 93.3%, respectively.