Research On Infrared Thermography Technique For Detecting Early Chanical Damage In Apples

The defects detection in fruit is very important in agriculture products processing. Especially the apple with early mechanical damage which has no apparent changes of color, smell and taste, it will lead to big economic losses result from infection of microbes and pulp rottenness. At present, defect detection in fruits in our country has been predominantly performed manually or by visible spectrum imaging systems, which are not capable of effectively distinguishing fruit with mechanical damage which occur a short time before inspection.Aiming at above problems, the infrared thermography nondestructive technique introduced in this paper can address it. According to the specific situation of China, Red-Fuji apple was chosen and the detailed study of infrared thermography detection of early mechanical damage in it was done from two perspectives mainly: heat transfer analysis and image processing. The main work and achievements are as follows:Firstly, it analyzed and compared the merits and demerits of fruit quality non-destructive techniques, and reviewed the achievements and potential problems and key technologies of infrared thermography technology introduced in this paper of fruit defects detection, and then presented the main content of our research and the way we use.The first three parts are analysis in accordance with heat transfer theory. Firstly, it introduced the basic theory and laws of infrared radiation, established the infrared thermography system and experimental method for detecting apples mechanical damages, analyzed the radiation laws of spherical surface, researched the non-destructive detection mechanism of infrared thermography of fruit, and build a heat transfer model of one-dimension unsteady state heat conduction equation for evaluating the cooling properties of spherical fruits. And then it analyzed the temperature behavior within different regions of apples in thermal images, and analyzed the dynamic temperature cooling curves from experimental measurements of defective apples. The experiments results revealed that the features of level temperature profile of apples in thermal images could be as an effective method to identify different mechanical damages, stalk and calyx. From the theory analysis and experimental study, we gained ftinction logarithmic curve of apples surface temperature and time ideally, it would provide another way to recognize different mechanical damages. Besides, from the cooling analysis, we found that the good temperature contrast between scratch and sound tissue could last 60 second to 240 second; the best detection time varied 0 second to 30 second after thermal excitation. And the good temperature contrast between bruise and sound tissue could last 20 second to 100 second; the best detection time varied 30 second to 60 second after thermal excitation.The last half parts tries to study the infrared thermography detection of fruit damages from digital image processing. Firstly, on the basis of analyzing the gray features of apple thermal images, it proposed a non-uniformity correction method based on histogram matching and Single-Scale-Retinex algorithm under low scale, which using the histogram matching algorithm to highlight the contrast of raw thermal images, and utilizing SSR algorithm to enhance the edges of fruits in raw thermal images, with combination of contrast adjustment image and edge enhanced model, final fused image can be obtained. Then selecting a threshold to segment defects and utilizing the loss rate at accurate segmentation of defects to assess the effectiveness of above methods. And then the post-processing measures such as defects dividing, morphological processing and edge extraction were done in order to get the contour profile of damage regions and research the thermal diffraction effects of these regions. The testing results demonstrate that the thermal images after non-uniformity correction show rich details and good contras. After thresholding, the biggest loss rate at accurate segmentation of defects is only 3.05%. Through numerical experiments show that the accurate detection rate of sound apples can reach to 92%, and the accurate detection rate of defective apples is still over 87%, which can be well used in fruit early mechanical damages detection.