Detection Of Navel Orange Surface Defects And R & D Of Grading Equipment Based On Machine Vision

Detection of fruit surface defects is an important step in fruit postharvest quality grading.The uneven distribution of brightness on the fruit surface caused by the fruit shape makes it hard to effectively segment and extract the surface defects.Therefore,compared with external quality indexes such as color,size,and shape of fruit,the rapid detection of surface defects is a difficult point in the fruit grading process.The external defect of fruit is also one of the potent factors determining its value,which is the most intuitive embodiment of fruit quality.With the application of machine vision technology in nondestructive testing of agricultural products,more and more scholars have joined in this research field.In this research,‘Newhall’ orange was selected as the experimental object.Based on machine vision technology,the detection algorithms of 14 common surface defects(e.g.,canker,thrips,scale infestation,insect injury,alternaria rot,wind damage,anthracnose,dehiscent,copper burn,blackspot,phytotoxicity,oil spotting,black rot,and mechanical injury)of the navel orange were developed and discussed in detail.A grading equipment for postharvest navel orange was developed,which could be used to on-line detect the size and color of navel orange and grade the fruit according to national standards or user-defined standards.It provided the theoretical basis and hardware support platform for the next research and development of the rapid on-line detection system of navel orange surface defects.The main research contents and conclusions of this paper were as follows:1)Aiming at this phenomenon that the surface brightness of navel orange was unevenly distributed,a new algorithm was designed and developed based on mask and brightness correction to detect the surface defects of the navel orange.A global threshold was used to segment the background by the bimodal method and then filled the holes to obtain a mask template.The mask template and the component image were subjected to a point operation to generate a fruit image.A multi-scale Gaussian function-based image brightness correction algorithm for brightness correction was proposed in this paper.The fruit component image and the constructed multi-scale Gaussian function were convolved to obtain a component surface illumination image.The fruit component image and the illumination component image were divided by points to get the brightness correction image.Finally,the single global threshold method was used to extract the surface defects of the navel orange.Based on this algorithm,defect sample detection rate was achieved 92.7% for 8 common surface defects(e.g.,mechanical injury,black rot,canker,thrips,oil spotting,alternaria rot,wind damage,anthracnose).This algorithm effectively solved the difficult problem of surface defects segmentation caused by the uneven brightness distribution of navel orange fruit.It provided technical support for the on-line accurate classification of navel oranges and a new method for the rapid detection of surface defects of other spherical fruits.2)A region brightness adaptive correction algorithm was proposed for orange surface detection in this paper.The aim of the whole image brightness correction was achieved through the brightness correction in the image region.Firstly,the neighborhood window size of the target image was set as w × w(neighborhood window size w set to 13),and the brightness of every pixel of fruit was determined by the brighter pixels in its neighborhood window.Then based on its surface brightness information,the target image with background removed was corrected by homogenization.After the brightness correction,the gray contrast between the surface defect area and the normal tissue area was obviously shown in the corrected images,so the single threshold method should be used to directly extract the surface defect from the brightness corrected navel orange image.Finally,the area filter was implemented to remove the stray points and noise.The method can effectively overcome the uneven brightness distribution of the navel orange surface.Based on this algorithm,8 kinds of common navel orange surface defects(e.g.,canker,thrips,scale infestation,insect injury,blackspot,wind damage,anthracnose,dehiscent)were detected,and the defect-recognition rate reached 95.8%.Compared with the other three defect detection algorithms,the correct defect-recognition rate was improved by 2.6% to 8.2%,the recognition speed was reduced by 0.27 s,0.14 s,and 1.45 s respectively.3)Based on the non-brightness correction model,an adaptive threshold algorithm based on image block was designed to detect surface defects of navel orange in this paper.The image was divided into multiple blocks.Theoretically,if the number of image blocks is sufficient,the surface brightness distribution of each block is approximately uniform.Using the proposed algorithm to extract the defects of each block and splice them together,the surface defects of navel orange could be detected.The proposed algorithm overcame the traditional complex algorithm to detect the defects,and it was more effective to solve the difficult problem of defect detection caused by uneven brightness distribution on the surface of spherical fruits.The samples with 12 kinds surface defects,including canker,thrips,scale infestation,insect injury,blackspot,anthracnose,dehiscent,black rot,phytotoxicity,copper burn,wind damage,and mechanical injury,were detected by this algorithm,and 97.1% of the fruits were correctly identified.4)A vehicle-mounted navel orange post-harvest on-line grading equipment was developed.The equipment could be used to online test the size and color of the navel orange,and classify the fruit at the field,and the grading speed achieved 5 fruits per second.