水果自动分级的机器视觉系统

2.image analysis procedure
Image segmentation
Morphological features
External features
2.image processing Step 1: acquisition of the first image (1)and the segmentation of its pixels into pre-defined classes by means of the above-mentioned table. smoothing procedure
Thank you ！
2. The machine vision system showed good results in detecting stem and blemishes.Damaged area is properly detected in apples,but the algorithms need to be tested more extensively in oranges and peaches. Further work should be done inorder to detect the defects that were not correctly discriminated, mainly because of its light colour, similar to the colour of sound skin.
The research status
The use of machine vision for the inspection of fruits and vegetables has increased during recent years.Nowadays, several manufacturers around the world produce sorting machines capable of pre-grading fruits by size, colour and weight.Nevertheless, the market constantly requires higher quality products and consequently,additional features have been developed to enhance machine vision inspection systems.
Step 2:extracte features to classify the fruits by size
binary image; except those considered as background or stem
extracte boundary;calculate the area and the size
Conclusions 1. The segmentation method is fast and appropriate for on-line processes, but depends much on the colour of the objects to be inspected.So it needs to be trained by expert operater.
Stem detection
93 images of oranges, 95 images of apples and 140 images of peaches,all of them acquired on-line; error analysis
On-line performance and repeatability
Detection methods
Hardware
colour camera frame grabber personal computer
Image analysis
1.off-line training The image analysis was performed by a specific software application developed at IVIA using the programming language C. off-line training: Using recorded images of fruit, an expert selected the different regions on the images and assigned all the pixels in every region to one of the pre-determined classes: background, primary colour, secondary colour, general damage type 1, general damage type 2, specific feature, stem and calyx.
3. Comparing these results with the average repeatability of human estimation of the size and the degree of skin damage, that were of 94 and 88%, respectively, and considering that the decision algorithm was trained and tested also by human operators,we consider that the results are acceptable.
Machine Vision System for Automatic Quality Grading of Fruit
Presenter : 华明亚 Student ID：14721353
Literature sources： AE：Automation and Emerging Technologies Received 14 August 2002; accepted in revised form 1 May 2003; published online 20 June 2003 Author introduction： J. Blasco Professor of University Jaume I, Campus Published articles: 1.Structural properties in R Fe 2 O 4 compounds ( R = Tm , Yb, and Lu) 2.Multivariate methods and artificial neural networks in the assessment of the response of infaunal assemblages to sediment metal contamination and organic enrichment
1.Apple economic crops 2.appearance influence sale 3.clean and sort products 4.High labor costs
1.automation of production 2. improve efficience 3.Cost savings 4.promote sales 5.Improve the economic value added
Results and discussion Evaluation of the segmentation procedure
Table 1 presents the pixel segmentation performanceof images of oranges, apples and peaches,pointing out a good estimation of the centroid and the size of the fruit. error analysis
1 2
The purpose and significance of the research
The research status Detection methods Results and discussion Conclusions
Catalog
3
4 5
The purpose and significance of the research
an error of 1mm can be allowed to the vision system when analysing the results of the on-line repeatability tests.
manual classification averaging 88%, mathine classification averaging 86% error analysis;time
Step 3：the area of each independent region was measured
all area measຫໍສະໝຸດ Baidured
stem location detection
the following features were measured for each fruit: (1) the length of the major damage--defined as the length of the major region, classified as damage, found in any of the four views; (2) the damaged area--equal to the sum of all the areas damaged, found in the four independent views; (3) the stem and calyx--considered to be present if found in any of the views; (4) the primary colour--calculated as the average ofthe primary colour estimated in eachindependent view; (5) the secondary colour--calculated as the average of the secondary colour estimated in each view; (6) the fruit size