DVC Machinevision b.v. Dutch Machine Vision Conference May 2018

Transcription

1 DVC Machinevision b.v. Dutch Machine Vision Conference May 2018

2 MEYN DVC and Poultry industry

3 Vision systems are used for Food Sustainability by monitoring results of hatching Grading Options Comply to government rules Higher standards of chicken living condition Performance of sorting animals

4 History Lot of waste

5 History: The start of Machine Vision Optimalisation, 6 Sigma delivers: Waste Reduction

6 Grading across the slaughterhouse? The ideal situation Bird Counter Vision Footpad Inspection Camera / Hockburn Inspection Camera DOA Bird Counter Vision Quality Grading System Multi-Spectral Grading System

7 Inspection across the slaughterhouse Using manual inspection some high lights Counting Veterinarian Inspection Inspection for dead birds Desease Inspection Sorting Bone inspection

8 Why the importance of grading? Effective analysis of product quality Whole bird quality Parts quality Animal welfare legislation Disease analysis Bacterial analysis Understanding the slaughterhouse performance X-Ray Analysis

9 Comply to government rules USDA (1977) Study in EU Slaugter house Between fast growing broilers and slower growing broilers Food pad Dermatis / (FDP) and Hockburns 22% / 20% 7% / < 1%

10 Comply to government rules USDA (2004) WellFare (started in the 80th) Sweden NL D Vision system Approval 2013

11 Why is it so difficult to get right? Variation of the product Bird size, variation in the process Quality demands of the End-customer Efficiency demands of the market Line speeds keep increasing Legislation becomes more stringent Footprint becomes more important Physical environment of slaughterhouses is tough for technology (IP69K)

12 Inspection across the slaughterhouse Using manual inspection some high lights Counting Veterinarian Inspection Inspection for dead birds Desease Inspection Sorting Bone inspection

13 What did we do before automation? Position of grading While hanging the birds at the shackle In line, birds hang already in shackle Afterwards when birds are dropped Purpose of grading Detect D.O.A Remove already unhealthy birds Remove clear quality defects (broken wings, feathers remaining) Remove sick birds Remove poorly processed birds Correct quality for retailer Final quality checks before packing / shipping Issues Not always detectable for D.O.A Speed + manual hanging = risk of mistakes Speed increasing = less time for operators to inspect More Too late? Cross contamination Order matching

14 Video of the grading

15 Manual Grading vs. Camera Grading Manual Grading The skills and the state of mind of the grader can vary Changes in grading strictness need to be updated through training Quality detection level influenced by human interpretation High speed quality grading is almost impossible Subjective measurements varies with the grader Due to limited inspection time the grader can judge a few control points only History of data is not available quickly Data registration takes time Automatic Grading The grade is based on critical limits previously set Many parameters are involved in grading each bird Breast side and back side can be identified Up to 7 grades possible High speed grading is no problem Automatic data registration for real time and historical referencing

16 What options are available? Bird Counter Vision Footpad Inspection Camera Hockburn Inspection Camera Quality Grading System - Whole Bird Quality Grading System Parts Quality Grading Leg Detection Quality Grading Head Detection Quality Grading Wing Pit Detection

17 Grading across the slaughterhouse? The ideal situation Bird Counter Vision Footpad Inspection Camera / Hockburn Inspection Camera Bird Counter Vision DOA (only life broilers can be processed) Quality Grading System X Ray Multi-Spectral Grading System

18 Differences in images captured by Meyn products Bird Counter Vision analyzes the contrast from the back light and the image to determine if a bird is in the shackle Head detection analyzes if the head is attached or not. This can be used for quality purposes. Hock burn camera focuses on the hock joint area and analyzes burn marks Leg Detection searches for a cut in the leg to direct the product for further processing QGS both front and back camera s analyze various parts of the griller, allocate a grading to each part and use that to determine a combined whole bird grading Footpad Inspection system analyzes lesions on the footpads and references against legally recognized classifications

19 Differences in Quality Grading Images per Chicken Type Yellow Chicken White Chicken Water Chill Air Chill Marinated Products Good Downgrade Capability to grade various types of chicken that have been through either air or spin chill Capability to grade marinated products, such as Pintado in Mexico

20 Examples of Quality Grading Images per Chicken Type

21 Bird Counter Vision Counting of birds in the bleeding line with an accuracy never seen before Line speed between 1,000 and 15,000 shackles per hour. Accuracy of > 99.95% Counting of birds with feathers on or off (not available with Optic) No limits as to bird size or type

22 Bird Counter Vision Understand your input and processing efficiency at set points throughout the slaughterline Advanced counting parameters The bird counter vision is also able to register the hanging performance. For this reason following values are measured as well: Birds hung with only one leg Birds hung in the shackle, not completely in the bottom of the shackle Loose feet or hocks in a shackle without bird, as indication for a bended shackle or a problem in the hock unloader

23 Footpad Inspection System (FI-S) Grade the severity of footpad lesions to determine animal welfare characteristics and product quality Smart camera captures image of the footpad as it passes Located after hock cutter / re-hanger from Kill > EV Image is analysed by software Images stored for re-configured amount of time Data stored for at least one year to be referenced if required (for legislation purposes)

24 Footpad Inspection System (FI-S) Grade the severity of footpad lesions to determine animal welfare characteristics and product quality Captured image is analysed by software Image is graded based upon a scorecard or reference table - set as the Swedish Scorecard as standard All information is captured and scores are given to a flock and farm based upon this Scores are used for animal welfare legislation Information can also be used for further quality grading and separation with a selective hock ejector. This for that plants who have a market to sell the foot.

25 Hockburn Inspection System (HI-S) Grade the severity of hockburns detected to determine animal welfare characteristics and product quality Used where it is not possible to capture an image of the footpad. Foot removed without hock Where feet are removed in the rehanger Positioned whilst birds are still shackled Same camera and machinery as the FI-S Score of 0 or 1 given (0 = acceptable / 1 = bad quality) Used for animal welfare traceability and supplier quality performance

26 Quality grading of a whole bird before the sorting drop point Quality Grading System for Whole birds (QGS-WB) Breast & Back-side detected Birds graded on classes from A-H Quality I automatically sent to cut-up Grading can be customized per customer Multiple grading settings can be set to supply multiple retailers Further configuration is then required for grading of specific parts Configuration required for parts routing into the correct processing lines Can be located in EV to grade products in advance of chilling, for order matching purposes

27 Quality grading of individual parts for sorting after cut-up Quality Grading System for parts Same camera system as whole bird grading More powerful software to analyse individual parts of the bird before cut-up Give each part a grading Grading determined by customizable parameters (per customer, retailer etc) Blood spots (bruises) Brown spots (burn spots) Torn skin Feather rests on wings

28 Quality grading of individual parts for sorting after cut-up Quality Grading System for parts Hocks Drums/Legs Breast Thigh Hocks Drums Back Tail Wings Wings

29 Quality grading of individual parts for sorting after cut-up Quality Grading System for parts Wing partly missing Broken Left Wing Thigh Skin Missing Blood Spot Right Wing

30 Leg Detection Intentionally make a cut in the leg of a shackled chicken so it can be selectively removed at a further stage of processing Line speeds are increasing and veterinarian workload is increasing Time to analyse each bird for all quality related inspection is reducing Time to remove a bird from the line is reducing The leg detection system allows the vet to make a quick incision on the leg of a bird to identify them as a quality downgrade or rejection The bird will be removed mechanically at a later point in the line The vet does not need to manually remove the bird so can spend more time inspecting the birds

31 Veterinarian cut The vet makes an incision at a set point on either the left or right leg The cut needs to made in the centre of the leg at the back The cut cannot be made on the front due to the hock joint Semi automatic using a few instructions to work with.

32 Customer sells birds with and without neck Neck Detection Depending of product quality system to have a choice to ensure the neck is fully removed or not before sorting Used for Panama Birds Without neck With neck Ensures right chicken enters the whole bird packaging area

33 Wing Pit Detection Downgrades from wing pit damage are big for some customers, as they cannot be identified until they have already been dropped Driven by large Canadian customer who have >20% of birds with Wing pit damage This damage is not identified until the birds have passed into the drop stations Manual re-classification is then required Meyn/DVC asked to develop a solution Similar questions have arisen in Europe If a bruise is present at location A, it is highly likely there will be a bruise at location B Camera is sensitive to red color Camera from backside is pointing downwards With this combination the camera can detect wing pit damage

34 Wing Pit Detection Downgrades from wing pit damage are big for some customers, as they cannot be identified until they have already been dropped Front camera does not have a position to detect wing pit damge Using the camera image taken before/ after the bird has passed the central point gives a view from the side of the bird This can be used to get a better angled view of the wing pit Similar analysis to part bird grading can then take place Testing has identified >80% success rate Investigation on improvements being made (install static wing stretcher before measurement) Image One (Full frontal) Straight View Image Two slightly from the side Side View

35 Pintado Grading Grading for marinated products developed for pintado products in Mexico. Can be matched to colour specification and sub quality products removed Ensure a consistent coverage of marinate Match the product specified colour to the actual Define if the product is within or outside of the spec Good Downgrade Mexican marigold

36 Analytics Have one central point of data where users can access the statistics for realtime or historical reporting Customers with more than one camera system have to use an interface for each system which becomes confusing Each camera system delivers detailed information, which can overwhelm the user The data needs to be stored, at its lowest level this can be a huge amount of data generated per hour, day, week We have seen that to much data is not helping the producer of the broilers. There is a need for a system that can extract the data, and make it usable for different functions in a slaughterhouse (from operator to management level)

37 Dashboard view Number of production lines (with camera s) KPI s bird defects System performance Live production and statistics dashboards

38 Monitoring Production Statistics Maximum Production Capacity Total % of A- grades produced across all lines Total Number of birds processed Line stop / Break Rolling / Realtime count per hour

39 Monitoring Production Statsicts Green bar shows the total count of defects The black line shows the cumulative defects e.g. in this case 50% of defects are caused by the first four bars Red dot indicates what the defect is as a % of total defects The final dot will always add up to 100 The defects along the bottom are identified in the Quality Grading System

40 Q&A Session At DVC Machinevision b.v. we are always open for new partnerships. If you have any questions, or simply want to learn more about machine vision applications, just get in touch!