Research On The Technology Of Poultry Manipulator Eviscerating And Edible Viscera Sorting Based On Machine Vision

Poultry eviscerating and the sorting of edible organs are the most difficult to realize mechanization and automation in poultry slaughtering and meat processing.It is of great significance to study the automatic eviscerating equipment for enhancing the automation level of poultry slaughtering production line and improving the quality of food safety.At present,most of the automatic eviscerating equipment on the market comes from developed countries such as Europe and America.Because Europeans and Americans generally don't eat the poultry viscera,the foreign eviscerating equipment is not suitable for poultry slaughtering of China.Most of the domestic enterprises adopt manual assisted stream line operation,so it is of great significance to study the automatic eviscerating equipment for poultry and the sorting and recognition of visceral organs in China.To ensure the integrity of the visceral organs and reduce the damage of vulnerable organs in the process of manipulator eviscerating,machine vision technology was used for positioning accurately of poultry carcass and viscera.Based on the actual eviscerating process requirements during poultry slaughtering and processing,the in-vitro recognition of edible visceral organs was determined in present study.The main research contents and results were summarized as follows:?1?Image segmentation of poultry carcass and the whole viscera.The technique of poultry carcass and overall viscera segmentation,and its relative position deviation were studied to facilitate the location of the eviscerating manipulator,which could solve the problem of visible carcass and invisible viscera in poultry eviscerating.The color feature was the remarkable characteristic of the poultry carcass,which was helpful for the identification of the poultry carcass parts.Firstly,the best color space of poultry carcass for image segmentation was selected according to the analysis of different color spaces.3different color Spaces?RGB,HSI and Lab?images were analyzed for the poultry carcass image.And b-component images of Lab color space were selected for segmentation.Secondly,the viscera images were segmented by two-step method,the whole visceral regions were divided into the heart-liver regions and the fat regions,which were segmented and then integrated based on the characteristics of poultry viscera images.Based on C-V model,an improved level set algorithm was proposed for the heart-liver region image segmentation and sovling many interfering factors in abdominal cavity.These results were compared with that of the OTSU multi-threshold segmentation algorithm and the traditional level set algorithm based on C-V model,which showed that this algorithm could greatly reduce the boundary leakage of the target region.Moreover,they had obvious advantages in three quantitative performance evaluation indicators?segmentation accuracy IOU,over-segmentation rate OR and under-segmentation rate UR?and the segmentation accuracy reached 98.9%.Finally,the relative position deviation of the poultry carcass and viscera was obtained through the coordinate transformation.These experimental results indicated that the average deviation in the horizontal direction was2.4mm and the deviation in the vertical direction was large.The average deviation value of the 1000¹500g and 1500²000g poultry were 22.9mm,32.4mm,respectively.?2?Image segmentation and location prediction in vivo of chicken heart and liver.Based on the image segmentation technique of chicken heart and liver,the quantitative prediction model of carcass and viscera was established and proposed to predict the visceral position,which was used to overcome the problem of vulnerable organs damage in the process of manipulator eviscerating.In view of the adhesion problem of the three visceral organs such as heart,left and right liver,the improved watershed segmentation algorithm was applied in this study through a numerical calculation and distance conversion proposed.In this algorithm,the over-segmentation was effectively suppressed according to the correct gradient image and the forced minimum technology.In this study,50 three-yellow chicken with carcass mass of 1000²260g were selected as the study object,the relative position relationship of the poultry carcass,overall viscera,heart and liver were acquired through the image processing algorithms and coordinate transformation.The correlation between these indexes was analyzed and 9 prediction models including Weibull model,Logistic model,Gompertz model and polynomial regression was established by SPSS,CurveExpert and other software.These results demonstrated the maximum fitting degree R² of the overall visceral length that predicted according to the carcass length was 0.952,and the maximum fitting degree R² of distance between upper liver and muzzle was predicted to be 0.929,which meet the estimated requirements.?3?The image segmentation and feature extraction of edible visceral organs for poultry in vitro.For the sorting of edible visceral organs,three edible organs,heart,liver and muscular stomach were extracted from the whole viscera image and the characteristic values were calculated in present study.Firstly,a new color model was established in the Lab model space through normalization processing for the image segmentation of edible viscera,and then a color viscera image was obtained from complex background after the dot multiplication.Secondly,in order to fully express the features of 3 edible viscera,the9 color features,11 shape features and 21 texture features,a total of 41 characteristic value were extracted.Aming them,color features contained the mean values of three channels of the RGB,HSI and Lab color space.And shape features included the perimeter,area,aspect ratio,degree of circular and 7 Hu invariant moment features M₁-M₇.Besides,the gray-level co-occurrence matrix consisted of 16 features such as the energy,entropy,moment of inertia and correlation characteristics on the direction 0°,45°,90°and 135°.Tamura texture features mainly contained the roughness,contrast,direction,linearity and regularity.Finally,qualitative analysis was made based on the change rule of characteristic values.The experimental results indicated that the mean value of color features from HSI channel and the correlation,coarseness,contrast of texture features had higher resolution on liver and muscle stomach,which the circumference and area of the shape features could help to distinguish the difference between the heart and other two kinds of internal organs.?4?Study on identification of edible viscera for poultry in vitro.Through the analysis of the shape features of the heart,liver and muscular stomach,the recognition of the heart was realized by the method of marking the shape features of the connected domain.In order to improve the stability of algorithm,all features were analyzed to reduce the dimension and the multiple feature fusion detection model of edible visceral organs was established in this study.Based on BP neural network modle,SVM modle and parameter optimization,the simulation and classification results of 313 samples data suggested that SVM model had better recognition effect than BP neural network model,which its identification accuracy reach 98.33%.Moreover,the SVM model had faster recognition speed?5.29ms?,which fully meets the requirements of edible visceral organ recognition.?5?Construction and experimental analysis of poultry eviscerating manipulator system.In order to solve the problem of visceral damage in the eviscerating process,the machine vision technology and visceral location prediction of poultry viscera were applied to the poultry eviscerating system,and the poultry eviscerating manipulator system was constructed and analyzed in present study.Firstly,through the analysis of the eviscerating process,the mechanical structure of the poultry body conveying device,the rectangular coordinate mechanical arm assembly,the poultry eviscerating manipulator were constructed in our experiment.Secondly,after the vision system was established,the poultry carcass and muzzle were fixed for accurate position through threshold segmentation,maximum inscribed circle acquisition algorithm and other image processing methods.Moreover,the consistency requirement of poultry was reduced and online eviscerated was completed through the motion path planning of manipulator by control system.Thus,the manipulator was leaded into poultry abdominal cavity and avoided the damage of vulnerable organs.Finally,the comprehensive performance test in the eviscerating system were carried out for two groups of poultry with different body weights,and the test results indicated that the body size had a certain impact on the visceral residue rate and visceral damage rate,highlighting that perhaps smaller poultry had a greater risk of visceral damage.The average visceral residual rate was 6.24%and the damage rate was 15%.In conclusion,based on the automatic eviscerating and sorting of poultry,the manipulator eviscerating visual system and identification methods of edible organs were carried out in this study,which could provide more technical support for the fully automatic operation of the poultry slaughtering production line.