Design Of Spraying Control System For Moisturizing Pig Carcass Based On Wet State Of Epidermis

In view of the existing automatic recognition methods for the dry or wet state of the pig carcass surface based on image processing technology,the feature parameters are single,and the segmentation threshold needs to be manually determined.It is difficult to obtain the optimum,resulting in low recognition accuracy and failure to comprehensively consider the actual demand of cold storage operations,based on the completion of the research on the selection of parameters to characterize the dry or wet state of the carcass and the learning of the SVM classification model,this paper developed a spray moisturizing operation control system based on the dry or wet state of the pig carcass epidermis and carried out experimental verification.The specific work is as follows:（1）In view of the unreasonable design of the original image acquisition device,which affected the operation of pig carcass in and out of the warehouse,and even caused the damage of the bracket,and then affected the image acquisition,a new image acquisition device was designed and five groups of pictures of pig carcass epidermis from wet to dry were taken.After image processing,the feature parameters including total white light area,maximum contour area,maximum contour length and maximum ellipse area are extracted.The experiment on the influence of different moisturizing time on the relevant parameters shows that the change of the feature quantity of the image extracted from the processed image can correctly reflect the current dry or wet state of the pig carcass epidermis under different moisturizing time.（2）In order to eliminate the subjective adverse effects of judging the dry or wet state of pig carcass epidermis manually by the relevant characteristic parameters,and improve the intelligence and automation level of the system,a support vector machine classification model for judging the dry or wet state of pig carcass epidermis was established on the basis of machine learning theory.When the RBF was selected as the kernel function,the classification accuracy could reach 96.3%.The reliability test results show that the average recognition accuracy is 95.1%;ANOVA analysis showed that different moisturizing time had no significant effect on the recognition accuracy of classification model（P>0.05）.The recognition accuracy of the 6-hour moisture retention classification model is 95.6%,and the operation stability is 97.21%,which meets the requirement that the stability of the control system is not less than 95%.（3）According to the production practice of slaughtering pigs one by one and half carcasses in batches,according to the objective requirements of spraying operation that may need to be implemented in single row or full warehouse,on the basis of analyzing the control logic required by the system,the serial communication module setting,the selection of executive components,the control system circuit layout,the control system software and the design of human-computer interface were completed,The running effect of the system is tested and verified.The results showed that the consistency of dry consumption of half carcass among different tracks was more than 99%,which indicated that the SVM classification model of dry or wet state of pig carcass epidermis based on machine learning theory was reliable and the logic of spray operation control system was effective.The results showed that the pig carcass weight loss of 24 h was 1.02%under the optimal 5 h moisturizing time selected by the operation effect verification test,and the average total number of colonies on the carcass surface was 2.4±0.15 logcfu.cm^-2,and no coliform was detected,the weight loss of pig carcass and the bacterial content on the surface all meet the production requirements of the enterprise.