Research On Quality Monitoring Technology Of Amblyseius Cucumeris In Industrialized Cultivation

Using mites to control mites as an important means of biological pest control has gained rapid development in recent years due to its advantages of environmental protection,sustainability,and efficiency compared to chemical control.At present,many enterprises at home and abroad have carried out industrialized cultivation of predatory mites and made biological control products for sale.The products to be sold before leaving the factory need to undergo quality testing(quality indicators are used to determine whether the number of predatory mites contained in the feeding medium per unit weight meets the standard)to prevent inferior products from entering the market,resulting in poor agricultural control effects and affecting the promotion of biological control concepts.At this stage,enterprises still use manual testing methods to evaluate the quality of products,which is not only inefficient,but also high employment costs.With the promotion and application of biological control,the production of predatory mites continues to increase,so that artificial detection methods cannot meet the needs of industrial mass production and quality inspection.In response to this issue,this thesis has conducted the following research work with the main predatory mite species,Amblyseius cucumeris(Oudemans),which is currently cultivated in factories:(1)Mite screening and image collection section.In order to solve the problem of complex breeding environment for A.cucumeris,which leads to the inability to directly observe and capture clear mite images,a set of device with functions of mite screening,collection,transmission,and image collection was designed.In order to ensure smooth and stable operation of the equipment,the hardware was selected according to the system requirements,and a complete set of hardware control design was completed using PLC,including stepper motor configuration and ladder diagram program design.(2)Data set production section.Using the constructed image acquisition device,2390 images were taken,including two species of mites,A.cucumeris and Acaroid mites(feed).Through the use of image enhancement methods such as rotation and image contrast transformation,6214 images were amplified.Finally,the mites dataset was produced using Label Img software.(3)Live object detection and counting section.By analyzing the principles and characteristics of five mainstream target detection algorithms,and conducting experiments on this mites dataset,the results show that YOLOv5 algorithm performs better in mites detection,but the detection effect is poor when complex situations such as multiple impurities,mites adhesion,and diverse body forms of mites appear in the image.To improve the effectiveness of mite detection,an improved YOLOv5 algorithm is proposed,which introduces lightweight networks,attention mechanism modules,and model clipping operations.The optimized algorithm was put into production site for experimental verification,and the results showed that the detection performance of the algorithm for mites has been improved,with good versatility,and can be extended to apply to the quality inspection of other predatory mite species.(4)This thesis designs monitoring software for the quality control system,and completes the design of human-computer interaction interfaces such as login,control,and historical data.This thesis designs monitoring software for the quality control system,and completes the design of human-computer interaction interfaces such as login,control,and historical data.Its monitoring interface has functions such as interlocking PLC to monitor hardware equipment and call algorithms to detect and count.The history interface includes functions such as recording inspection results of each batch,and prompting for quality conformance.Through the research of this topic,first of all,a set of device with functions of mites screening,collection,transmission,and image collection was designed and completed to achieve pure mites collection and clear mites shooting targets.Secondly,an improved YOLOv5 detection algorithm was proposed and put into production field testing.The experiments showed that the proposed improved algorithm improved the detection accuracy of mites,and the correct detection rates for A.cucumeris and acaroid mites in cucumbers reached 97.16% and 96.57%,respectively.Finally,a monitoring software was designed,which has functions such as lower computer hardware control,call algorithm detection,historical data,and so on,which can meet the quality inspection requirements of the industrial production of A.cucumeris.The methods proposed in this thesis fill the gap in the field of quality inspection in the biological control and predatory mite breeding industry,and bring new impetus to the development of the industry.