Research On Eligible Detection Method Of Syringe Needles Based On BP Neural Network

In order to prevent the spread of AIDS and other infectious diseases, disposablesyringes plays an important role. However, disposable syringes in the productionprocess will generate about one millionth of defective needles, how to remove thesedefective needle is an important research direction of needle detection. Aiming at theproblem of the qualified examination syringe needle, we put forward a eligibledetection method of syringe needles based on BP neural network. The main contentsare as follows:(1) In order to acquire higher quality images of disposable syringe needle, we setup the needle detection hardware environment.(2) We study the method of needle image preprocessing. In order to get a clearimage of the tip profile, we design a set of effective pretreatment process, includingimage filtering, image segmentation, pseudo object removal, needle contour extraction,image correction and needle extraction.(3) We study the method of needle feature extraction. Based on the analysis ofeach feature of the qualified needle, bent needle and inverted needle, we choose thelow order moments of image boundary region invariant moments Φ1、Φ2、Φ3, and tiparound the edge of maximum curvature Kmaxland Kmaxras the five characteristicparameters of the main tip, and give the calculation methods of the five parameters.(4) We study the method of BP neural network detection. First, we brieflyintroduces the basic principle of BP neural network; secondly, we design a three layersBP neural network for the eligible detection of disposable syringe needles; third,training samples, the designed BP neural network were trained by multiple sets ofdifferent needles, which made it have the ability to correctly detect needles eligibility.(5) In order to verify the effectiveness of the detection scheme in many aspects,this paper designed four groups of needle eligibility experiments: qualified needledetection experiment, flip needle detection experiment, curved needle detectionexperiment and mixed needle experiment. The first three experiments for three kinds ofneedle detection experiment respectively, to verify the effectiveness of the scheme tosingle type needle detection; mixed needles experiment subjects for three kinds ofmixed needle, to verify that the proposed scheme for different types of needles mixed still had the validity. Test results show that the flip and curved needles had100%detection rate, which ensured the absolute safety of products; And qualified needledetection rate reached more than98%, so this scheme can be used in practicalproduction. The effectiveness and superiority of the image preprocessing method andthe design of BP neural network is validated through the experimental results, thismethod had high detection accuracy, which can be applied in practical production.