Research On Rapid Detection Method Of Salmonella In Food

As the major enteric pathogens, Salmonella are widely distributed in nature,easily cause food poisoning.However,with the shortcomings of long detection period,complex operation,low sensitivity and specificity,the existing detection methods for Salmonella are difficult to be used in field detection.To solve such a problem,this paper presents a rapid detection method that is based on immunogold labeling and computer vision technology.In the paper immunogold nanoparticles is used as seed crystal,and Ascorbic acid is chosen to be growth solution of reducing agent.Homogenous growth of gold nano crystals is used in Salmonella specific staining.The optimum staining conditions are o btained in experiment.In the staining process:2μl specimen is placed on the glass slid e.After the specimen is fixed 6～9min by 75% alcohol,2μl primary antibody(1:2000) and immunogold nanoparticles(1:40) are successively dropped into the specimen follo wed with 30min reaction time at 37℃.And then 4μl growth solution is dropped into the specimen in twice with a total reaction time 4～8min.The specimens with clean b ackground and clear stained thallus can be fabricated in this method,and no cross rea ctionwith other microorganism is found.The paper developed a rapid detection depending on computer vision technogy.Itcan do microscopic image acquisition and processing automatically in real-time,whichmake quantitative detection of Salmonella come true.This whole detection system consists of control module,video acquisition module,image processing module and patternrecognition module.The upper PC control the movement of the microscope stage with the help of Micro Controller Unit(MCU), and monitors the video acquisition module to take microscopic images according to the position feedback.Video acquisition module is developed by DirectShow streaming media.And the CaptureImage Class is designed to collect,save and display the microscope images.In the image processing module,filtration combined with gray level transform method is used in image preprocessing to reduce background noise.The Salmonella cell images are separated from background by iterative threshold method.Then the morphological open-close operation and connected component labeling methods are introduced to segment images,wipe off isolated impurity images, smooth cell image edges and connect breaks.At last,eight-chain code tracing method is used to calculate form factor,rectangularity,eccentricity,and complexity of the image edge of Salmonella cells which is taken by fill hole etching.Pattern recognition is performed by BP neural network.And the BP neural network recognizer is designed based on the training set,the recognition purpose and the influence ofnetwork parameters to the network performance.The model test results indicate that the recognizer's accuracy for forecast is up to 96.82%.Finally,the software design of the whole detection system is completed since the network recognizer is added into the project by Matlab engineer port.The contrast experiment is introduced to analyze the accuracy.And the results ind icate that this rapid detection could acquire ideal outcome with a remarkably high co rrelation coefficient(R2>0.99)with traditional methods.In addition,low significant differe nce(P>0.05)between results,short detection time(2h),high repeatability,low minimal dete ction limit(<10cells/ml),make it appropriate to be used in practical detection of Salmo nella.