Research On Image-based Biochip Spotting Process Detection Method

In recent years,biochip detection technology has been widely used in many fields,and the quality of biochips determines the accuracy of detection results.Based on biochips produced by mechanical spotting method,combined with image-related technology,the detection methods for the problems involved in the process of biochip spotting are studied,which is significant for improving the detection efficiency and precision in automated manufacture.Research on image-based methods of biochip spotting process detection has been done,whose detection contents mainly include the following three aspects:(1)Detection of biochip workpiece size and position misalignment.The image acquisition device was designed and constructed,and the camera was calibrated by the standard checkerboard method to realize the conversion between the actual physical size and the image pixels.Then,the image acquisition device is used to acquire the biochip workpiece image.For the problem that the chip workpiece area and the background area are similar in gray scale,the images are enhanced and the contrast is improved between the target and the background.Aiming at the noise sensitive problem of traditional differential edge detection operator,the active contour model(snake model)is proposed to extract the workpiece edge.And then the contour model is initialized by Canny edge detection and the curve energy function is minimized by iteration.Thereby the target edge contour is obtained.Finally,the straight line is fitted by the least square method to obtain the line parameters,and the workpiece size and the deviation angle are calculated.(2)Detection of the center position of the biological chip spotting point.For the problem that the spot-point target is small and noisy,a sub-pixel-based edge detection method is proposed.Firstly,the gray-scale threshold segmentation method is used to extract the area around the point and the Canny edge detection is used to perform pixel-level edge positioning on the spot.Then the sub-pixel edge localization method of fitting method,interpolation method and moment method is studied and analyzed.The sub-pixel edge positioning of the spot is performed by the quadratic curve fitting method of gray gradient direction.In order to solve the problem that the edge is not continuous for the positioning,the feature is to divide and merge the edge arc by a certain rule to get the arc that fits the actual edge.Finally,the obtained arc is fitted by the geometric least square method to obtain the center coordinates of the spot.(3)Detection of classification of biochip products.For the biochip's quality detection problem,the method of convolutional neural network is used to classify and detect the spot.By employing Faster R-CNN as the basic model,the feature extraction network structure is improved according to the image characteristics of the biochip spotting point.The improved VGG-16 network is used as the extraction network for the spotting feature.We redesign the candidate anchor frame size according to the shape feature and use the improved network model to classify the biochip spot points.The results show that the accuracy of the method can reach 97.33% and can effectively detect the spot points of biochips.