Counting Of Overlapping Quasi-circular Granule Based On Image Processing And Research And Development Of APP

The identification and counting of quasi-circular granule targets is of great significance in practical applications.The traditional image processing system is implemented on the PC side.Such systems have limitations such as complex construction,high hardware cost,inconvenient movement and strong professionalism This limitation is particularly prominent in fast real-time counting occasions Therefore,based on the Android system,this paper designs a portable mobile phone counting APP for the recognition and counting of round-shaped particle targets.The APP has many advantages such as strong real-time performance,portability,strong portability,low cost and favorable promotionBased on digital image processing technology,this paper studies the related theoretical algorithms for counting quasi-circular granule targets.Based on digital image processing technology,this paper studies the related theoretical algorithms for counting round particle targets.In the pre-processing part of the target image,the image enhancement is first performed on the target image,and the top-Hat transform and the gray-scale stretching are used for the correction of the illumination unevenness in the image acquisition process.Since the edge information of the quasi-circular granule is needed in the subsequent image segmentation of overlapping particles,a double filtering algorithm is proposed for the smooth denoising of the target image.This method protects the edge information of the particle to some extent while removing noise.Then use the Otsu algorithm to perform threshold segmentation,and improve from the threshold criteria and the block processing to obtain the final binary image;Finally,a filter is designed to perform morphological processing on the binary image.This method eliminates the partial adhesion between the particle targets while keeping the particle target edge profile deformation to a minimum.In the segmentation part of the target image,the overlapping particle targets are separated and extracted from the original image by the shape factor.For the non-overlapping quasi-circular granule targets,the H-minima transform-based marker control watershed algorithm is used for segmentation,for overlapping quasi-circular granule targets.The marker-controlled watershed algorithm based on fuzzy theory gradient convolution kernel is used to segment.Considering the ambiguity of the shape and size of the quasi-circular granule,the algorithm introduces the membership function and designs a fuzzy gradient convolution kernel,which is made by convolution operation.The edge pixels of each particle target are enhanced for the degree of deviation from the standard circle,and the mark-controlled watershed algorithm is used to complete the effective segmentation of the overlapping quasi-circular granule targets.In the recognition and counting part of the target image,the segmented single particle is identified and counted by using the backtracking detection algorithm based on the connected region,and the result is imported into the mobile phone database and the PC end database for subsequent query and statisticsBased on the research of image processing algorithms,this paper designs and implements the functions of each module of APP based on Android development technology and OpenCV visual library,and tests the performance of the counting application system.The final result shows that the APP runs stably,there is almost no error in counting individual particles and adhesion particles,and the average accuracy of counting on overlapping particle targets is about 97%.The system can complete the basic quasi-circular granule counting problem under the requirement of portable and low cost,and has good practical value.