Research On Identification Algorithm Of Maize Seeds Purity Based On Color Clustering

The false seed or the seed of substandard purity which can result in crop decrease is one of the most important factors in our country even in the world.The key link from the pure Maize Seeds to the substandard purity is between breeding seeds company and retail enterprises.The identification method recommended by The ICC(International Grains Council)?The ISTA(International Seed Testing Association)and The MOA(Ministry of agriculture of the People's Republic of China)is The PAGE(Polyacrylamide Gel Electrophoresis).The method is complicate in operation,high in professional knowledge,long appraisal period and not suitable for on-line identification and field recognition of hybrid seed breed.In order to solve the problem of no portable equipment for the purity detection in the maize seed circulation,to improve the purity of the corn seed,to guarantee the planting of the maize crop and to raise the farmers' income,the portable maize seed purity identification instrument was developed and developed,and the method of the purity identification of the maize seed was studied.The main research contents and innovations include:(1)The research and development of portable maize seed purity identification instrument.By applying the demand analysis and design,maize seed purity identification device mainly consists of two parts of hardware and software.The hardware is mainly composed of image imaging system,illumination system,maize seed kernel transmission system and control system.This instrument can automatically collect the tip of the seeds of multiple maize seeds and the image information of the non-embryo side,and the image is uniformly preserved in BMP format.The software algorithm is made up of image enhancement,color feature extraction optimization and purity cluster recognition algorithms.(2)This part studies the pretreatment of the top image of maize seed and the image of the non-embryo side,and proposes the method of using gray scale adaptive stretch to enhance the single-channel enhancement of the maize seed image,then the three channels are fused,and the contrast enhancement of maize seed image is realized,which provides reliable data source for the extraction and identification of grain color.(3)It studies that method to complete the division extraction of maize seed image area.To improve the adaptability of the system,the method of taking a single threshold is to split the top image of the maize seed and the image from the side,and the four color regions of the seed are extracted;By contrast analysis,the average size of the RGB,HSI and the Lab is the color parameters of each region,and the maize seeds are extracted in four color areas,and there are four areas,nine eigenvectors,36 dimensions of color information as the color signature of a maize seed.(4)Research completed 36 dimensions to optimize the extraction of maize seed color characteristics.A two-step optimization method was proposed to optimize the color information,and the variation coefficient and significance analysis were used to eliminate the color information of the irrelevant color information and the low degree of relevancy,and the 36 dimension color characteristics were reduced to 20 dimensions.It use multiscale wavelet decomposition packet to extract that color information of 20 dimensional,and then makes a significant analysis of the 48-dimensional color information obtain by the decomposition,and retains 20 dimensional color details as the recognition classification vector.The purity identification clustering algorithm of maize seed was studied.The method of probability purity of the multi-cluster model,based on the classic k-means,SOM and the two-step group of corn seeds,to identify the corn seeds,and to increase the recognition rate as a final clustering category in the probability scale,to finish the method of the seed purity recognition of the corn seeds.Through the actual verification of the corn seeds purchased by the market,the purity identification of a batch of maize seeds can be completed,which is superior to the traditional identification method,and the identification precision reaches the inspection and quarantine standard in the circulation place,and the target is set.(5)The purity identification algorithm of the portable dynamic corn seed purity identification instrument is designed and tested.First,filter the clustering data,then Multi-cluster Probability Model is used to identificate the purity of 8 maize seeds purchased on the market.And the same time,Identification of purity of 8 maize seeds by SSR.Analyze the significance between the purity identification data by Multi-cluster Probability Model and SSR.And we found no significant difference in their purity means.This method can be used to identify the purity of maize seeds.