Research On Parameters Acquisition Of Maize Seeds’ Traits Based On Computer Vision

Corn is one of the most important foods and the most widely produced feed grain in the world.The assessment of seed is about the color,shape and damage,which is an indispensable step in the cultivation of high-quality varieties.The corn is traditionally detected by observing with the naked eye or measuring with simple tools such as balances,vernier calipers,etc,which is so repetitive and tiring that the error rate is high.With the rapid development of computer vision technology and the wide demand of corn assessment,it is necessary for measuring corn automatically to apply computer vision.Therefore,a corn assessment system based on computer vision technology is designed in this paper,which can result in automatic and high-throughput acquisition of corn phenotypic parameters with high precision,high efficiency and automation.The main research contents in this paper are as follows:(1)Image processing and parameter extraction algorithm on corn ear.Firstly,each corn ear was segmented integrally.This step included image preprocessing,segmentation and recognition,where the most critical procedure was identifying and separating the touching corn ears.The segmentation result directly affected the measurement accuracy of the corn phenotypic trait parameters.The test result showed that the average correct rate of the proposed algorithm for maize ear recognition is 96%.Then determining the method for extracting the ear parameters which included the ear number,ear length,ear width and bald tip length.The test result showed that the average accuracy of the method for measuring the ear number,ear length and ear width are 0.5%,1.13%and 2.96%,respectively.The proposed method for measuring the bald tip length was most convenient at present.The above results showed that the algorithm was high-precision,high-efficiency and real-time.(2)Image processing and parameter extraction algorithm on corn kernel.Firstly,each corn kernel was recognized correctly.This step included image preprocessing,segmentation and recognition,the key of which was edge enhancement and kernel recognition.The recognition accuracy directly affected the measurement accuracy of the corn kernel phenotypic parameters.Eight corn varieties were selected to test the proposed algorithm.The single test showed that the average recognition accuracy exceeded 93.6%and the batch test showed that the average correct rate was over 94%,which was about 10%higher than the HoughCircle detection algorithm.Then the method for extracting the kernel parameters including kernel number,kernel length and kernel width was determined.The results showed the method was effective and could meet the requirement of real-time detection.(3)Completing the whole design of corn assessment system which included image acquisition,image processing algorithm integration and user graphical operation interface.The test on the stability and compatibility of the software system showed that the software could met the users’ requirements.