Research And Development Of Online Identification Algorithm And Equipment For Weedy Rice Seeds

Rice is one of the important food crops for human beings.Its total output is only second to that of corn and wheat,ranking third,and occupies an important position in China’s agricultural production and national economy.In recent years,direct seeding rice has become an important way of rice simple cultivation.With the increase of rice direct seeding area year by year,the harm of weedy rice has gradually emerged.It has early emergence and rapid growth,and competes with cultivated rice for sunlight,nutrients,water and growth space,which is extremely harmful.If weedy rice area is more than 40%,rice maturity will be covered by weedy rice,no target yield formation;in addition,weedy rice is easy to fall and yield is low,which not only affects the yield of rice but also affects the processing and appearance quality of rice,resulting in significant economic losses to agricultural production.Rice seeds are the key factor for the high yield of direct seeding rice.Screening weedy rice seeds and preventing them from entering the growth stage of rice are of great significance to improve the quality of cultivated rice seeds.Artificial screening of weedy rice seeds is not only susceptible to subjective influence but also inefficient.Therefore,based on machine vision and image processing technology,this paper proposes an online identification method for weedy rice seeds,and develops an online identification equipment for weedy rice seeds to improve the purity of rice seeds.The main research work and conclusions of this paper are as follows:（1）Development of an on-line identification equipment for weedy rice seeds.The equipment is mainly composed of seed metering device,image acquisition device,transmission device,rotary encoder and control system.When working,the rice seeds were poured into the seed-metering device.The main control machine was used to control the seed-metering drive module to drive the rotation of the seed-metering motor to make the rice seeds fall on the conveyor belt,and the transmission device was used to transport them forward.The control system controlled the image acquisition device by the feedback signal of the rotary encoder to sample and identify and locate the rice weedy rice seeds,and the number and location of the weedy rice seeds were identified by designing a visual interface display.（2）A concave point segmentation method based on ECMM is proposed.Firstly,the collected image is preprocessed,the adhesion contour with shape factor less than 0.4 is extracted,and the edge of the extracted contour is smoothed by one-dimensional Gaussian convolution kernel.Secondly,the curvature of the smooth contour curve and its mean curvature are calculated,and several points that are quite different from the mean curvature are found as corners.Finally,according to the positive and negative area of the vector triangle,whether the corner point is a real concave point is judged,and the angle range（0°-180°）of the normal direction composed of the concave point,the previous point and the subsequent point is searched.In this angle range,the matching concave point pair is searched to complete the adhesion segmentation.The results show that the average accuracy of the algorithm is 92.90%,which is 19.82%higher than that of the limit corrosion method and 12.85%higher than that of the watershed algorithm.（3）A rice weedy rice seed identification and localization algorithm is proposed.Firstly,a total of five samples of Linhan No.1 seed,Zhenghan No.10 seed,original upland rice seed,Danhan rice seed and weedy rice seed were selected for research,and their material properties（length,width,area,perimeter,aspect ratio,RGB color model and HSV color model）were measured,and the key characteristics of aspect ratio and R channel pixel ratio for distinguishing weedy rice seed were found through analysis and comparison.Secondly,the length-width ratio and R-channel pixel ratio of seeds in each contour were calculated,and the results were compared with the key characteristics of cultivated rice seeds to identify weedy rice seeds.Finally,the center coordinates of the contour are calculated and used as the image pixel coordinates of the seed to realize the positioning.（4）Complete the prototype processing and performance test of weedy rice seed identification equipment.The prototype of weedy rice seed recognition equipment was processed,and the recognition performance tests of Linhan No.1 seed,Zhenghan No.10 seed,Danhan rice seed and original dry rice seed mixed with weedy rice seeds were carried out.The results showed that the recognition accuracy of weedy rice seeds was 100%,100%,98.33%and 98.82%,respectively.