Research On Methods Of Lingwu Long Jujubes’ Real-time Localization Based On True Color Images

Lingwu Long Jujube,an important economic forest-fruit crop in Ningxia,has short picking period about 20 days.At present,the packing way of Lingwu Long Jujube depends mainly on manpower.Which is disadvantage with inconvenient for picking,heavy labor intensity and low packing efficiency.Therefore,there is an increasing demand for automatic picking technology of Lingwu Long Jujube.The key of the automatic picking technique of Lingwu Long Jujube is how to determine the position of Jujubes.So this paper studies mainly the real-time localization method of Lingwu Long Jujube based on true color image under natural environment to lay a theoretical and technical foundation for the development of Lingwu Long Jujube automatic picking robot.The main research contents and related result are as follows.1.Pre-processing of Lingwu Long Jujubes,which put forward a image filtering method and improved image enhancement technology.The method of the image noise reduction based on optimal channel-processing quadratic optimization.For the purpose of optimal noise reduction effect,the first step is to reduce the dimension of the mixed noise(reduce the types of noise or reduce the parameter)to achieve the effect that noise tends to be consistent.This is the first optimization.The second optimization is to eliminate the coarse error point to avoid the impact of larger error points.Compared with mean filtering,median filtering and adaptive filtering,the image quality processed by the proposed algorithm is generally improved by more than 1.2.Due to the influence of light,the color information of the collected image will change.So it is very important to maintain the invariability of the image color information.For the polarization distribution of the image gray value with brightness affected by light(distributed in the low pixel or high pixel area),a mapping function is designed based on the power function.Which maps the pixel values of the polarized distribution to the normal region.Through experiments it has been found that the effects of light can be overcome to some extent2.According to the color characteristics of Lingwu Long Jujubes,three image segmentation algorithms are proposed under natural environment The running time and the correct recognition rate of proposed algorithms can meet the requirements.A image segmentation method of Lingwu Long Jujubes based on a new extraction model of hue is proposed.a extraction model of hue aiming at images of Lingwu Long Jujubes based on stage treatment of R component is built to extract hue information.And the difference between the target object and background object is increased by using color difference.A color image segmentation method of Lingwu Long Jujubes based on the maximum entropy is proposed.It gets a better hue information by extracting hue image with adjustment Then the image is segmented by optimal threshold obtained by combining the maximum entropy and the mathematical criteria to achieve the adaptive adjustment of the segmentation threshold.Finally,segmented image will be obtained through dealing with mathematical morphology.A image segmentation method of Lingwu Long Jujubes based on marker-controller watershed algorithm is proposed.Which segments image accurately by funding marker-image and mask-image.By comparing the segmentation effect of 30 Lingwu Long Jujubes images with artificial method and other methods,the accuracy of segmentation rate is up to 92.6883%,89.60%and 92.525%respectively.The running time that proposed two algorithms is 1.3107s,1.3132s and 1.21792s respectively.3.the premise of accurate location is to be able to segment accurately the image of Lingwu Long Jujubes and determine the position of it in the two-dimensional image.Here,the centroid of Lingwu Long Jujube is took as the position of it.In this paper,local information is used to count the number of Lingwu Long Jujubes.Firstly,the image of Lingwu Long Jujubes is corroded by mathematical morphology to make the different dates not communicate with each other.and the local connected regions are counted.Using the corroded parts of Lingwu Long Jujubes determine locations of local centroid.In fact,the centroids found at this time have large errors,but each centroid position must be located inside of Lingwu Long Jujubes.Using the statistical information of regions determined by adjacent local centroids get the boundary with adhesion.And then the respective area of each Lingwu Long Jujubes is identified.Through the experiment of 30 images,the accurate recognition rate of Lingwu Long Jujube is more than 93%.In this paper,the research results solve the real-time localization of objects,which will lays the foundation for the visual system of picking equipment.