Research On Detection And Intelligent Recognition Method Of Grain Quantity Information

In our country, Grain depot’s geographic dispersion and the limitation of artificialsupervision led to many illegal affairs, such as empty warehouse, steal, falsely reporting for moresubsidy and so on. These irregularities result in huge economy loss for our nation. So in thispaper, the following two grain quantity intelligent recognition methods are proposed.The grain quantity intelligent recognition method based on SFS is an exploratory study onthree dimensional measurement of the bulk, and focuses on the camera calibration method, andthe method of obtaining three dimensional coordinates of the other points except the referencepoints using interpolation based on SFS. The work is as follows:(1) A new method of camera calibration is proposed-the camera calibration result (theinterior and external parameters of the camera) is gotten through the projective transformationbetween the sub-pixel center coordinate of the laser spot in the image and the3D coordinate ofthe laser spot in the world coordinate system.Obtaining the3D coordinate by SFS is based on one2D image,and camera calibration cancorrect the image deformation caused by camera distortions.①The3D coordinate of the spot location is calculated by use of the distance acquired bythe laser rangefinder, the rotation angle of the PTZ, the projection of the space point on everyaxis of the world coordinate system.②The spot in the image is extracted by average background model method. The sub-pixelcenter coordinates of the spot are extracted by curve fitting sub-pixel center location algorithmbased on barycentre. Experimental results show that the position of the spot extracted is right, theprecision of the sub-pixel center coordinates is high.③The projective transformation between the sub-pixel center coordinates and the3Dcoordinates of the laser spot is done, and then the interior and external parameters of the cameraare gotten.The comparison between the collected images before and after calibration shows that themethod presented in this paper overcomes the camera distortion.(2)The interpolation method is proposed based on SFS technologyThe spots within the grain boundary are regarded as the reference points, and the3Dcoordinates and sub-pixel center coordinates of the reference points are calculated by①and ②in (1); Combine these spots in the same image and the two-dimensional Delaunaydecomposition of the sub-pixel center coordinates of the reference points is done, and thus theVoronoi polygons are obtained; There is only one reference point (spot) in every polygon, andthe three dimensional coordinates of the other points except the reference points are computed byinterpolation based on SFS technology.(3)The grain volume calculation is realizedLocate the pixel location of the points on the surface of grain in the image, and thefollowing calculation is done line by line: Each point and its three adjacent points on its right,right down and down constitute a hexahedron which is approximately regarded as a cuboid, andthe sum of all the cuboids’ volume is the volume of grain.The experiments in and outside the laboratory show the error of the grain quantitymeasurement method based on SFS technology has a certain fluctuation, but the scope of theerror is acceptable. The value of standard deviation meets the project requirements.Grain quantity recognition method based on laser3D coordinate instrument focuses on thecalculation method of3D coordinates of spot position, the invariance of rotation and translationof3D coordinates, the preprocessing of laser scanning3D data, and grain total volumecalculation method. The following work is done:(1)The calculation method of the3D coordinates of the spot position is inferredAfter the coordinate system of the laser3D coordinates instrument and parameters are set,the3D coordinates of the spot position is inferred using the position relationship between therow scan and field scan motors, the change of the angle when the motors rotate, and theprojection of a space point on each axis in a coordinate system.(2)Rotation and translation invariance of3D coordinate are realizedThe rotation matrix and translation vector between the coordinate system of thethree-dimensional laser detector and the world coordinate system are calculated by use of theprojection on the axes of the world coordinate system of the data point in the coordinate systemof the three-dimensional laser detector. Then the coordinates of the data point are converted tothe world coordinate system, that is, the rotation and translation invariance are implemented.(3)The preprocessing method of laser scanning3D data is proposed①Zero point correction. The zero points in the datum collected are interpolated to restore using piecewise linear interpolation.②Useless points are removed. Useless points include the points on the beam and wall. Theisolated points (mutations in the data points) on the beam are directly removed. The points on thewalls may be judged by x or y coordinates of the points being equal. The points on the walls areremoved.③The judgment whether there is the shelter on the grain pile and the restoration of thedatum sheltered. The points on the grain surface include the points on the plane and on theprotuberances. Owing to the location and height of the laser coordinate instrument, grain pilemay be partially sheltered. The judgment of the shelter is done according to the differences of theprojections on the x axis of the points on grain pile when there is shelter and not. The datarecovery of the part sheltered is done by cross ratio theorem in projective transformation andsymmetry of the grain pile.(4)Grain volume calculation is realized by Delaunay decompositionThe pre-processed3D datum are decomposed by Delaunay triangulation, and the cylindersare gotten whose tops are triangles. The volume of every cylinder is accumulated, and then thetotal grain volume is obtained.Experiments of the grain in the depot show that the error of grain quantity intelligentrecognition method based on laser3D coordinate instrument is more stable than SFS technology.The error is about3%and meets the management requirement of the audit department basically.The two grain quantity recognition methods based on SFS technology and laser3Dcoordinate instrument do not require human intervention after the device is installed. Theintelligently contactless detection of the grain quantity is realized. The costs of supervision andauditing can be greatly reduced, and the market prospect is broad.