Based On Performance Testing Of The Precision High-speed Camera Metering Device

The homogeneity of seeding is one of the most important specifications to reflect the work performance and quality of the precise seed metering. The parameter is difficult to detect. Its detecting method directly affects the evaluation of precise seed metering, and influences not only the equilibrium of plant absorbing nutrient and the grain yield but also the producing cycle and manufacturing cost of precise seed metering. So it is very necessary to find a simple and feasible method to detect the properties of seed metering.In order to improve the detecting precision of the properties of seed metering and the automation of seed metering, by analyzing the development of computer equipment, CPU, RAM of large capacity, HIGH SPEED VIDICON and showing system, the author proposed a detecting technique of the properties of seed metering based on computer vision. Basing on computer vision, wavelet transform, mathematical morphology and genetic algorithm and so on, the author solved the key technique problems about using computer Vision to detect the properties of seed metering, and developed the detecting software package of the properties of seed metering (VDSSP). The feasibility and accuracy of the system is shown by experiments. The main research work and conclusions are as following:1.In this paper, a detecting system of seed metering properties based on computer vision is established. The author analyzed image-grabbing features of objects, and pointed out the key techniques of establishing this system. An image grabbing system which is consist of a camera and a mirror is established, so the motion law of seeds can be represented by grabbing the motions of falling seeds from different views. Then the author developed a software package of this detecting system (VDSSP). The functions of software used in the system are described in this paper.2.In this paper, the method to detect the motion objects (seeds) based on a kind of complicated technique is proposed. First, difference methods to processing images are applied to removing the complex background to catch motion object quickly. The changed velocity linear motion IS regarded as uniform velocity linear motion, so the mathematics model to restore image blurred by seed motion is established to obtain the initial position of seeds in the images and calculate coordinate in the images. Second, according to the distribution characteristic of signals and noises in the frequency field, the mixed signals with variousfrequencies are decomposed by wavelet transform in different frequency block. Then noises in the images are reduced by different change tendency of the maximum value of module for signals and noises in wavelet transform. The noises generally come from experiment desk vibration, natural light and surrounding environment. Then author proposed a self-adjusting image segmentation algorithm based on between-class inside-class difference and mathematics morphology. With defining series and parallel connection of objects in seed images, the seeds, which are superposed due to object shelter, are detached successfully. Last, the method to searching mass center of seeds quickly based on matrix method is also established.3.The author brought forward the multi-sensor calibration method based on the compute vision. According to the imaging traits of the camera and the plane mirror, the non-linearity method is used to establish the mathematic model, which is provided by the image grabbing system of a camera and a plane mirror. And the parameters of the grab image system are obtained. At the same time, the author took into account of the lens distortion of the camera and corrected the parameters of the grab image system. The calibration of image grabbing system is more scientifically and more accurately by this method. So the precision of the calibration and the detection are improved in this system by this method.4.The point-pattern matching algorithm of the single-frame seed image is brought forward based on the genetic algorithm. The distributing traits of t...