The Object Recognition Method In The Needle-like Crystal Image

Crystal industry is an important part of the world economy, crystal products has a wide range of applications including pharmaceutical, food and chemical industries etc.Since the measurement of crystal size distribution is necessary for efective control, several measurements have been developed recently. At present, vision measurement method is attacting more interests in the crystallization process, in which images of crystals are obtained and interested information is extracted from the images.In this dissertation, the object recognition method in the needle-like crystal image is researched based on machine vision and digital image processing technology.The main contributions are summarized as follows.(1) The development and application of the control and measurement technology in crystallization process is overviewd and the machine vision technology is introduced. An image acquisition environment is designed according to the shape characteristics of needle-like crystals for the convenience of the subsequent processing.(2) A segmentation algorithm based on gray distribution steepest descent method is proposed. The image segmentation threshold is selected to be the fastest decline point on the probability density function of the crystal image's gray distribution so as to avoid the over-segmentation of the crystal image and reduce the loss of the crystal object infomation effectively. Furthermore, the enhancement method based on morphological is applied to recover the lost crystal region and enhance the linear features of the crystal object edge.(3) A feature extraction method is proposed by fitting the rectangular profile of the crystal. The length and width of the crystal region is defined and extracted through the edge and the characteristic line of the crystal region to fit the rectangular outline of the needle-like crystal.The target region of the crystals can be hit accurately by this method. (4) The overlapped phenomenon in the needle-like crystal image is analyzed and an improved segmentation method is proposed based on the concave point matching. A discriminant condition of split-line based on characteristic line of the crysral region is adopted to achieve perfect division of the overlapped region.(5) An object recognition system for the needle-like crystal on the basis of the proposed method is developed and the effectiveness of the method is demonstrated by the experimental results.