| With the rapid development of artificial intelligence technology,machine vision and image processing technology are widely used in various industries.The evolving image processing technology is also gradually increasing the requirements for image processing quality and speed,and needs to meet the needs of long-distance and high-speed rate transmission and storage.In addition,image edge detection and segmentation technologies have great potential for applications in smart manufacturing,agriculture,medicine,and automation.Therefore,in order to realize the requirements of high-precision detection,strong real-time and long-distance and high-speed transmission in image processing,this paper investigates the key issues of image edge detection,image segmentation and image data transmission on the FPGA platform.The research mainly includes:(1)In order to solve the problem that the traditional Sobel edge detection is inaccurate in positioning and cannot automatically change the threshold value according to the external light,that is,the threshold value changes with the change of gray level,an adaptive threshold edge detection algorithm based on FPGA is proposed.The algorithm uses an 8-directional Sobel operator to separate out the edge information of the image and uses the mean value as the adaptive threshold for edge detection,achieving adaptive changes in the threshold as the grey level changes.At the same time,this paper designs an FPGA-based image detection system,which completes the functions of real-time image data acquisition,detection and display,and realizes adaptive threshold Sobel edge detection of complex image information.Among them,for the problem of inaccurate positioning of traditional edges,this paper adds operators for convolution in 45°direction and 135°direction,and adopts double thresholding to do edge judgment,which greatly increases the completeness and accuracy of edge details.To address the problem that edge detection thresholds must change with external factors,this paper splits a frame into a number of pixel windows of a 3×3 matrix and uses the mean value as the threshold for edge detection,realizing that the threshold changes with the change in grey level.(2)A local threshold segmentation method based on FPGA is proposed for the problems of uneven light distribution and reflection in image segmentation.The method uses the idea of a mean value of local neighborhood blocks and a Gaussian-weighted sum of local neighborhood blocks to solve the problem of large changes in local grey levels and to achieve local threshold segmentation.To address the problem that the algorithm is complex and not easy to implement on FPGA,the algorithm is disassembled and analyzed,and the window cache module,data superposition module and mean filtering module are designed.Finally,the designed algorithm is verified by MODELSIM simulation software,and the fixed thresholding algorithm and OTSU algorithm are designed on VS2015 for comparison tests on images.The experiments show that the algorithm can effectively reduce the interference caused by uneven illumination and reflection phenomenon on the image and improve the segmentation effect.(3)The design of an image processing system based on FPGA Ethernet transmission is proposed for the problems of addressing the issues of how to transmit and store images over long distances and at high rates.In this paper,a long-distance,high-efficiency transmission storage system is designed.The transmission of image data is realized through the UDP protocol in Ethernet.Firstly,the image data is captured via the camera OV5640;secondly,the captured data is pre-processed;then,the processed image is cached via SDRAM;finally,it is transferred to the host computer for storage via the network cable using the UDP transmission protocol.The method is suitable for large data transmission volumes,long transmission distances and high transmission efficiency,and can save the processed image data in real time,laying the foundation for subsequent processing. |
PDF Full Text Request