Research And Application Of FPGA-based Multi-motion Target Detection Technology For Intelligent Cutting Machines

With the rapid development of digital video technology and industrial intelligence,the manufacturing industry is increasingly demanding automated analytical image acquisition equipment.To achieve intelligent image acquisition in industry,image acquisition equipment needs to be able to detect moving target images in real time and digitize the images using a computer.In intelligent cutting and transport scenarios,the system needs to recognize and detect moving objects in order to control and regulate the motion of the target object.Therefore,this paper needs to analyze and solve the problems arising from traditional motion target detection algorithms in this transport scenario,where the research problems are focused on the following aspects:In the implementation stage of the image pre-processing algorithm,the following methods are used to improve the hardware computing speed and save hardware resources in view of the hardware design characteristics of the FPGA: in the image format conversion process,shift operations are used instead of floating point operations for the format conversion of image data in order to reduce the amount of data for later image processing;in the filtering operation and image segmentation processing,a convolution kernel is designed to calculate the gradient value of the In the process of filtering operation and image segmentation,the method of calculating gradient values by convolution kernel is designed to meet the hardware requirements of digital image processing;in the process of morphological image processing,based on the characteristics of digital logic,the method of logic gate circuit is designed to realize the design requirements in order to achieve the requirements of eliminating noise and bridging breakpoints.In the analysis stage of the single motion target detection algorithm,in order to solve the problems of the traditional inter-frame differential algorithm in the lack of edge information in the motion vertical direction of the moving target and the low recognition of the moving target due to the fixed threshold,this paper proposes an improved three-frame differential motion target detection algorithm.The algorithm uses three adjacent frames for differential operation and fuses edge detection algorithms to achieve real-time detection of single moving targets.In addition,the algorithm adds differential adaptive thresholding and morphological filtering operations to improve the accuracy of the algorithm.Experimental results show that the algorithm effectively reduces the effect of changes in the light intensity in the environment on the results,overcomes the problem of missing edge information of regular moving objects in the vertical direction of motion,and also identifies the texture information on the surface of moving objects more clearly.Compared to traditional motion target detection algorithms,this algorithm can better detect and identify motion targets.In the analysis stage of multi-motion target detection algorithm,in order to solve the problem of industrial transportation system needing to identify multiple motion targets and the problems of traditional digital image processing platform in terms of large space occupation and weak system expansion function,this paper proposes a multi-motion target detection algorithm based on FPGA.The algorithm adopts an improved four-frame inter-differential operation and utilizes a multi-port SDRAM transfer storage technique to achieve real-time detection of multiple motion targets.Experimental results show that the algorithm can effectively achieve real-time detection of multiple motion targets and also meet the accuracy of multi-motion target recognition.