Study On Key Preprocessing Algorithms Of Vision Sensing For Multiple Application Scenarios

As an extension of human vision,vision sensing has broad application prospects in transportation,security and industry.And preprocessing of vision sensing also plays a critical role.It is the necessary prerequisite for subsequent practical vision sensing applications.The preprocessing must cooperate with specific case studies due to the complexity of practical application scenarios.This has been reflected in several national and provincial projects of the laboratory,such as "Research and Application of Key Technologies in Real-Time Video Monitoring and Recognizing System for Ships","Disastrous Weather and Emergency Warning and Decision Support System for Express Ways Based on Visual Sensor Networks" and"Ceramic Tiles Sorting System Based on Vision Sensing".Based on analysis of recent research progress at home and abroad,the meaning and role of preprocessing in vision sensing systems are described.For a variety of practical applications,difficulties and encountered problems in key modules and algorithms are analyzed.Then corresponding algorithms or optimized solutions are proposed.Study on key preprocessing algorithms include irradiance calibration algorithm developed on the basis of energy function,brightness correction algorithm based on irradiance response model,High Dynamic Range(HDR)image synthesis algorithm and background brightness compensation algorithm based on irradiance response model.Theses preprocessing algorithms have been experimental verification and applied in multiple application prospects such as vision sensing based ceramic tiles sorting,express ways abnormal event detection.In outdoor cases,such as security and transportation,light conditions are often uncontrollable and with strong noise interference.In order to solve the problems mentioned above,this dissertation improves the irradiance calibration algorithm developed on the basis of energy function.Most traditional algorithms only consider from the perspective of parameter model or constraint items of convergence unilaterally and are thus very susceptible to image noise.The improved algorithm analyzes a variety of factors which affect the calibration accuracy of irradiance comprehensively and establishes a three-part general form of energy function composed by camera irradiance response curve,residual penalty term and convergence constraint terms.The optimized implementation of each part is given out,and a parameter convergence strategy considering both speed and accuracy is adopted.In order to solve the uneven distribution problem of image brightness in industry,this dissertation proposes brightness correction algorithm based on irradiance response model by analyzing light source,lens and camera luminance response characteristics.The image gray value is mapped to the scenario brightness value through irradiance response curve to eliminate complex correction calculation introduced by nonlinear characteristic of the response curve.Subsequently,the correction algorithm calibrated by gray panel is used to realize brightness correction to meet the real-time requirements of practical applications.In order to obtain brightness distribution histogram required by the brightness correction algorithm,this dissertation establishes gray panel de-fold algorithm based on edge detection,morphological expansion and space-time filtering,solving the fold problem of gray panel which is difficult to be avoided in traditional algorithms.For image noise interference is static scenarios,this dissertation improves HDR image synthesis algorithm based on multiple-exposure image sequence and balanced suppression of all kinds of image noises is thus achieved.To further expand application scenarios,this dissertation establishes a dynamic-scenario HDR image synthesis algorithm to reduce interferences,such as camera offset and the movement of objects.In the proposed algorithm,accurate image registration and fast image registration algorithms are applied to complex camera offset and slight camera offset respectively.For physical movement interferences,an automatic calculation algorithm of exposure factors is also presented.The unimodal distribution histogram of estimates of exposure factors is utilized to realize HDR image synthesis when some camera exposure factors are unknown.The automatic exposure of camera in traffic scenarios can introduce sudden change of background brightness,to solve the problem,this dissertation presents a background brightness compensation algorithm based on irradiance response model.Traditional background modeling algorithms are difficult to adapt to the rapid change of background brightness,resulting in conspicuous wrong detection problems in moving foreground scenarios.Combining irradiance response model,this dissertation analyzes changes of background brightness introduced by automatic exposure quantitatively,and then utilizes luminance compensation loosely-coupled structure to upgrade traditional background modeling methods,compensates changes of background brightness introduced by automatic exposure by calculating the irradiance response curve and exposure factors.In summary,by experimental verification and practical application,the research results of preprocessing algorithms is credible,and have important theoretical significance and application value.The main contributions of this dissertation are summarized as follows:?The irradiance calibration algorithm based on energy function construction is improved which is no need for calibration object,and the influence of image noise on the calibration accuracy is effectively reduced.?Brightness correction algorithm is improved based on irradiance response model to suppress the uneven distribution problem of image brightness by light source,lens and camera,and the fold problem of gray calibration board which is difficult to be avoided is solved.?HDR image synthesis algorithm is established and this algorithm can suppress all kinds of image noises evenly,display the image detail of high dynamic range scenes completely,and solve camera offset and motion disturbances in dynamic scenes.?Background brightness compensation algorithm based on the response model of camera irradiance is proposed to compensate the sudden change of background brightness introduced by automatic exposure,and then suppress the error detection of moving foreground,extract the true foreground region from images.