| The accurate extraction of the vehicle target area is the key to the detection of expressway anomaly events based on video surveillance.In the expressway scene,the existing vehicle shadow interference area,making the vehicle area distortion,expansion,connectivity or even lost.In addition,there is much noise interference in the expressway scene image and the image quality is lower,resulting in that the traditional shadow suppression method is still difficult to apply.Therefore,it is of great theoretical and practical significance to study the method of vehicle shadow interference suppression in highway environment and improve the accuracy of vehicle target detection.Through analyzing the characteristics of the highway under the scenario of shadow interference,some important researches in background modeling,the suppression of the shadow interference in the open-air scene and in the tunnel scene have been done.And the vehicle target shadow interference suppression scheme in highway scene is proposed.For background modeling,an improved background modeling method based on time-domain inter-frame color gradient mean feature is proposed.For the gradient mean feature can't be well defined foreground and background pixels,resulting in poor modeling results.By using of the target pixel value in the three channels in the maximum rate of change to establish the pixel is the foreground or background confidence to determine its Background model reliability.At the same time,in order to reduce the interference of noise to background modeling,the mean value of the maximum value of the three-channel color gradient in the local area constitutes the weight of the background modeling contribution rate,thus improving the effectiveness of background modeling.The comparison experiment shows that the modeling method can weaken the influence of foreground pixels on background modeling,and can meet the requirements of real-time and accuracy.For shadow suppression in open-air scene,aiming at the noise interference in open-air scene,the shadow suppression method which fused the color grayscale feature and HLGP(Histogram of Local Gradient Patterns)features is proposed.In this method,the shadow is first determined by chromaticity and brightness similarity.For the local area with rich texture information,the local texture is extracted with noise robustness LGP(Local Gradient Patterns),and the local gradient histogram is used to compose the HLGP feature.Degeneration was proved.On this basis,the result of color shading is corrected to reduce its false detection area.The experimental results show that the method can overcome the noise interference and obtain high shadow detection rate and foreground discrimination rate.For shadow suppression in tunnel scene,aiming at the serious noise and blurred image quality in the tunnel scene,the shadow suppression method based on the LPQ(Local Phase Quantization)feature and wavelet decomposition is proposed.This method introduces the LPQ feature analysis with fuzzy robustness and illumination invariance,and uses the wavelet analysis method to extract the details of the fuzzy information.Through the image merging operation on the results of the two extracted,so as to obtain a more accurate shadow area and vehicle target area.The results of the comparison show that the method can take the shadow detection rate and the foreground discrimination rate accurately and the vehicle target area can be extracted accurately when the color information is insufficient and the picture quality is blurred.Finally,the combination of the above research results,formed a set of the shadow interference suppression technology in expressway.And the experimental data of expressway video data collected by field are validated.The experimental results show that compared with the traditional shadow suppression method,the proposed method has better effect of shadow interference suppression in the highway scene,so as to support the detection of more accurate vehicle target area. |
PDF Full Text Request