Papers by Author: Hong Ke Xu

Abstract: In this paper, an improved algorithm of general radial basis (RBF) function neural network is introduced, based on improved algorithm, the neural network realized quickly fault diagnosis and self-update of neural network structure, and the neural network is applied to the on-line fault diagnosis expert system. The expert system deals with the fault data that send from on-line monitoring equipment by using neural network, and it can discover the fault type and give reasonable solution by forward reasoning. Meanwhile, the expert system has the ability of achieving new knowledge based on the application of self-update ability of RBF neural network.

Abstract: In order to study the rapid and efficient identified method of accident-prone section in montane highway, the method of principal component - gray clustering analysis has been proposed. By deep analysis of the characteristics of accident-prone section, the identified indexes of accident-prone section have been screened out, the reducing dimensionality of principal component analysis and incomplete information processing of gray clustering analysis have been organically integrated, and the clustering weight coefficients are creatively determined based on the information content. Based on data investigation and treatment, using the identified method of principal components - gray clustering analysis, the security level of sections is achieved by programming. The results show that this identified method has high precision and convenience in aspects of aggregative indicators selected and clustering value calculated. The identified method can effectively identify the security level of accident-prone section, and divide the section security level into 4-grade. Aiming at the identified results, the security measures are further researched. So the identified method has practical value.

Abstract: The current self-organizing feature map (SOFM) neural network algorithm used for image compression, of which a large amount of network training time and the blocking effect in the reconstructed image existed in codebook design vector calculation. Based on the above issue, this paper proposed an improved SOFM. The new SOFM introduced normalized distance between the sum of input vectors and the sum of the codeword vectors as a constraint in the process of searching for the winning neuron, which can remove redundant Euclidean distance calculation in the competitive process. Furthermore, this paper has done image compression by combining wavelet transform with the improved SOFM (WT & improved SOFM). The method firstly conducted wavelet decomposition for the image, retained low-frequency sub-band, then put the high-frequency sub-band into improved SOFM network, and achieved the purpose of compression. Experimental results showed that this algorithm can greatly reduce the network training time and enhance the learning efficiency of neural network, while effectively improve the PSNR (increased 0.6dB) of reconstructed.

Abstract: In order to improve the safety of driving under different visibility meteorological conditions on the expressway, this paper analyses the composition and operation principle of the expressway lane departure warning and navigation system based on machine vision. And an effective lane mark identification algorithm suiting for different visibility situations is proposed. Firstly the image contrast between the expressway lane and its background is increased by using histogram equalization technology, improving detecting range of lanes and accuracy. Secondly the edges in the lane directions are detected by utilizing specific Sobel operator. In order to suit for different visibility situations and improve detecting efficiency, the method of maximum classes square error is applied to threshold segmentation. Finally, lane is abstracted according to expressway lane features after image Hough transform. Based on lanes identification, this paper designs a navigation algorithm of driving direction. This algorithm performs driving direction navigation decisions according to two characteristic parameters which are deviation angle and deviation distance. The experimental results indicate that the developed system exhibits good detection performances in recognition reliability and navigation decision. It has proved that this system has high accuracy, large detection range and high practicability.

Abstract: Aiming at regulating the toll evasion behaviors in highway weight charges and reducing charge disputes caused by jumping, this article studied the algorithm that tracks vehicle beating when it is passing the scale. Based on license plate location, vehicle movement could be characterized by tracking the plate centroid using Lucas-Kanade optical flow algorithm. The optical flow vector of the centroid was calculated frame by frame, which could be used for drawing trajectory of centroid coordinates, and calculating beating parameters. In order to expand the detection range and adaptability of the algorithm, through calculating optical flow hierarchically combined with Gaussian pyramid, then tracking centroid from high lever to low in the image pyramid, it could achieve the capture of the vehicle' fast moving. Through experiments, trajectory reflected vehicle beating information well, which provides a strong evidence means for levy problem of the highway weight charges.

Abstract: The present traffic controllers are most run in 2,3 or 4 phase mode at single point intersections, but the transition of different phase modes during traffic controlling is not flexible. In view of this deficiency, a multi-stage traffic signal controller of variable phases was designed and realized. It is equipped with a single chip microcomputer as the core, signal lights driver circuit, real-time clock, keys and LCD display modules etc. By means of physical simulation, the results show that this controller can run reliably according to system parameters and achieve automatic transition of 2, 3 or 4 phase mode in different stages, thus it also is proved that this controller is more flexible and efficient to control traffic flow.

Abstract: Based on the study on traffic flow characteristics of the intersection, and current signal timing model of intersection, this paper selected the stop delay, the number of stops and parking traffic capacity as the indexes, and translated them into a single nonlinear objective function which is the fitness of genetic algorithm. In order to meet the changes of intersection traffic flow, this paper improved the basic genetic algorithm. The improved algorithm with two genetic layers carried on signal timing optimization for middle traffic flow and peak traffic flow situation. Experiments show that the model is reasonable, and the effect caused by timing parameters optimization is obvious.

Abstract: Toll evasion is a known challenging problem in expressway transportation. Existing solutions mainly rely on the audit by people on the spot. However, with the increase of expressway networks, this kind of methods become less practical and lead to low efficiency. Considering this deficiency, this paper presents a method based on data mining to automatically assist the toll audit. Our method not only takes use of the data recorded from the expressway toll system, but also brings convenience to the auditing process.

Abstract: To meet the requirements of smoothness and flexibility, an intelligent track-searching vehicle is designed to run fast and smoothly on different tracks. It is equipped with a MC9S12DG128 SCM as core, a steering engine, a motor-control module and other modules. The continuous path recognition algorithm is applied to collect track information, and the optimized PID control algorithm is used to control the intelligent vehicle’s steering engine and motor. The experimental results show that the detection accuracy of intelligent vehicle is significantly increased; its sensitivity and stability are also obviously improved, achieving the fundamental goal of high speed of intelligent track-searching vehicle.