016 International Conference on Control and Automation (ICCA 016) ISBN: 978-1-60595-39-8 Breaker Contact Movement State Recognition Based on Image Processing Tian-zheng WANG 1, Yan-song LI 1, Zhong-jie WANG 1, Li-qiang MA 1 and Ya-xiao WANG,a,* 1 State Grid Shanxi Electric Power Research Institute, Taiyuan, Shanxi 030001, China North China Electric Power University, Baoding, Hebei 071003, China a ncepu_wyx@163.com, TEL: +86-1500096851 *Corresponding author Keywords: Particle filter algorithm, Image, Breaker. Abstract. This paper briefly describes a new method of breaker contact movement state recognition. This method is based on image target tracking. Particle filter algorithm can solve the non-linear processing problem, so it is better robustness and accuracy to the image target tracking. Using high-speed motion picture camera to obtain the circuit breaker image can solve some difficulties which are done by traditional contact measurement methods. Experimental results show that breaker contact movement state recognition based on image processing is very effective. Introduction High voltage circuit breaker is a very important power system electrical equipment [1]. The power grid safe and stable operation is related to normal operation of the circuit breaker. To improve the stability of the grid, many experts begin to engage in breaker performance especially mechanical properties. Though the structure of the circuit breaker is very complex, the main cause of the failure is determined by its mechanical properties. In the existing circuit breakers, actuator failures reach to 64. 8%, in which refusing to move, malfunction, slow separating and three- phases closing at different time account for at least 80% of these institutions failures []. The main equipment used at home and abroad is touch sensor. Most touch sensors are mounted on the actuator or spindle, which may cause some impact on operation of the circuit breaker. In the installation process, there is some gap between actuator and sensor, which can lead to unnecessary measurement error. Because of breaker models various, the sensor jig models are different, and their installation processes are too cumbersome. Furthermore, since the circuit breakers tend to be more compact, its internal structure will be more compact, which has impact on the installation of the sensor. Sometimes the sensor even may not be mounted [3]. This paper presents a circuit breaker contact movement state recognition method based on image processing to overcome the contact test method insufficient. Namely insulating rod connected with the circuit breaker contacts as an object image acquisition, sets a high-performance camera outside the circuit breaker to the coil current signal as a trigger to shoot, film the process of moving through the circuit breaker window, gets contact movement state through the series of pictures. Particle filtering technology is based on each pixel in the image as a particle filter in each frame of the target. It can filter the observation status. Finally the mechanical properties of the circuit breaker can be calculated by data processing. Breaker Moving Contact Sport Track Moving Target Marking Direct observation of the movable contact is difficult for the breaker moving contact is located in arc quenching chamber. Therefore we choose insulated rod and rigidly connected to the movable contact as an observation object. Insulated rod movement and rotation indirectly reflects the
movement of the moving contact. In order to distinguish observed objects clearly from the background, we paste auxiliary markers which have distinctive grayscale and color characteristics of the lever. When the circuit breaker switch-on-off, we observe the auxiliary markers through high-speed and high-definition video camera and calculate the mechanical properties of the moving contacts. Therefore, this paper use auxiliary markers on insulate rod as an observation target. Tracking specific processes is shown in Fig. 1. Illumination Acquiring the actual size of target by second makers Expert systems or neural networks Breaker Auxiliary markers in the insulating rod affixed with the logo and bright colors as an observation target High-speed, highdefinition cameras capture the target image Obtain an image sequence Tracking target by particle filter algorithm Obtain breaker mechanical properties, make decision Malfunction Normal Manual intervention Figure 1. Particle filter tracking algorithm flowchart. Observations Target Tracking Patches with bright colors are chosen as observed target, which are seen as auxiliary marker. It is easy to distinguish the auxiliary markers from insulated rod and other components inside. Therefore, this method has better robustness and the entire implementation. It is also more convenient. By observing position change of the ion target image, its trajectory can be gotten and the mechanical properties of the circuit breaker operating mechanism can be calculated. Therefore, the key of breaker contact movement state recognition based on image processing is detected through the image sequence analysis to track the observation target in the circuit breaker with the movement contact switch-on-off. Color histogram s feature is a method used widely. Series images taken by the high-speed camera is a color model. The image is used by the three primary colors which it product color space mixed mode. So the light change, the color change. Three primary colors can be transformed into each other. So RGB color histogram model does not apply to the image processing for color uniformity and independence [7]. HSV color histogram is adopted to process image. Compared to the RGB space, HSV color space is line with the human perception of color, and the color of the basic attributes of hue, saturation and brightness is independent of each other. H and S can be considered insensitive to light and allocated a larger quantization level, and V components are very sensitive to light and allocated a smaller quantization level. So the total number of quantization steps is N nh nv n. S x0 is seen as the center of the target area normalized histogram: n xi x 0 qcolour ( u) Ck b( xi ) u u 1,,, N i1 h. (1) 1 x, k(x) 0, x 1. () Others h is tracking window width (including the length and width of ch cl)and it is defined as h (ch/ ) (cl/ ). is the Kronecker delta function. bx ( i ) is the index of x i in histogram. n xi x 0 C 1/ k is the normalization constant. i1 h Mechanical properties is acquired by observing target, it is shown in Fig. :
Figure. Mechanical properties of the breaker movable contact acquisition method schematic. Angular displacement characteristic can be calculated by the rotation angle in the image of moving contact switch-on-off. This is because the circuit breaker moving contact is rigidly connected to the insulated rod, The substance heart connection line of two auxiliary flags is the shaft angle of rotation. The shaft angle indirectly reflects the angular displacement of the moving contact. Before the moving contact action, the connection between two auxiliary flags heart is l 1. After the moving contact action, the connection between two auxiliary flags is l. Then is the angle of the breaker moving contact displacement from l to l. [0, ) 1 and it is calculated as: k k1 tan 1 k k 1 (3) k 1 is the slope of the centroid connection l. 1 k is the slope of the centroid connection l. Using this formula to calculate the displacement angle is based on the premise that l 1 and l have slope. If none exist, then l 1 and l are perpendicular to the selected horizontal axis, and the angle displacement is zero; if and only if there is a slope to exist, there are four cases: The slope l does not exist, and the angle between the horizontal axis and l is 1 1 /, then / ; 1 The slope l does not exist, and the angle between the horizontal axis and l is 1 1 /, then / ( 1 ); The slope l 1 does not exist, and the angle between the horizontal axis and l is /, then / ; The slope l 1 does not exist, and the angle between the horizontal axis and l is /, then /; Motion Tracking Algorithm Selection Based on the characteristic features of different templates, tracking algorithms are divided into many kinds. For example, based on the tracking area, based on the active contour tracking, based on characteristics tracing and based on the model tracking; Based on different identify and predict motion parameters methods, these algorithms can be divided into Mean Shift algorithm,camshif algorithm, Kalman filter and its improved algorithms, particle filter algorithm [8,9] and so on. Meanshift algorithm is affected to the external environment. This may miss the target. Camshif algorithm is not conducive to automate tracking operation. Although Kalman filter and its improved algorithm don t have these shortcomings, these do not apply to trace nonlinear curve movement. Particle filter algorithm is apply to non-linear, non-gaussian systems target tracking algorithm, and it has a strong multi-modal capabilities. In measuring nonlinear movement models and tracking multi-form movement targets, the particle filter has excellent robustness [10]. The selection of the
particle filter tracking methods algorithm is based on the characters of mechanical properties parameter measurement. Experimental Analysis By analyzing H component in image sequences, H component distribution histogram can be d described. It is shown in Fig. 3: Figure 3. H component of auxiliary markers distribution histogram. The video image from the first frame to the last frame are transformed into HSV space, then H component in each area can extracted. Where if the H component s characteristic shape and area is same as auxiliary markers, the position will be extracted. Then the particle filter tracking process will be initialized. Using the particle filter algorithm to analyze auxiliary marker motion in each images, the motion characteristics of the circuit breaker contacts can be achieved. The experiment is implemented on a model of LW59-5/4000-50 of 0kV high-voltage AC sulfur hexafluoride circuit breaker. Using this method to obtain the moving contact lines and angular displacement characteristic, the time curves can be got.they are shown in Fig. 4 and 5.These curves are consistent with the circuit breaker factory report switch-on and switch-off stroke curve. This shows that the method is effective. This method also can be analyzed to all types of high-voltage circuit breaker moving contact motion. Figure 4. Linear displacement curve of breaker switch-on.
Figure 5. Angular displacement characteristic curve of Breaker switch-off. Acknowledgements This paper is subsidized by State Grid (NO.SGTYHT/14-JS-188) References [1] Huaijun Zhao, Peng Ma, Weiping Fu, Yuangui Cheng. DSP-based breaker mechanical properties detection device, J. High-voltage Electrical. (006) 38-384. (In Chinese) [] G Mazza, R Michaca. The First International Enquiry on Circuit-Breaker Failures and Defects in Service. Elektra. 1998 [3] Xingquan Huang, Yuxiao Zhang, Qi Li, Fei Pan, Chengrong Li. High-speed imaging method to detect mechanical properties of the circuit breaker, J. North China Electric Power University (Natural Science),011,43-47. (In Chinese) [4] Okuma K., Taleghani A., De Freitas N., et al. A boosted particle filter: Multitarget detection and tracking, M. Computer Vision-ECCV 004. Springer Berlin Heidelberg, 004, 8-39. [5] Yang J., Schonfeld D., Mohamed M.. Robust video stabilization based on particle filter tracking of projected camera motion, J. Circuits and Systems for Video Technology, IEEE Transactions on, 009, 945-954. [6] Chang C, Ansari R. Kernel particle filter for visual tracking, J. Signal processing letters, IEEE, 005, 4-45. [7] Wei Zeng, Guibin Zhu, Jie Chen, Dingding Tang. Robust particle filter tracking algorithm by multi-feature fusion, J. Computer Application, 010, 643-645. (In Chinese) [8] Takahito Kawanishi, Takayuki Kurozumi, Kunio Kashino, Shigeru Takagi. A Fast Template Matching Algorithm with Adaptive Skipping Using Inner-Sub templates Distances. Proceedings of the 17th International Conference on Pattern Recognition (ICPR 04). [9] Heipke C. Overview of image matching techniques. Proc. of the OEEPE Workshop on the Application of Digital Photogrammetric Workstations. 1996 [10]Fangfang Du. Vision-based particle filter target tracking algorithm and its application in mobile robots, D. Hangzhou University of Electronic Science and Technology, 009. (In Chinese)