SIFT-based object recognition for tracking in infrared imaging system
Changhan Park and Sunghun Jung
SIAT, Samsung Thales Co., Ltd., 304, Chang-li, Namsa-Myun, Cheoin-Gu, Yongin, Gyeonggi 449-886, KOREA Abstract In this paper, we propose an automatic object tracking and recognition in unknown environments by using scale-invariant feature transform (SIFT) in PowerPC-based infrared (IR) imaging system. Proposed method consists of two stages. First, we must localize the interest point in position and scale of moving objects. Second, we must build a description of the interest point and recognize moving objects. Proposed SIFT method for an effective feature extraction in PowerPC-based IR imaging system consists of scale space, extrema detection, orientation assignment, key point description, and feature matching. SIFT descriptor sets up extensive range about 1.5 times than visual image when feature value of SIFT in IR image is less than visual image. Based on experimental results, the proposed method is extracted objects feature values in our system, and the result is presented by experiment. I. INTRODUCTIONAND BACKGROUND BJECT recognition techniques have been actively investigated and have wide application in various fields. An infrared (IR) image and a visual image are a potential solution to improve an objects detection, tracking, recognition, and image fusion performance. Tracking and recognition based on the visual image is sensitive to variations in illumination conditions [1][2]. The most popular solution for the robust object recognition method is scale-invariant feature transform (SIFT) approach that transforms an input image into a large collection of local feature vectors, each of which is invariant to image translation, scaling, and rotation [3]. In this paper, we propose an automatic object tracking and recognition by using SIFT in PowerPC-based infrared (IR) imaging system [4]. Proposed method consists of two stages. First, we must localize the interest point in position and scale of moving objects. Second, we must build a description of the interest point and recognize moving objects. Proposed method uses SIFT for an effective feature extraction in our system. Proposed SIFT method consists of scale space, extrema detection, orientation assignment, key point description, and feature matching. SIFT descriptor sets up extensive range about 1.5 times than visual image when feature value of SIFT in IR image is less than visual image. Because, object of IR image is analogized by field test that it exists more expanse form than visual image. Therefore, proposed SIFT descriptor is constituted at more expanse term for a precise matching of object. This paper is organized as follows: Section 2 describes object segmentation by using the morphological operation and feature extraction by using the SIFT in our system. Section 3 presents robust algorithm against Gaussian noise and experimental, and Section 4 concludes the paper. II. FEATURE EXTRACTION In this section, we deal with morphological operation for object segmentation before SIFT in PowerPC-based IR imaging system. Proposed method for object segmentation can be used our algorithm which the method presented in [5]. We also deal with feature extraction for object recognition by using SIFT in our system. Proposed method for object recognition can be used our algorithm which the method presented in [3]. And we are simulated and apply to our system it. Figure 1 shows proposed method for feature extraction. Figure 1. Proposed method for feature extraction 2.1 Morphological operation for object segmentation The basic morphological operations are dilation and erosion, which are performed by convolving the structuring element (SE) with the image [5]. The other morphological operations are the evolvement result of the above two basic operations, such as opening operation, closing operation, and top-hat operation. We use morphological top-hat operation for object segmentation. The morphological top-hat operation for gray-scale images is part of the basic toolbox of morphological operations. Its function is to detect contrasted objects on non-uniform backgrounds. Depending upon whether we are dealing with thermal objects on a dark background, the operation is defined as: )) ( min ( max ) ( ) , ( A A B A A B A Tophat B B $ (1) 2.2 Feature extraction using SIFT In this session, we deal with feature extraction for object recognition in PowerPC-based IR imaging system. Image features have many properties that make them suitable for matching differing images of an object in IR sensor. The features are invariant to image scaling and rotation, and partially invariant to change in illumination. Proposed method uses SIFT for an effective feature extraction in our system. O 978-1-4244-5417-4/09/$26.00 2009 IEEE 978-1-4244-5417-4/09/$26.00 2009 IEEE Following are the major stages of computation [3] used to generate the set of IR image features: 1) Scale-space extrema detection 2) Key point localization 3) Orientation assignment 4) Key point descriptor III. ROBUST ALGORITHMAGAINST GAUSSIANNOISE The performance of the SIFT-based recognition is affected by noise, especially zero-mean Gaussian noise. To mitigate the noise effect, various details can be used. Under the assumption that an object is brighter than the background, the intensity of the object is likely to be larger than the noise standard deviation. Let us consider the situation where an object is moving from the ) 1 ( k th frame to the k th frame. If the object is small and moving, the probability that the difference between the intensity of the object pixel in the current frame and that of the same-positioned pixel in the previous frame is larger than standard deviation of noise is close to 1. Let be defined as at the i ) , ( c r I i ) , ( c r I th frame, and be Gaussian noise on at the i ) , ( c r n i ) , ( c r th frame, which r and represent a row and a column in image, respectively. We assume: c ) , 0 ( ~ ) , ( 2 n i N c r n V . (2) Figure2 shows the recognition and tracking result of a moving object with zero-mean Gaussian noise . Figure 2(a) shows extracted feature vectors of selected object, Figure 2(b) shows occlusion while the object tracking from proposed method. Figure 2(c) shows coast tracking it that disappears the selected object. Figure 2(d) shows tracking the object when it appears an object. Figure 3 shows object recognition and tracking result of thermal moving objects. If it disappears the recognized object now, our system change tracking mode. Figure 4 shows object segments of the proposed algorithm with zero-mean Gaussian noise . ) 15 , 0 ( 2 N ) 20 , 0 ( 2 N (a) 24 th frame (b) 28 th frame (c) 32 nd frame (d) 40 th frame Figure 2. Object tracking and recognition results of the proposed algorithm with Gaussian noise . ) 15 , 0 ( 2 N (a) 36th frame (b) 40th frame Figure 3. Object tracking and recognition results of the proposed algorithm with Gaussian noise ) 20 , 0 ( 2 N (a) 36th frame (b) 40th frame Figure 4. Segmented results by the proposed algorithm with Gaussian noise ) 20 , 0 ( 2 N IV. CONCLUSIONS In this paper, we proposed real-time intelligence tracking and recognition of moving objects in PowerPC-based IR imaging system. We proposed the approach that can reduce the feature extraction time in SIFT by using morphological operation. The main idea is search for segmented thermal objects rather than to perform exhaustive search in the whole difference of Gaussian pyramid. Proposed method consists of two stages. First, we must localize the interest point in position and scale of thermal moving objects using morphological top-hat operation. Second, we must build a description of the interest point and recognize thermal moving objects. Experimental result, the proposed method can acquired real-time intelligence tracking and high recognition rate. REFERENCES [1] J. Wang, J. Liang, H. Hu, Y. Li, and B. Feng, Performance evaluation of infrared and visible image fusion algorithms for face recognition, Proc. International Conf. Intelligent Systems and Knowledge Engineering (ISKE 2007), pp. 1-8, 2007. [2] J. Lee, Y. Kim, C. Park, C. Park, and J. Paik. Robust feature detection using 2D wavelet transform under low light environment, Proc. Intelligent Computing in Signal Processing and Pattern Recognition (ICIC2006), vol. 345, pp. 1042-1050, 2006. [3] D. Lowe. Distinctive image features from scale-invariant keypoints, Int. Journal of Computer Vision Papers, vol. 60, no. 2, pp. 91-110, 2004. [4] J. Lee, J. Youn, and C. Park, "PowerPC-based system for tracking in infrared image sequences, Proc. SPIE, vol. 6737, pp. 67370S-9, 2007. [5] C. Xiaochun, H. Yihua, H. Guilan, T. Xiaohong, and L. Wuhu, Morphology based adaptive preprocessing method of infrared image sequence, Proc. SPIE 27th Int. Congress on High-Speed Photography and Photonics 6279, 62793C-6 (2007).