Moving Object Stitching Method Based on Single Fixed Camera Image Sequence

Abstract

The invention relates to a single fixed camera image sequence-based moving target splicing method. The method comprises the following steps of detecting out a foreground region and / or a background region in an input image sequence of a single fixed camera; identifying whether the foreground region in an image is a foreground region formed by moving targets or a foreground region comprising the moving targets; taking an input image frame comprising a moving target region as a target image frame for splicing input images; correcting and / or removing distortion of the moving target region; extracting image features of the moving target region and performing feature matching; and according to a matching result and a corresponding matching relationship, calculating image conversion parameters, and performing image splicing of the moving target region to form a complete moving target image. The moving targets can be spliced; the complete moving target image is formed by splicing the local images comprising the moving targets and shot at different moments by the single camera; and the method adapts to occasions of bayonet video monitoring and the like.

Description

technical field

[0001] The invention relates to a moving target splicing method based on a single fixed camera image sequence, which is suitable for bayonet video monitoring and other occasions. Background technique

[0002] Image mosaic is to combine several partial images with overlapping parts into a large seamless panoramic image. The basic technical route is usually to re-project multiple images onto a common surface after registration, and then fuse the images. To generate a panoramic image, the core technology is image registration and transformation.

[0003] The current relatively mature image stitching technology is to stitch multiple images collected by multiple fixed cameras. The processing process is as follows: (1) extract and match image features, calculate the transformation matrix between adjacent images and Called the homography matrix. (2) Using an iterative algorithm, decompose the internal and external parameter matrices of the cameras corresponding to...

Examples

Embodiment Construction

[0070] The present invention will be further described below in conjunction with the accompanying drawings.

[0071] See attached Figure 1-6 , what the present invention considers is the mosaic of moving object, and mosaic object is the moving object in image sequence (such as figure 1 shown), and therefore the camera translation during the image acquisition process cannot be ignored, and the infinity assumption is not satisfied.

[0072] Due to the above reasons, the key problems to be solved in the present invention are: (1) Distinguish moving frames and still frames in the image sequence, and use moving frames containing moving objects as splicing input. (2) Distinguish the motion area and the still area in the image, and only process the motion area in the image when the image is matched. (3) Considering that the distance between the camera and the target will cause image distortion, the image needs to be corrected for distortion. (4) Select an appropriate image transf...

technical field

[0001] The invention relates to a moving target splicing method based on a single fixed camera image sequence, which is suitable for bayonet video monitoring and other occasions. Background technique

[0002] Image mosaic is to combine several partial images with overlapping parts into a large seamless panoramic image. The basic technical route is usually to re-project multiple images onto a common surface after registration, and then fuse the images. To generate a panoramic image, the core technology is image registration and transformation.

[0003] The current relatively mature image stitching technology is to stitch multiple images collected by multiple fixed cameras. The processing process is as follows: (1) extract and match image features, calculate the transformation matrix between adjacent images and Called the homography matrix. (2) Using an iterative algorithm, decompose the internal and external parameter matrices of the cameras corresponding to...

Examples

Embodiment Construction

[0070] The present invention will be further described below in conjunction with the accompanying drawings.

[0071] See attached Figure 1-6 , what the present invention considers is the mosaic of moving object, and mosaic object is the moving object in image sequence (such as figure 1 shown), and therefore the camera translation during the image acquisition process cannot be ignored, and the infinity assumption is not satisfied.

[0072] Due to the above reasons, the key problems to be solved in the present invention are: (1) Distinguish moving frames and still frames in the image sequence, and use moving frames containing moving objects as splicing input. (2) Distinguish the motion area and the still area in the image, and only process the motion area in the image when the image is matched. (3) Considering that the distance between the camera and the target will cause image distortion, the image needs to be corrected for distortion. (4) Select an appropriate image transf...

Images