Apparatus and method for analyzing fisheye camera image
Disclosed is an image analysis apparatus for analyzing a camera image. The image analysis apparatus for analyzing a camera image segments an input fisheye camera image into segmented images with a preset size field of view and superimposes the segmented images so that some regions overlap, performs dewarping on each of the segmented images, then combines the segmented images on which the dewarping is performed using a preset combination method, generates an analysis image, and detects objects included in the analysis image. In this case, the image analysis apparatus removes a result recognized as a duplicate from a detection result of the object by post-processing.
1. An image analysis apparatus for analyzing a camera image, which analyzes a fisheye lens camera image, the image analysis apparatus comprising:
a memory; and
a processor, wherein the processor, when executing program instructions stored in the memory, is configured to perform:
a pre-processing operation of segmenting an input image into segmented images with a preset size field of view so that some regions of the displayed image overlap in the segmented images, performing dewarping on each of the segmented images, then combining the segmented images on which the dewarping is performed using a preset combination method, and to generate an analysis image;
an image analysis operation of receiving the analysis image and detecting an object; and
a post-processing operation of removing a result recognized as a duplicate from a detection result of the object of the image analysis operation,
wherein the analysis image is generated by concatenating the segmented images on a predetermined grid,
wherein the segmented images are images of different areas of a fisheye lens camera image, and
wherein the post-processing operation includes determining a duplicate object on the basis of coordinates of a region of the object in the fisheye lens camera image converted from coordinates of the segmented image of a bounding box and removing the result recognized as the duplicate from the detection result of the object.
2. The image analysis apparatus of claim 1 , wherein the post-processing operation further includes comparing the images of the detected object to determine similarity of the images of the object and removing the result recognized as the duplicate.
3. The image analysis apparatus of claim 1 , wherein the processor is further configured to perform: a user interface operation of receiving a size of a field of view to be segmented and a method of combining segmented images from a user and setting the size and the method.
4. The image analysis apparatus of claim 3 , wherein:
the user interface operation comprises setting a region of interest (ROI) in the fisheye lens camera image according to a user's input; and
the pre-processing operation comprises segmenting the input image with the preset size field of view so that boundary portions of the ROI overlap.
5. The image analysis apparatus of claim 1 , wherein at least a part of the object is shown in an overlapped region in at least two of the segmented images.
6. The image analysis apparatus of claim 5 , wherein the analysis image is different from any of the at least two of the segmented images.
7. The image analysis apparatus of claim 5 , wherein the analysis image comprises a part of one of the at least two of the segmented images and a part of the other of the at least two of the segmented images.
8. The image analysis apparatus of claim 1 , wherein any portion of the input image belongs to one or two segmented images.
9. The image analysis apparatus of claim 1 , wherein the analysis image comprises a plurality of dewarped segmented images which are concatenated on the predetermined grid, the object belongs to at least two dewarped segmented images, and one of the at least two dewarped segmented images completely displays the object.
10. The method of claim 1 , wherein each point in the fisheye lens camera image is converted to at least one point in one of the segmented images.
11. The method of claim 1 , wherein each point in any of the segmented images is converted to one point in the analysis image.
12. The method of claim 1 , wherein every point in the fisheye lens camera image is converted to at least one point in the analysis image.
13. A method of analyzing an image using an image analysis apparatus for analyzing a camera image, in which the image analysis apparatus analyzes a fisheye lens camera image, the method comprising:
an image segmentation operation of segmenting an input image into segmented images with a preset size field of view so that some regions of the displayed image overlap in the segmented images;
a dewarping operation of performing dewarping on each of the segmented images;
an image combination operation of combining the segmented images on which the dewarping is performed using a preset combination method to generate an analysis image;
an image analysis operation of receiving the analysis image and detecting an object; and
a post-processing operation of removing a result recognized as a duplicate from a detection result of the object in the image analysis operation,
wherein the analysis image is generated by concatenating the segmented images on a predetermined grid,
wherein the segmented images are images of different areas of a fisheye lens camera image, and
wherein the post-processing operation includes determining a duplicate object on the basis of coordinates of a region of the object in the fisheye lens camera image converted from coordinates of the segmented image of a bounding box and removing the result recognized as the duplicate from the detection result of the object.
14. The method of claim 13 , wherein the post-processing operation further includes comparing the images of the detected object to determine similarity of the images of the object and removing the result recognized as the duplicate.
15. The method of claim 13 , further comprising a region of interest (ROI) setting operation of setting an ROI in the fisheye lens camera image according to a user's input,
wherein the image segmentation operation includes segmenting the input image with the preset size field of view so that boundary portions of the ROI overlap.
16. The method of claim 13 , wherein at least a part of the object is shown in an overlapped region in at least two of the segmented images.
17. The method of claim 16 , wherein the analysis image is different from any of the at least two of the segmented images.
18. The method of claim 16 , wherein the analysis image comprises a part of one of the at least two of the segmented images and a part of the other of the at least two of the segmented images.
19. The method of claim 13 , wherein any portion of the input image belongs to one or two segmented images.
20. The method of claim 13 , wherein the analysis image comprises a plurality of dewarped segmented images which are concatenated on the predetermined grid, the object belongs to at least two dewarped segmented images, and one of the at least two dewarped segmented images completely displays the object.