Method and apparatus for performing object detection based on images captured by a fisheye camera and electronic device
This disclosure provides an apparatus and method for performing object detection based on images captured by a fisheye camera and an electronic device. The apparatus includes a memory and a processor coupled to the memory. The processor according to an embodiment is configured to: project an original image captured by the fisheye camera onto a cylindrical or spherical projection model, and perform reverse mapping to obtain at least two reversely mapped images, angles of view of the at least two reversely mapped images being towards different directions, detect objects in the reversely mapped images, respectively, and detect an object that is the same among the objects detected in the reversely mapped images. According to this disclosure, information in the wide field of view images obtained by capturing by the fisheye camera may be fully utilized.
1. An apparatus for performing object detection based on images captured by a fisheye camera, the apparatus comprising:
a memory;
a processor coupled to the memory and configured to:
project original images captured by the fisheye camera onto a cylindrical or spherical projection model, and perform reverse mapping to obtain at least two reversely mapped images corresponding to a same original image, angles of view of the at least two reversely mapped images being towards different directions,
detect objects in the at least two reversely mapped images, respectively, and
detect an object that is the same among the objects detected in the at least two reversely mapped images,
wherein, in performing the reverse mapping, the processor sets an offset related to a width and/or a height of an original image among the original images, the offset being set to be greater than a value and less than the value respectively, so as to obtain the at least two reversely mapped images,
wherein, when the offset is equal to the value, the angles of view face a center of the width and/or the height of the original image,
wherein the processor is further configured to:
map, for objects existing in overlapped regions of the at least two reversely mapped images, contour frames of the objects in the overlapped regions of the at least two reversely mapped images into the original images to obtain mapped contour frames, and
detect a degree of overlapping of the mapped contour frames, and determine that the object is the same when the degree of overlapping is greater than a threshold.
2. The apparatus according to claim 1 , wherein,
the processor performs detection on the at least two reversely mapped images based on a foreground and a background, and/or performs detection of the object based on a classifier obtained based on a training.
3. The apparatus according to claim 1 , wherein the processor is further configured to:
detect a motion state of the object according to a result of the detection of the object.
4. The apparatus according to claim 3 , wherein the processor is further configured to:
calculate, when determining that the objects detected in the at least two reversely mapped images are the object that is the same among the objects, motion velocities of the object respectively based on the at least two reversely mapped images, and
detect the motion state of the object according to a difference between the motion velocities of the object calculated respectively based on the at least two reversely mapped images.
5. An electronic device, comprising the apparatus as claimed in claim 1 .
6. A method for performing object detection based on images captured by a fisheye camera, the method comprising:
projecting original images captured by the fisheye camera onto a cylindrical or spherical projection model, and performing reverse mapping to obtain at least two reversely mapped images corresponding to a same original image, angles of view of the at least two reversely mapped images being towards different directions;
detecting objects in the at least two reversely mapped images, respectively; and
detecting an object that is the same among the objects detected in the at least two reversely mapped images,
wherein, in performing the reverse mapping, an offset related to a width and/or a height of an original image among the original images, the offset being set to be greater than a value and less than the value respectively, so as to obtain the at least two reversely mapped images;
wherein, when the offset is equal to the value, the angles of view face a center of the width and/or the height of the original image,
wherein the detecting of the object that is the same among the objects detected in the at least two reversely mapped images comprises:
mapping, for objects existing in overlapped regions of the at least two reversely mapped images, contour frames of the objects in the overlapped regions of the at least two reversely mapped images into the original images to obtain mapped contour frames, and
detecting a degree of overlapping of the mapped contour frames, and judging that the object is the same when the degree of overlapping is greater than a threshold.