Image processing method and vehicle auxiliary apparatus

Abstract

An image processing method and an automotive auxiliary device are disclosed. The image processing method includes: generating a first target bounding box in an image to surround a target area by analyzing an image acquired by an image capturing device; obtaining multiple detection points corresponding to multiple key objects by analyzing the target area; obtaining a virtual representative extension line intersecting the first target bounding box at a first intersection point and a second intersection point based on these detection points, and determining a selected positioning point from the multiple positioning points based on the virtual representative extension line; removing the selected positioning point from the positioning points, and generating a second target bounding box based on the remaining positioning points, the first intersection point, and the second intersection point; and issuing a warning message in response to the overlap between the area selected by the second target bounding box and a collision warning area. This patent can significantly reduce false alarms and improve the accuracy of warnings.

Claims

1. An image processing method executed via a processor disposed in a vehicle, the image processing method comprising: An image is acquired via an imaging device; a first target bounding box is generated in the image to surround a target region by analyzing the image, wherein the first target bounding box includes multiple edges and multiple positioning points formed by the edges; multiple detection points corresponding to multiple key objects are obtained by analyzing the target region; a virtual representative extension line is obtained based on the detection points, and a selected positioning point is determined from the positioning points based on the virtual representative extension line, wherein the virtual representative extension line intersects a first edge and a second edge of the first target bounding box at a first intersection point and a second intersection point, respectively, and the first edge and the second edge form the selected positioning point; the selected positioning point is removed from the positioning points, and a second target bounding box is generated based on the remaining positioning points, the first intersection point, and the second intersection point; and a warning message is issued in response to the region selected by the second target bounding box overlapping with a collision warning region.

2. The image processing method as described in claim 1, further comprising, after generating the first target bounding box: In response to the fact that there are two key objects included in the target area, two detection points are obtained from each of the two key objects; And based on the two detection points, the virtual representative extension line is generated.

3. The image processing method as described in claim 1, further comprising, after generating the first target bounding box: In response to the fact that the number of key objects included in the target area is three or more, the detection points are obtained from the key objects respectively; multiple candidate extension lines are generated based on the detection points; and one of the candidate extension lines that does not pass through the key objects is selected as the virtual representative extension line.

4. The image processing method as described in claim 1, further comprising, after generating the first target bounding box: A central baseline is set by dividing the image into two equal parts, left and right. And determine the area proportion of the target area located to the left and right of the central baseline, thereby determining whether the target object selected by the first target frame is located to the left or right of the vehicle.

5. The image processing method of claim 1, wherein the step of generating a first target bounding box in the image to surround the target region by analyzing the image includes: The image is analyzed by an object detection module to find a target object in the image, thereby generating the first target bounding box, an object category parameter, and an object confidence index.

6. The image processing method as described in claim 1, wherein the step of obtaining the detection points corresponding to the key objects by analyzing the target area includes: The target area is analyzed by an object detection module to identify key objects, and the point with the smallest coordinate value in a vertical direction is extracted from the outline of each key object as the corresponding detection point.

7. The image processing method as described in claim 1 further includes: The collision warning zone is set based on a safe reaction distance.

8. The image processing method as described in claim 1, wherein the step of issuing the warning message includes: This causes an indicator light to illuminate based on the warning message.

9. The image processing method as claimed in claim 1, wherein the positioning points of the first target bounding box include a first positioning point, a second positioning point, a third positioning point, and a fourth positioning point, the fourth positioning point formed by the first side and the second side is determined as the selected positioning point, and the second target bounding box is generated based on the first positioning point, the second positioning point, the third positioning point, the first intersection point, and the second intersection point.

10. A vehicle auxiliary device, installed inside a vehicle, the vehicle auxiliary device comprising: An image-capturing device acquires an image; The image acquisition device is coupled to a processor and configured to: generate a first target bounding box in the image by analyzing the image to surround a target region, wherein the first target bounding box includes a plurality of edges and a plurality of positioning points formed by the edges; obtain a plurality of detection points corresponding to a plurality of key objects by analyzing the target region; obtain a virtual representative extension line based on the detection points, and determine a selected positioning point from the positioning points based on the virtual representative extension line, wherein the virtual representative extension line intersects a first edge and a second edge of the first target bounding box at a first intersection point and a second intersection point, respectively, the first edge and the second edge forming the selected positioning point; remove the selected positioning point from the positioning points, and generate a second target bounding box based on the remaining positioning points, the first intersection point, and the second intersection point; and issue a warning message in response to the region selected by the second target bounding box overlapping with a collision warning region.

11. The vehicle assistance device as claimed in claim 10, wherein the processor is configured to: obtain two detection points from each of the two key objects in response to the number of key objects included in the target area being two; and generate the virtual representative extension line based on the two detection points.

12. The vehicle assistance device as claimed in claim 10, wherein the processor is configured to: obtain detection points from the key objects respectively in response to the number of three or more key objects included in the target area; generate a plurality of candidate extension lines based on the detection points; and select one of the candidate extension lines that does not pass through the key objects as the virtual representative extension line.

13. The vehicle assistance device as claimed in claim 10, wherein the processor is configured to: after generating the first target frame, set a central baseline such that the image is divided into left and right sides; and determine the area proportion of the target region located on a left side and a right side of the central baseline, thereby determining whether a target object selected by the first target frame is located on the left or right side of the vehicle.

14. The vehicle assistance device as claimed in claim 10, wherein the processor is configured to: analyze the image through an object detection module to identify a target object in the image, thereby generating the first target bounding box, and generating an object category parameter and an object confidence index.

15. The vehicle assistance device as claimed in claim 10, wherein the processor is configured to: analyze the target area through an object detection module to identify key objects, and extract a point with the minimum coordinate value in a vertical direction from a contour of each of the key objects as a corresponding detection point.

16. The vehicle assistance device as claimed in claim 10, wherein the processor is configured to: set the collision warning zone based on a safe reaction distance.

17. The vehicle assistance device as claimed in claim 10, wherein the processor is configured to: drive an indicator light to illuminate based on the warning message.

18. The vehicle assistance device as claimed in claim 17, wherein the indicator light is disposed on the rearview mirror of the vehicle.

19. The vehicle assistance device as claimed in claim 10, wherein the positioning points of the first target frame include a first positioning point, a second positioning point, a third positioning point, and a fourth positioning point, the fourth positioning point formed by the first side and the second side is determined as the selected positioning point, and the second target frame is generated based on the first positioning point, the second positioning point, the third positioning point, the first intersection point, and the second intersection point.

Images