Column inverted image misjudgment prevention identification method and system

Abstract

The invention discloses a stand column inverted image misjudgment prevention identification method and system, and the method comprises the steps: continuously obtaining an image containing a stand column behind a vehicle when the vehicle is automatically parked in an underground garage, carrying out the saturation feature extraction of the image, and carrying out the segmentation of the image according to the saturation of the image; performing specific pixel scanning on the image according to regions, and re-segmenting the image after similarity matching; and sequentially carrying out visual single-frame filtering and visual multi-frame filtering on the segmented image to identify a column entity and an inverted image. The system can accurately recognize the inverted image of the stand column, and the driving efficiency and safety of memory parking during intelligent driving are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of vehicle environmental perception detection, and particularly to a method and system for preventing misjudgment of column reflections. Background Art

[0002] At present, intelligent driving of vehicles has some related functions, especially the memory parking function with the smooth epoxy floor underground garage as the core scenario. The smooth epoxy floor of the underground garage will produce reflected images of columns, and these reflected images may have various impacts on the intelligent driving of vehicles.

[0003] During the process of the vehicle using the memory parking function in the underground garage, in the image captured by the camera, as Figure 1 shown, the column reflections on the smooth epoxy floor may be confused with the real columns, affecting the accuracy of image recognition of the intelligent driving system. When the intelligent driving system misjudges the column reflections on the smooth epoxy floor as real obstacles, it may trigger unnecessary obstacle avoidance actions, such as emergency braking or steering avoidance, thus affecting the driving efficiency and safety of the vehicle. Frequent misjudgments and unnecessary obstacle avoidance actions will also reduce the user experience of the intelligent driving system and lower the user's trust in intelligent driving technology. Taking the situation shown in Figure 1 as an example, the column reflection and the column itself are integrated, and the intelligent driving system may not be able to distinguish them. Within the visual perception range, it may misjudge it as an intrusion into the lane. When the vehicle is driving according to the memory trajectory, it may be misjudged as an intrusion into the lane by the column reflection and forced to stop, resulting in jamming. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for preventing misjudgment of column reflections in view of the deficiencies of the prior art. The system can accurately identify column reflections and improve the driving efficiency and safety of memory parking during intelligent driving.

[0005] To achieve the above object, according to one aspect of the present invention, a method for preventing misjudgment of column reflections is provided, including:

[0006] When the vehicle automatically parks in the underground garage, continuously obtain an image containing columns behind the vehicle, extract the saturation feature of the image, and divide the part of the image containing columns into a first region and a second region according to the saturation of the image, wherein the average saturation value of the first region is greater than that of the second region;

[0007] Specific pixel scanning is performed on the image containing the column part that divides the first area and the second area by area. After matching the similarity, the image containing the column part is re-divided into a third area and a fourth area, and rectangular recognition frames are marked on the column in the third area image and the column in the fourth area image;

[0008] In the single-frame image containing the column, compare the actual length of the long side of the side line of the column in the third area image with the first length, and compare the actual length of the short side of the side line of the column in the third area image with the second length. If the actual length of the long side of the side line of the column in the third area image is less than the first length and the actual length of the short side of the side line of the column in the third area image is less than the second length, it is determined that the third area image contains a column reflection and the single-frame image containing the column is filtered; where both the first length and the second length are preset values;

[0009] In the continuous-frame image containing the column retained after filtering, continuously compare the distance between the bottom-end points of the rectangular recognition frame of the column in the third area image and the corresponding bottom-end points of the rectangular recognition frame of the column in the fourth area image, and continuously compare the distance between the midpoints of the two side lines of the column in the third area image. If the distance between the bottom-end points of the rectangular recognition frame of the column in the third area image and the corresponding bottom-end points of the rectangular recognition frame of the column in the fourth area image is greater than the third length and the distance between the midpoints of the two side lines of the column in the third area image is greater than the fourth length, it is determined that the column in the third area image is a column entity, and it is determined that the column in the fourth area image is a column reflection; where both the third length and the fourth length are preset values.

[0010] In the above solution, the method further includes: determining that the column in the third area image is a column entity also needs to satisfy that the length of the long side of the side line of the column in the third area image is greater than the fifth length.

[0011] In the above solution, the image of the vehicle containing the column is in RGB format.

[0012] In the above solution, the method for extracting the saturation feature of the image and dividing the part of the image containing the column into the first area and the second area according to the saturation of the image is: converting the RGB format image into an HIS format image, and extracting the saturation feature of the HIS format image based on general image processing principles. Specifically:

[0013] Calculate the average saturation of the HIS - format image as T, and set the segmentation saturation threshold as T1, where T1 is an updated value. First, segment the HIS - format image containing the column part into two parts according to the initial saturation calculation value R1. Calculate the average saturations Ta and Tb of the two parts respectively, and the average saturation of the sum of the two parts. Update the threshold T1 with the average saturation of the sum of the two parts. Repeatedly use the updated threshold T1 to segment the HIS - format image and update the threshold T1 until the absolute value of the difference between the average saturation T and T1 reaches a preset appropriate value, and complete the segmentation of the image containing the column part into the first region and the second region, where the average saturation of the first region is greater than that of the second region; where the initial saturation calculation value R1 is a preset value.

[0014] Where the calculation formulas for the intensity I and saturation S of the HIS - format image are as follows:

[0015]

[0016] The purpose of this step is that in the prior art, color extraction and segmentation are generally for a fixed saturation, which is mostly obtained by camera and real - vehicle calibration. It is good for real - scene recognition and discrimination, but not applicable to reflections. It is necessary to dynamically select an appropriate initial saturation extraction threshold.

[0017] In the above solution, the method of specifically scanning the pixels of the image containing the column part segmented into the first region and the second region by region and re - segmenting the image containing the column part into the third region and the fourth region after matching the similarity is as follows:

[0018] Convert the image containing the column part segmented into the first region and the second region into a grayscale image, scale it to a fixed - size pixel block, calculate the average pixel grayscale value of each pixel block, calculate the similarity of each pair of pixel blocks according to the average pixel grayscale value, and re - refine and segment the image containing the column part segmented into the first region and the second region into the third region and the fourth region according to the size of the similarity, where the similarity of the third region is greater than that of the fourth region.

[0019] Where the calculation formula for specifically scanning the pixels is as follows:

[0020]

[0021] is the average pixel grayscale value, x i is the grayscale value of each pixel.

[0022] In the above solution, the side lines of the column in the third area image are the two sides of the rectangle where the column contacts the ground; the longer one of the two side lines is the long side, and the shorter one of the two sides is the short side;

[0023] The bottom side of the column rectangular recognition frame in the third area image is the side parallel to and close to the ground of the rectangular recognition frame; the bottom side endpoints of the column rectangular recognition frame in the third area image are the two end points of the bottom side of the column rectangular recognition frame in the third area image;

[0024] The bottom side of the column rectangular recognition frame in the fourth area image is the side parallel to and close to the ground of the rectangular recognition frame, and the bottom side endpoints of the column rectangular recognition frame in the fourth area image are the two end points of the bottom side of the column rectangular recognition frame in the fourth area image.

[0025] In the above solution, the first length is 40 cm; the second length is 5 cm.

[0026] In the above solution, the third length is 20 cm; the fourth length is 30 cm.

[0027] In the above solution, the fifth length is 30 cm.

[0028] According to another aspect of the present invention, there is provided a system for preventing misjudgment of column reflections, including:

[0029] A visual reflection judgment saturation segmentation module, which is used to continuously obtain an image containing a column behind the vehicle when the vehicle automatically parks in an underground garage, extract the saturation features of the image, and divide the part of the image containing the column into a first area and a second area according to the saturation of the image, wherein the average saturation value of the first area is greater than that of the second area;

[0030] A visual reflection judgment consistency segmentation module, which is used to perform specific pixel scanning on the image of the part containing the column divided into the first area and the second area by area, re-divide the part of the image containing the column into a third area and a fourth area after matching the similarity, and label a rectangular recognition frame on the column in the third area image and the column in the fourth area image;

[0031] A visual single-frame filtering module is used to compare the actual length of the long side of the column edge line in the third area image with the first length and compare the actual length of the short side of the column edge line in the third area image with the second length in the single-frame image containing the column. If the actual length of the long side of the column edge line in the third area image is less than the first length and the actual length of the short side of the column edge line in the third area image is less than the second length, it is determined that the third area image contains a column reflection and is filtered;

[0032] A visual multi-frame filtering module is used to continuously compare the distance between the bottom edge endpoints of the column rectangular recognition frame in the third area image and the corresponding bottom edge endpoints of the column rectangular recognition frame in the fourth area image, and continuously compare the distance between the midpoints of the two edge lines of the column in the third area image in the continuous frame image containing the column retained after filtering. If the distance between the bottom edge endpoints of the column rectangular recognition frame in the third area image and the corresponding bottom edge endpoints of the column rectangular recognition frame in the fourth area image is greater than the third length and the distance between the midpoints of the two edge lines of the column in the third area image is greater than the fourth length, it is determined that the column in the third area image is a column entity and the column in the fourth area image is a column reflection.

[0033] Generally speaking, compared with the prior art by the above technical solutions conceived by the present invention, the following beneficial effects can be achieved:

[0034] (1) The present invention provides a system for preventing misjudgment and recognizing column reflections, which can accurately identify column reflections and understand the environment, improve the accuracy of vehicle trajectory judgment, and thus improve the driving efficiency and safety of memory parking during intelligent driving.

[0035] (2) The present invention provides a system for preventing misjudgment and recognizing column reflections, which can improve the performance of reflection and real system judgment at zero cost under the condition of the same function being carried; it is applicable to all vehicle models, can interact with various types of reverse radar systems, and operates modularly. Description of the Drawings

[0036] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Moreover, throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0037] Figure 1 Images of column reflections and real columns on a smooth epoxy floor.

[0038] Figure 2It is a schematic flowchart of a method for preventing misjudgment and recognition of the inverted image of a column in Embodiment 1 of the present invention.

[0039] Figure 3 It is a schematic diagram of the column in the third region in a single-frame image containing a column in Embodiment 1 of the present invention.

[0040] Figure 4 It is a schematic diagram of the bottom edge of the column rectangular recognition frame in a single-frame image containing a column in Embodiment 1 of the present invention. Detailed implementation mode

[0041] In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention. In addition, the technical features involved in the various embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0042] It should be understood that the magnitudes of the sequence numbers of the steps in the embodiments do not mean the order of execution. The order of execution of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0043] Embodiment 1

[0044] One aspect of the embodiments of the present application provides a method for preventing misjudgment and recognition of the inverted image of a column. Please refer to Figure 2 , including:

[0045] S1. When the vehicle automatically parks in the underground garage, continuously obtain an image containing a column behind the vehicle, extract the saturation feature of the image, and divide the part of the image containing the column into a first region and a second region according to the saturation of the image, where the average saturation value of the first region is greater than that of the second region.

[0046] Specifically, in the embodiments of the present application, it can be understood that when the vehicle automatically parks in the underground garage, the memory parking function is used to reverse the vehicle. During the reversing process, an image containing a column behind the vehicle is continuously obtained, and the inverted image of the column is also included in the image. In the embodiments of the present application, the image containing a column obtained behind the vehicle is in RGB format, and the RGB format image containing a column is converted into an HIS format image, and the saturation feature of the HIS format image is extracted based on general image processing principles. Specifically:

[0047] Calculate the average saturation of the HIS - format image as T, and set the saturation threshold for segmentation as T1, where T1 is an updated value. First, segment the HIS - format image containing the column part into two parts according to the initial saturation calculated value R1, calculate the average saturations Ta and Tb of the two parts respectively and the average saturation of the sum of the two parts, and update the threshold T1 with the average saturation of the sum of the two parts. Repeat using the updated threshold T1 to segment the HIS - format image and update the threshold T1 until the absolute value of the difference between the average saturation T and T1 reaches a preset appropriate value. In this embodiment, this preset value is T W is 1%; after the segmentation of the image containing the column part into the first region and the second region is completed, the average saturation of the first region is greater than that of the second region, where the average saturation of the first region is greater than that of the second region; the initial saturation calculated value R1 is a preset value, and in this embodiment, R1 = 0.5;

[0048] The calculation formulas for the intensity I and saturation S of the HIS - format image are as follows:

[0049]

[0050] Specifically, in the embodiment of this application, the saturation of the first region is 0.2, and the saturation of the second region is greater than 0.7.

[0051] S2, perform a specific pixel scan on the image containing the column part segmented into the first region and the second region by region. After matching the similarity, re - segment the image containing the column part into the third region and the fourth region, and label rectangular recognition frames on the columns in the third - region image and the columns in the fourth - region image.

[0052] Specifically, in the embodiment of this application, it can be understood that the image containing the column part segmented into the first region and the second region is converted into a grayscale image, scaled to a 5 * 5 pixel block, calculate the average pixel grayscale value of each pixel block, calculate the similarity of each pair of pixel blocks according to the average pixel grayscale value, and re - refine and segment the image containing the column part segmented into the first region and the second region into the third region and the fourth region according to the size of the similarity, where the similarity of the third region is greater than that of the fourth region.

[0053] In the embodiment of this application, the formula used for the specific pixel scan is:

[0054]

[0055] where is the average pixel grayscale value within the region, and x i is the grayscale value of each pixel.

[0056] S3. In a single-frame image containing a column, compare the actual length of the long side of the side line of the column in the third-region image with the first length, and compare the actual length of the short side of the side line of the column in the third-region image with the second length. If the actual length of the long side of the side line of the column in the third-region image is less than the first length and the actual length of the short side of the side line of the column in the third-region image is less than the second length, it is determined that the third-region image contains a column reflection, and the single-frame image containing the column is filtered out.

[0057] Specifically, in this embodiment, as Figure 3 shown, the side lines of the column in the third-region image are the two sides of the rectangle where the column contacts the ground; the longer one of the two sides of the side line is the long side, and the shorter one of the two sides is the short side. In the embodiment of the present application, the first length is 40 cm, and the second length is 5 cm.

[0058] S4. In the continuous-frame images containing the column that are retained after filtering, continuously compare the distance between the bottom-end points of the column rectangle recognition frame in the third-region image and the corresponding bottom-end points of the column rectangle recognition frame in the fourth-region image, and continuously compare the distance between the midpoints of the two side lines of the column in the third-region image. If the distance between the bottom-end points of the column rectangle recognition frame in the third-region image and the corresponding bottom-end points of the column rectangle recognition frame in the fourth-region image is greater than the third length and the distance between the midpoints of the two side lines of the column in the third-region image is greater than the fourth length, it is determined that the column in the third-region image is a column entity, and it is determined that the column in the fourth-region image is a column reflection.

[0059] Specifically, in this embodiment, as Figure 4 shown, the bottom side of the column rectangle recognition frame in the third-region image is a side parallel to and close to the ground of the rectangle recognition frame, that is, Figure 4 the blue line segments in the top view and the front view; the bottom-end points of the column rectangle recognition frame in the third-region image are the two end points of the bottom side of the column rectangle recognition frame in the third-region image;

[0060] The bottom side of the column rectangle recognition frame in the fourth-region image is a side parallel to and close to the ground of the rectangle recognition frame, that is, Figure 4 the red line segments in the top view and the front view, the bottom-most side of the yellow rectangle recognition frame (reflection); the bottom-end points of the column rectangle recognition frame in the fourth-region image are the two end points of the bottom side of the column rectangle recognition frame in the fourth-region image.

[0061] Specifically, in this embodiment, the third length is 20 cm; the fourth length is 5 cm.

[0062] Particularly, all distances in this embodiment are actual distances, not distances in the image.

[0063] In summary, after distinguishing the column entity and its reflection, the vehicle can successfully perform automatic parking.

[0064] Another aspect of the embodiments of the present application provides a system for preventing misjudgment and recognition of column reflections, including:

[0065] A visual reflection judgment saturation segmentation module, which is used to continuously obtain an image containing a column behind the vehicle when the vehicle automatically parks in an underground garage, extract the saturation features of the image, and divide the part of the image containing the column into a first region and a second region according to the saturation of the image, where the average saturation value of the first region is greater than that of the second region;

[0066] A visual reflection judgment consistency segmentation module, which is used to perform specific pixel scanning on the image of the column-containing part divided into the first region and the second region by region, and after matching the similarity, re-divide the image of the column-containing part into a third region and a fourth region, and mark rectangular recognition frames on the columns in the third-region image and the columns in the fourth-region image;

[0067] A visual single-frame filtering module, which is used to compare the actual length of the long side of the side line of the column in the third-region image with a first length, and compare the actual length of the short side of the side line of the column in the third-region image with a second length in a single-frame image containing a column. If the actual length of the long side of the side line of the column in the third-region image is less than the first length and the actual length of the short side of the side line of the column in the third-region image is less than the second length, it is determined that the column in the third-region image contains a column reflection and is filtered;

[0068] A visual multi-frame filtering module, which is used to continuously compare the distance between the bottom-end points of the rectangular recognition frame of the column in the third-region image and the distance between the midpoints of the two side lines of the column in the third-region image with the corresponding bottom-end points of the rectangular recognition frame of the column in the fourth-region image in the continuous frame image containing a column that remains after filtering. If the distance between the bottom-end points of the rectangular recognition frame of the column in the third-region image and the corresponding bottom-end points of the rectangular recognition frame of the column in the fourth-region image is greater than a third length and the distance between the midpoints of the two side lines of the column in the third-region image is greater than a fourth length, it is determined that the column in the third-region image is a column entity, and it is determined that the column in the fourth-region image is a column reflection.

[0069] Embodiment 2

[0070] One aspect of the embodiments of the present application provides a method for preventing misjudgment and recognition of column reflections, including:

[0071] S1. When the vehicle automatically parks in an underground garage, continuously obtain an image containing a column behind the vehicle, extract the saturation features of the image, and divide the part of the image containing the column into a first region and a second region according to the saturation of the image, where the average saturation value of the first region is greater than that of the second region.

[0072] S2. Perform a specific pixel scan on the image containing the column part that is divided into the first area and the second area by area. After matching the similarity, re-divide the image containing the column part into the third area and the fourth area, and label the columns in the third area image and the columns in the fourth area image with rectangular recognition frames.

[0073] S3. In a single-frame image containing a column, compare the actual length of the long side of the side line of the column in the third area image with the first length, and compare the actual length of the short side of the side line of the column in the third area image with the second length. If the actual length of the long side of the side line of the column in the third area image is less than the first length and the actual length of the short side of the side line of the column in the third area image is less than the second length, determine that the third area image contains a column reflection and filter the single-frame image containing the column; where both the first length and the second length are preset values.

[0074] S4. In the continuous-frame images containing columns that are retained after filtering, continuously compare the distance between the bottom endpoints of the column rectangular recognition frame in the third area image and the corresponding bottom endpoints of the column rectangular recognition frame in the fourth area image, and continuously compare the distance between the midpoints of the two side lines of the column in the third area image. If the distance between the bottom endpoints of the column rectangular recognition frame in the third area image and the corresponding bottom endpoints of the column rectangular recognition frame in the fourth area image is greater than the third length and the distance between the midpoints of the two side lines of the column in the third area image is greater than the fourth length; where both the third length and the fourth length are preset values;

[0075] The above content is basically the same as the corresponding content in Embodiment 1. The difference is that this embodiment further includes: when the length of the long side of the side line of the column in the third area image is greater than the fifth length, determine that the column in the third area image is a column entity, and determine that the column in the fourth area image is a column reflection; where the fifth length is a preset value, which is 30 cm in this embodiment.

[0076] It should be noted that according to the needs of implementation, each step described in this application can be split into more steps, or two or more steps or partial operations of steps can be combined into new steps to achieve the purpose of the present invention.

[0077] Those skilled in the art can easily understand that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent replacements, and improvements made within the spirit and principle of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for preventing misjudgment recognition of column reflections, characterized in that, Including: When the vehicle automatically parks in the underground garage, continuously obtain an image of the area behind the vehicle that includes columns, extract the saturation feature of the image, and divide the part of the image that includes columns into a first region and a second region according to the saturation of the image; Perform specific pixel scanning on the image of the part including columns that is divided into the first region and the second region by region. After matching the similarity, re-divide the part of the image that includes columns into a third region and a fourth region, and label a rectangular recognition frame on the column in the third region image and the column in the fourth region image; In the single-frame image including columns, compare the actual length of the long side of the side line of the column in the third region image with a first length, and compare the actual length of the short side of the side line of the column in the third region image with a second length. If the actual length of the long side of the side line of the column in the third region image is less than the first length and the actual length of the short side of the side line of the column in the third region image is less than the second length, determine that the third region image includes a column reflection and filter the single-frame image including columns; In the continuous-frame images including columns that are retained after filtering, continuously compare the distance between the bottom-end points of the column rectangular recognition frame in the third region image and the corresponding bottom-end points of the column rectangular recognition frame in the fourth region image, and continuously compare the distance between the midpoints of the two side lines of the column in the third region image. If the distance between the bottom-end points of the column rectangular recognition frame in the third region image and the corresponding bottom-end points of the column rectangular recognition frame in the fourth region image is greater than a third length and the distance between the midpoints of the two side lines of the column in the third region image is greater than a fourth length, determine that the column in the third region image is a column entity and determine that the column in the fourth region image is a column reflection.

2. A method for preventing misjudgment recognition of the column reflection according to claim 1, characterized in that The image obtained of the vehicle including columns is in RGB format.

3. The method for preventing misjudgment recognition of the column reflection according to claim 2, wherein The method for extracting the saturation feature of the image and dividing the part of the image that includes columns into a first region and a second region according to the saturation of the image is: convert the RGB format image into an HIS format image, and extract the saturation feature of the HIS format image based on general image processing principles. Specifically: Calculate the average saturation of the HIS - format image as T, and set the saturation threshold for segmentation as T1, where T1 is an updated value. First, segment the HIS - format image containing the column part into two parts according to the initial saturation calculation value R1, and calculate the average saturations Ta and Tb of the two parts respectively, as well as the average saturation of the sum of the two parts. Update the threshold T1 with the average saturation of the sum of the two parts. Repeat using the updated threshold T1 to segment the HIS - format image and update the threshold T1 until the absolute value of the difference between the average saturation T and T1 reaches a preset appropriate value, and complete the segmentation of the image containing the column part into the first region and the second region, where the average saturation of the first region is greater than that of the second region. Here, the initial saturation calculation value R1 is a preset value. The calculation formulas for the intensity I and saturation S of the HIS - format image are as follows:

4. The method for preventing misjudgment recognition of the column reflection according to claim 3, characterized in that, The method of specifically scanning pixels of the image containing the column part segmented into the first region and the second region by region and re - segmenting the image containing the column part into the third region and the fourth region after matching the similarity is as follows: Convert the image containing the column part segmented into the first region and the second region into a grayscale image, scale it to a fixed - size pixel block, calculate the average pixel grayscale value of each pixel block, calculate the similarity of each pair of pixel blocks according to the average pixel grayscale value, and re - refine and segment the image containing the column part segmented into the first region and the second region into the third region and the fourth region according to the size of the similarity, where the similarity of the third region is greater than that of the fourth region. The calculation formula for specifically scanning pixels is as follows: is the average pixel gray value, x i is the gray value of each pixel.

5. A method for preventing misjudgment recognition of a column reflection according to claim 1, characterized in that, The side lines of the column in the third - region image are the two sides of the rectangle where the column contacts the ground. The longer one of the two side lines is the long side, and the shorter one of the two side lines is the short side. The bottom side of the column rectangular recognition frame in the third - region image is the side parallel to and close to the ground of the rectangular recognition frame. The bottom - side endpoints of the column rectangular recognition frame in the third - region image are the two end points of the bottom side of the column rectangular recognition frame in the third - region image. The bottom side of the column rectangular recognition frame in the fourth - region image is the side parallel to and close to the ground of the rectangular recognition frame. The bottom - side endpoints of the column rectangular recognition frame in the fourth - region image are the two end points of the bottom side of the column rectangular recognition frame in the fourth - region image.

6. A method for preventing misjudgment recognition of the column reflection according to claim 1, characterized in that, The method further includes: determining that the column in the third - region image is a column entity and also needs to satisfy that the length of the long side of the side lines of the column in the third - region image is greater than the fifth length.

7. A method for preventing misjudgment recognition of a column reflection according to claim 1, characterized in that, The first length is 40 cm; the second length is 5 cm.

8. A method for preventing misjudgment recognition of the column reflection according to claim 1, characterized in that, The third length is 20 cm; the fourth length is 30 cm.

9. A method for preventing misjudgment recognition of the inverted image of a column according to claim 2, characterized in that, The fifth length is 30 cm.

10. An anti-misjudgment recognition system for column reflections, characterized in that, Including: The visual reflection judgment saturation segmentation module is used to continuously obtain an image containing columns behind the vehicle when the vehicle automatically parks in an underground garage, extract the saturation features of the image, and segment the part of the image containing columns into a first area and a second area according to the saturation of the image, where the average saturation value of the first area is greater than that of the second area; The visual reflection judgment consistency segmentation module is used to perform specific pixel scanning on the image of the part containing columns segmented into the first area and the second area by region, and after matching the similarity, re-segment the part of the image containing columns into a third area and a fourth area, and mark rectangular recognition frames on the columns in the third area image and the columns in the fourth area image; The visual single-frame filtering module is used to compare the actual length of the long side of the side line of the column in the third area image with the first length and compare the actual length of the short side of the side line of the column in the third area image with the second length in the single-frame image containing columns. If the actual length of the long side of the side line of the column in the third area image is less than the first length and the actual length of the short side of the side line of the column in the third area image is less than the second length, it is determined that the third area image contains a column reflection and is filtered; The visual multi-frame filtering module is used to continuously compare the distance between the bottom endpoints of the rectangular recognition frame of the column in the third area image and the corresponding bottom endpoints of the rectangular recognition frame of the column in the fourth area image, and continuously compare the distance between the midpoints of the two side lines of the column in the third area image in the continuous frame image containing columns after filtering. If the distance between the bottom endpoints of the rectangular recognition frame of the column in the third area image and the corresponding bottom endpoints of the rectangular recognition frame of the column in the fourth area image is greater than the third length and the distance between the midpoints of the two side lines of the column in the third area image is greater than the fourth length, it is determined that the column in the third area image is a column entity, and it is determined that the column in the fourth area image is a column reflection.

Images