A target identification method and device eliminating obstacle interference

Abstract

The present invention discloses a target recognition method and device for eliminating obstacle interference, and relates to the technical field of image target detection. The device collects near-infrared images at different times, then extracts contour points of the near-infrared images, determines the roundness value of the near-infrared images and the radius of the target fitting circle by the contour points, and then implements the screening of the geometric features of the target by the roundness value and the radius of the target fitting circle. Through the screening of multiple geometric features, the target can be more accurately identified; through the screening of the radius of the target fitting circle at different times, the combination of the time dimension and the radius of the fitting circle can effectively identify the slight position change of the target under the condition of obstacle occlusion, reduce the detection jitter and error caused by obstacle occlusion, thereby improving the stability and accuracy of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments.

Description

Technical Field

[0001] The present invention belongs to the technical field of image target detection, and in particular relates to a target recognition method and device that eliminates obstacle interference. Background Art

[0002] Near-infrared image is an image captured using light within the near-infrared spectrum. It has unique advantages, such as being insensitive to atmospheric conditions and visible light reflection, and being able to obtain relatively clear image information, especially under complex weather conditions such as haze and smoke. Therefore, target detection based on near-infrared images has good application prospects in environmental monitoring, road monitoring, industrial defect detection and other fields. However, in practical applications, obstacles such as vehicles and personnel often interfere with target detection. Obstacles can block the target in the near-infrared image or cause the target to be distorted, affecting target recognition.

[0003] The current target recognition method usually identifies the position of the target through multi-scale transformation based on grayscale images. However, in practical applications, the position of the target is constantly changing dynamically. The interference of obstacles during the change period leads to large errors in the obtained near-infrared image. Simply performing multi-scale transformation on the image cannot effectively eliminate the interference of obstacles and it is difficult to accurately capture the actual position of the target, resulting in inaccurate measurement of the target position, thus affecting the practical application in the fields of environmental monitoring, road monitoring, industrial defect detection, etc. Therefore, there is an urgent need for a target recognition method and device that eliminates obstacle interference. Summary of the invention

[0004] The present invention aims to provide a target recognition method and device that eliminates obstacle interference to solve the above technical problems. The target can be recognized through the roundness value and radius of the near-infrared image at different times, thereby improving the accuracy of target identification.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a target recognition method that eliminates obstacle interference, including:

[0006] Collecting a number of initial near-infrared images according to a preset time interval, and extracting contour points of the initial near-infrared images;

[0007] The initial near-infrared image is screened according to a preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images;

[0008] A target fitting circle of each of the first near infrared images is generated based on the contour points, and a target is identified based on a radius of the target fitting circle.

[0009] It can be understood that, compared with the prior art, the present invention collects near-infrared images at different times, then extracts contour points of the near-infrared images, determines the roundness value of the near-infrared images and the radius of the target fitting circle by the contour points, and then uses the roundness value and the radius of the target fitting circle to screen the geometric features of the target. By screening multiple geometric features, the target can be identified more accurately; by screening the radius of the target fitting circle at different times, the combination of the time dimension and the radius of the fitting circle can effectively identify slight position changes of the target in the case of obstacle occlusion, reduce detection jitter and errors caused by obstacle occlusion, thereby improving the stability and accuracy of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments.

[0010] As a preferred solution, the method of collecting a plurality of initial near-infrared images according to a preset time interval and extracting contour points of each of the initial near-infrared images specifically includes:

[0011] Collecting a number of initial near-infrared images according to a preset time interval;

[0012] Filter each of the initial near-infrared images according to a preset image filtering algorithm to determine a second near-infrared image;

[0013] performing binarization processing on the second near infrared image to obtain a third near infrared image;

[0014] Performing edge detection on the third near infrared image according to a preset edge detection algorithm to obtain an edge of each of the third near infrared images;

[0015] Contour points are extracted from the edge of each of the third near-infrared images according to a preset contour point extraction algorithm to determine the contour points of the initial near-infrared image.

[0016] This preferred solution obtains contour points by filtering, binarizing and edge detecting the initial near-infrared image in sequence, removes noise and interference information in the image through filtering, and reduces the complexity of the image through binarization, thereby improving the accuracy and efficiency of subsequent contour point extraction, and further improving the accuracy and efficiency of target identification based on contour points.

[0017] As a preferred solution, the initial near-infrared image is screened according to the preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images, specifically including:

[0018] Screening initial near infrared images whose number of contour points is greater than a preset threshold value of the number of contour points to obtain a plurality of fourth near infrared images;

[0019] The roundness value of each of the fourth near-infrared images is calculated according to a preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness values ​​to obtain a plurality of first near-infrared images.

[0020] This preferred solution can quickly screen out images that meet the target recognition requirements by screening the initial near-infrared image based on the number of contour points, and can further reduce the interference of noise and obstacles on target recognition by calculating the roundness value, thereby improving the accuracy of target recognition in complex environments and improving the accuracy of target recognition.

[0021] As a preferred solution, the roundness value of each of the fourth near-infrared images is calculated according to a preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness value to obtain a plurality of first near-infrared images, specifically including:

[0022] Determining a contour graph of each of the fourth near infrared images according to the contour points of each of the fourth near infrared images;

[0023] Obtaining the perimeter and area of ​​the contour graphic according to a preset graphic calculation function;

[0024] Determining a graphic roundness value of the contour graphic according to the perimeter and area of ​​the contour graphic;

[0025] Determining a roundness value of each of the fourth near-infrared images based on the graphic roundness value of the contour graphic;

[0026] The fourth near infrared images having a roundness value greater than a preset roundness threshold are screened to obtain a plurality of first near infrared images.

[0027] This preferred solution can screen near-infrared images based on the geometric feature of roundness value by calculating the roundness value, thereby eliminating near-infrared images that do not contain targets, reducing the amount of data that needs to be processed later and improving the efficiency of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments.

[0028] As a preferred solution, the step of generating a target fitting circle for each of the first near-infrared images based on the contour points, and identifying the target based on the radius of the target fitting circle specifically includes:

[0029] Fitting the contour points of each of the first near-infrared images according to a preset fitting algorithm, generating a target fitting circle of each of the first near-infrared images, and determining a target fitting circle set; wherein the target fitting circle includes: center coordinates and radius;

[0030] Sorting the target fitting circles in the target fitting circle set according to the size of the radius, and obtaining a first target fitting circle with a maximum radius and a second target fitting circle with a minimum radius;

[0031] After determining that the radius difference between the first target fitting circle and the second target fitting circle is less than a preset radius threshold, the first target fitting circle and the second target fitting circle are removed from the target fitting circle set to obtain a first target fitting circle set;

[0032] Calculating the average values ​​of the center coordinates and the radius in the first target fitting circle set to obtain the target center coordinates and the target radius;

[0033] The target position is determined according to the target center coordinates and the target radius to complete the identification of the target.

[0034] This preferred solution screens the radius of the target fitting circle at different times. By combining the time dimension and the radius of the fitting circle, it can effectively identify slight position changes of the target when it is obstructed by obstacles, reduce detection jitter and errors caused by obstacle obstruction, and thus improve the stability and accuracy of target identification.

[0035] Accordingly, an embodiment of the present invention provides a target recognition device for eliminating obstacle interference, comprising: a contour point extraction module, a roundness value calculation module and a target recognition module;

[0036] The contour point extraction module is used to collect a number of initial near-infrared images according to a preset time interval, and extract contour points of the initial near-infrared images;

[0037] The roundness value calculation module is used to screen the initial near-infrared image according to a preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images;

[0038] The target recognition module is used to generate a target fitting circle for each of the first near-infrared images based on the contour points, and to recognize a target based on the radius of the target fitting circle.

[0039] As a preferred solution, the contour point extraction module includes: a contour point extraction unit;

[0040] The contour point extraction unit is used to collect a number of initial near-infrared images according to a preset time interval;

[0041] Filter each of the initial near-infrared images according to a preset image filtering algorithm to determine a second near-infrared image;

[0042] performing binarization processing on the second near infrared image to obtain a third near infrared image;

[0043] Performing edge detection on the third near infrared image according to a preset edge detection algorithm to obtain an edge of each of the third near infrared images;

[0044] Contour points are extracted from the edge of each of the third near-infrared images according to a preset contour point extraction algorithm to determine the contour points of the initial near-infrared image.

[0045] As a preferred solution, the roundness value calculation module includes: a roundness value calculation unit;

[0046] The roundness value calculation unit is used to screen the initial near infrared images whose number of contour points is greater than a preset contour point number threshold, to obtain a plurality of fourth near infrared images;

[0047] The roundness value of each of the fourth near-infrared images is calculated according to a preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness values ​​to obtain a plurality of first near-infrared images.

[0048] As a preferred solution, the roundness value calculation unit includes: a roundness value calculation subunit;

[0049] The roundness value calculation subunit is used to determine the contour graph of each of the fourth near infrared images according to the contour points of each of the fourth near infrared images;

[0050] Obtaining the perimeter and area of ​​the contour graphic according to a preset graphic calculation function;

[0051] Determining a graphic roundness value of the contour graphic according to the perimeter and area of ​​the contour graphic;

[0052] Determining a roundness value of each of the fourth near-infrared images based on the graphic roundness value of the contour graphic;

[0053] The fourth near infrared images having a roundness value greater than a preset roundness threshold are screened to obtain a plurality of first near infrared images.

[0054] As a preferred solution, the target recognition module includes: a target recognition unit;

[0055] The target recognition unit is used to fit the contour points of each of the first near-infrared images according to a preset fitting algorithm, generate a target fitting circle of each of the first near-infrared images, and determine a target fitting circle set; wherein the target fitting circle includes: center coordinates and radius;

[0056] Sorting the target fitting circles in the target fitting circle set according to the size of the radius, and obtaining a first target fitting circle with a maximum radius and a second target fitting circle with a minimum radius;

[0057] After determining that the radius difference between the first target fitting circle and the second target fitting circle is less than a preset radius threshold, the first target fitting circle and the second target fitting circle are removed from the target fitting circle set to obtain a first target fitting circle set;

[0058] Calculating the average values ​​of the center coordinates and the radius in the first target fitting circle set to obtain the target center coordinates and the target radius;

[0059] The target position is determined according to the target center coordinates and the target radius to complete the identification of the target.

[0060] It can be understood that, compared with the prior art, the present device collects near-infrared images at different times, then extracts the contour points of the near-infrared images, determines the roundness value of the near-infrared images and the radius of the target fitting circle by the contour points, and then uses the roundness value and the radius of the target fitting circle to screen the geometric features of the target. By screening multiple geometric features, the target can be identified more accurately; by screening the radius of the target fitting circle at different times, the combination of the time dimension and the radius of the fitting circle can effectively identify slight position changes of the target in the case of obstacle occlusion, reduce detection jitter and errors caused by obstacle occlusion, thereby improving the stability and accuracy of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 : A flowchart of a target identification method for eliminating obstacle interference provided by an embodiment of the present invention;

[0062] Figure 2 : A flowchart of another target identification method for eliminating obstacle interference provided by an embodiment of the present invention;

[0063] Figure 3 : A schematic diagram of the structure of a target recognition device for eliminating obstacle interference provided by an embodiment of the present invention;

[0064] Among them, 201: contour point extraction module; 202: roundness value calculation module; 203: target recognition module. DETAILED DESCRIPTION

[0065] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0066] Embodiment 1

[0067] Please refer to Figure 1 , which is a step flow chart of a target identification method for eliminating obstacle interference provided by an embodiment of the present invention, including steps S101 to S103.

[0068] Step S101: collecting a number of initial near-infrared images at preset time intervals, and extracting contour points of the initial near-infrared images.

[0069] In this embodiment, the step of collecting a plurality of initial near-infrared images according to a preset time interval and extracting contour points of each of the initial near-infrared images specifically includes:

[0070] Collecting a number of initial near-infrared images according to a preset time interval;

[0071] Filter each of the initial near-infrared images according to a preset image filtering algorithm to determine a second near-infrared image;

[0072] performing binarization processing on the second near infrared image to obtain a third near infrared image;

[0073] Performing edge detection on the third near infrared image according to a preset edge detection algorithm to obtain an edge of each of the third near infrared images;

[0074] Contour points are extracted from the edge of each of the third near-infrared images according to a preset contour point extraction algorithm to determine the contour points of the initial near-infrared image.

[0075] In an optional embodiment, the time interval is set to 1s, and the near-infrared camera is controlled to collect a near-infrared image once every 1s, so as to determine several initial near-infrared images, and then each initial near-infrared image is filtered using a Gaussian filtering algorithm (i.e., a preset image filtering algorithm) to determine a second near-infrared image, and then the second near-infrared image is binarized using a global threshold method to obtain a third near-infrared image; then the third near-infrared image is edge detected using a Canny edge detection algorithm (i.e., a preset edge detection algorithm) to obtain the edge of each of the third near-infrared images, and finally cv2.findContours() (i.e., a preset contour point extraction algorithm) of the OpenCV library is used to extract contour points from the edge of each of the third near-infrared images to determine the contour points of the initial near-infrared image.

[0076] It should be noted that the preset image filtering algorithm described in this embodiment includes but is not limited to Gaussian filtering algorithm, mean filtering algorithm and other algorithms for filtering out image noise; the edge detection algorithm includes but is not limited to Roberts edge detection algorithm, Canny edge detection algorithm, Laplacian edge detection algorithm and other algorithms for identifying a set of pixel points with drastic brightness changes in an image; the binarization processing can also use grayscale gradient-based methods, adaptive threshold methods and other algorithms; OpenCV library (Open Source Computer Vision Library) is a cross-platform computer vision and machine learning software library, which is often used for processing and calculating images, and cv2.findContours() is often used to extract contour points.

[0077] This embodiment obtains contour points by filtering, binarizing and edge detecting the initial near-infrared image in sequence, removes noise and interference information in the image by filtering, and reduces the complexity of the image by binarization, thereby improving the accuracy and efficiency of subsequent contour point extraction, and further improving the accuracy and efficiency of target identification based on contour points.

[0078] Step S102: screening the initial near-infrared image according to a preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images.

[0079] In this embodiment, the initial near-infrared image is screened according to the preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images, specifically including:

[0080] Screening initial near infrared images whose number of contour points is greater than a preset threshold value of the number of contour points to obtain a plurality of fourth near infrared images;

[0081] The roundness value of each of the fourth near-infrared images is calculated according to a preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness values ​​to obtain a plurality of first near-infrared images.

[0082] In an optional embodiment, the preset contour point number threshold is set to 0. When the number of contour points of the initial near-infrared image is less than the contour point number threshold, it indicates that the initial near-infrared image is not the target to be identified, thereby eliminating interference and not identifying it in subsequent steps. When the target is completely blocked by an obstacle, the acquired near-infrared image cannot identify any valid contour points. Therefore, this embodiment sets the preset contour point number threshold to 0, which can eliminate near-infrared images that cannot identify the target, thereby improving the efficiency and accuracy of subsequent target identification.

[0083] This embodiment can quickly screen out images that meet the target recognition requirements by screening the initial near-infrared image based on the number of contour points, and can further reduce the interference of noise and obstacles on target recognition by calculating the roundness value, thereby improving the accuracy of target recognition in complex environments and improving the recognition accuracy of the target.

[0084] In this embodiment, the roundness value of each of the fourth near-infrared images is calculated according to the preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness value to obtain a plurality of first near-infrared images, specifically including:

[0085] Determining a contour graph of each of the fourth near infrared images according to the contour points of each of the fourth near infrared images;

[0086] Obtaining the perimeter and area of ​​the contour graphic according to a preset graphic calculation function;

[0087] Determining a graphic roundness value of the contour graphic according to the perimeter and area of ​​the contour graphic;

[0088] Determining a roundness value of each of the fourth near-infrared images based on the graphic roundness value of the contour graphic;

[0089] The fourth near infrared images having a roundness value greater than a preset roundness threshold are screened to obtain a plurality of first near infrared images.

[0090] In an optional embodiment, the cv2.arcLength() function and the cv2.contourArea() function in the OpenCV library are used to calculate the perimeter and area of ​​the contour graphic respectively, and then the graphic roundness value of the contour graphic is calculated according to the roundness calculation formula, so as to determine the roundness value of each of the fourth near-infrared images;

[0091] Among them, e is used to represent the roundness value (i.e. the roundness value of the figure), C is used to represent the circumference, S is used to represent the area, and combined with the constant π, the roundness calculation formula is:

[0092]

[0093] The preset roundness threshold is set to 0.8, and the fourth near-infrared images with roundness values ​​greater than 0.8 are screened to obtain a plurality of first near-infrared images.

[0094] It should be noted that the roundness value can be used to calculate the degree of proximity between the target in the near-infrared image and the circle. When the roundness value is closer to 1, it means that the contour figure is closer to a circle, that is, closer to the shape of the target. When the target is blocked by an obstacle, it will cause the figure in the captured near-infrared image to be deformed, and the place that is not similar to a circle can be effectively identified by calculating the roundness value, thereby improving the accuracy of target identification.

[0095] By calculating the roundness value, this embodiment can screen the near-infrared images based on the roundness value as a geometric feature, thereby eliminating the near-infrared images that do not contain the target, reducing the amount of data that needs to be processed later and improving the efficiency of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments.

[0096] Step S103: generating a target fitting circle for each of the first near-infrared images based on the contour points, and identifying the target based on the radius of the target fitting circle.

[0097] In this embodiment, generating a target fitting circle of each first near infrared image based on the contour points, and identifying the target based on the radius of the target fitting circle specifically includes:

[0098] Fitting the contour points of each of the first near-infrared images according to a preset fitting algorithm, generating a target fitting circle of each of the first near-infrared images, and determining a target fitting circle set; wherein the target fitting circle includes: center coordinates and radius;

[0099] Sorting the target fitting circles in the target fitting circle set according to the size of the radius, and obtaining a first target fitting circle with a maximum radius and a second target fitting circle with a minimum radius;

[0100] After determining that the radius difference between the first target fitting circle and the second target fitting circle is less than a preset radius threshold, the first target fitting circle and the second target fitting circle are removed from the target fitting circle set to obtain a first target fitting circle set;

[0101] Calculating the average values ​​of the center coordinates and the radius in the first target fitting circle set to obtain the target center coordinates and the target radius;

[0102] The target position is determined according to the target center coordinates and the target radius to complete the identification of the target.

[0103] In an optional embodiment, the least squares method is used to fit the contour points of the first near-infrared image to generate a target fitting circle for each of the first near-infrared images, and the center coordinates and radius of each target fitting circle are determined; then, the radii of the target fitting circles are sorted from small to large, and the smallest radius is subtracted from the largest radius. When the difference is less than a preset radius threshold, the target fitting circles corresponding to the largest radius and the smallest radius are eliminated, and the average radius and average center coordinates of other target fitting circles are calculated, and the average radius and average center coordinates are used as the radius and center of the target to complete the identification of the target.

[0104] It should be noted that the preset radius threshold can be determined based on the actual radius of the target to be identified; in addition to the least squares method, the contour points can be fitted using fitting algorithms such as the random sampling consensus algorithm and the gradient-based method to determine the target fitting circle; when the radius difference of the target fitting circle at two different times is greater than the preset radius threshold, it indicates that between the two times, there is an obstacle that interferes with the target, causing the target to shake and deform, so by comparing the difference in the radius of the target fitting circle, the interference of the obstacle on the target identification can be eliminated, thereby improving the accuracy of target identification.

[0105] In an optional embodiment, in addition to collecting a number of initial near-infrared images according to a preset time interval, the initial near-infrared images at different times can also be refined and distinguished in frames, for example, greater than or equal to 3 frames, and then the target is identified through the above steps S101 to S103. By further refining the intervals of the near-infrared images, the accuracy of target identification can be further improved.

[0106] This embodiment screens the radius of the target fitting circle at different times and combines the time dimension with the radius of the fitting circle to effectively identify slight position changes of the target when it is obstructed by obstacles, reduce detection jitter and errors caused by obstacle obstruction, and thus improve the stability and accuracy of target identification.

[0107] In an alternative embodiment, please refer to Figure 2 , is a flowchart of another target identification method for eliminating obstacle interference provided by an embodiment of the present invention, such as Figure 2 As shown, first obtain a near-infrared image, then extract contour points, and after judging that the number of contour points is greater than a threshold, calculate the roundness of the near-infrared image. After judging that the roundness is greater than the threshold, calculate the radius difference of consecutive frames of the near-infrared image. After judging that the radius difference is less than the threshold, output the target recognition result.

[0108] This embodiment collects near-infrared images at different times, then extracts contour points of the near-infrared images, determines the roundness value of the near-infrared images and the radius of the target fitting circle by the contour points, and then implements the screening of the geometric features of the target by the roundness value and the radius of the target fitting circle. By screening multiple geometric features, the target can be identified more accurately; by screening the radius of the target fitting circle at different times, the combination of the time dimension and the radius of the fitting circle can effectively identify the slight position change of the target when it is blocked by obstacles, reduce the detection jitter and error caused by obstacle blocking, and thus improve the stability and accuracy of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments.

[0109] Embodiment 2

[0110] Please refer to Figure 3 , which is a structural schematic diagram of a target recognition device for eliminating obstacle interference provided by an embodiment of the present invention, including: a contour point extraction module 201, a roundness value calculation module 202 and a target recognition module 203.

[0111] The contour point extraction module 201 is used to collect a number of initial near-infrared images according to a preset time interval, and extract contour points of the initial near-infrared images.

[0112] In this embodiment, the contour point extraction module 201 includes: a contour point extraction unit;

[0113] The contour point extraction unit is used to collect a number of initial near-infrared images according to a preset time interval;

[0114] Filter each of the initial near-infrared images according to a preset image filtering algorithm to determine a second near-infrared image;

[0115] performing binarization processing on the second near infrared image to obtain a third near infrared image;

[0116] Performing edge detection on the third near infrared image according to a preset edge detection algorithm to obtain an edge of each of the third near infrared images;

[0117] Contour points are extracted from the edge of each of the third near-infrared images according to a preset contour point extraction algorithm to determine the contour points of the initial near-infrared image.

[0118] The roundness value calculation module 202 is used to screen the initial near-infrared image according to a preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images.

[0119] In this embodiment, the roundness value calculation module 202 includes: a roundness value calculation unit;

[0120] The roundness value calculation unit is used to screen the initial near infrared images whose number of contour points is greater than a preset contour point number threshold, to obtain a plurality of fourth near infrared images;

[0121] The roundness value of each of the fourth near-infrared images is calculated according to a preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness values ​​to obtain a plurality of first near-infrared images.

[0122] In this embodiment, the roundness value calculation unit includes: a roundness value calculation subunit;

[0123] The roundness value calculation subunit is used to determine the contour graph of each of the fourth near infrared images according to the contour points of each of the fourth near infrared images;

[0124] Obtaining the perimeter and area of ​​the contour graphic according to a preset graphic calculation function;

[0125] Determining a graphic roundness value of the contour graphic according to the perimeter and area of ​​the contour graphic;

[0126] Determining a roundness value of each of the fourth near-infrared images based on the graphic roundness value of the contour graphic;

[0127] The fourth near infrared images having a roundness value greater than a preset roundness threshold are screened to obtain a plurality of first near infrared images.

[0128] The target recognition module 203 is used to generate a target fitting circle for each of the first near-infrared images based on the contour points, and recognize the target based on the radius of the target fitting circle.

[0129] In this embodiment, the target recognition module 203 includes: a target recognition unit;

[0130] The target recognition unit is used to fit the contour points of each of the first near-infrared images according to a preset fitting algorithm, generate a target fitting circle of each of the first near-infrared images, and determine a target fitting circle set; wherein the target fitting circle includes: center coordinates and radius;

[0131] Sorting the target fitting circles in the target fitting circle set according to the size of the radius, and obtaining a first target fitting circle with a maximum radius and a second target fitting circle with a minimum radius;

[0132] After determining that the radius difference between the first target fitting circle and the second target fitting circle is less than a preset radius threshold, the first target fitting circle and the second target fitting circle are removed from the target fitting circle set to obtain a first target fitting circle set;

[0133] Calculating the average values ​​of the center coordinates and the radius in the first target fitting circle set to obtain the target center coordinates and the target radius;

[0134] The target position is determined according to the target center coordinates and the target radius to complete the identification of the target.

[0135] This embodiment collects near-infrared images at different times, then extracts contour points of the near-infrared images, determines the roundness value of the near-infrared images and the radius of the target fitting circle by the contour points, and then implements the screening of the geometric features of the target by the roundness value and the radius of the target fitting circle. By screening multiple geometric features, the target can be identified more accurately; by screening the radius of the target fitting circle at different times, the combination of the time dimension and the radius of the fitting circle can effectively identify the slight position change of the target when it is blocked by obstacles, reduce the detection jitter and error caused by obstacle blocking, and thus improve the stability and accuracy of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments.

[0136] In summary, the embodiment of the present invention collects near-infrared images at different times, then extracts contour points of the near-infrared images, determines the roundness value of the near-infrared images and the radius of the target fitting circle by the contour points, and then implements the screening of the geometric features of the target by the roundness value and the radius of the target fitting circle. By screening multiple geometric features, the target can be identified more accurately; by screening the radius of the target fitting circle at different times, the combination of the time dimension and the radius of the fitting circle can effectively identify slight position changes of the target in the case of obstacle occlusion, reduce detection jitter and errors caused by obstacle occlusion, thereby improving the stability and accuracy of target identification; by calculating the roundness value, the interference of noise and obstacles on target identification can be further reduced, thereby improving the accuracy of target identification in complex environments.

[0137] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A target recognition method eliminating obstacle interference, characterized in that: include: Collecting a number of initial near-infrared images according to a preset time interval, and extracting contour points of the initial near-infrared images; The initial near-infrared image is screened according to a preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images; A target fitting circle of each of the first near infrared images is generated based on the contour points, and a target is identified based on a radius of the target fitting circle.

2. A target recognition method for eliminating obstacle interference as claimed in claim 1, characterized in that: The step of collecting a plurality of initial near-infrared images according to a preset time interval and extracting contour points of each of the initial near-infrared images specifically includes: Collecting a number of initial near-infrared images according to a preset time interval; Filter each of the initial near-infrared images according to a preset image filtering algorithm to determine a second near-infrared image; performing binarization processing on the second near infrared image to obtain a third near infrared image; Performing edge detection on the third near infrared image according to a preset edge detection algorithm to obtain an edge of each of the third near infrared images; Contour points are extracted from the edge of each of the third near-infrared images according to a preset contour point extraction algorithm to determine the contour points of the initial near-infrared image.

3. A target recognition method for eliminating obstacle interference as claimed in claim 1, characterized in that: The step of screening the initial near infrared image according to the preset roundness value algorithm and the contour points to obtain a plurality of first near infrared images specifically includes: Screening initial near infrared images whose number of contour points is greater than a preset threshold value of the number of contour points to obtain a plurality of fourth near infrared images; The roundness value of each of the fourth near-infrared images is calculated according to a preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness values ​​to obtain a plurality of first near-infrared images.

4. A target recognition method for eliminating obstacle interference as claimed in claim 3, characterized in that: The step of calculating the roundness value of each of the fourth near infrared images according to the preset roundness value algorithm and the contour points, and screening the fourth near infrared images based on the roundness value to obtain a plurality of first near infrared images specifically includes: Determining a contour graph of each of the fourth near infrared images according to the contour points of each of the fourth near infrared images; Obtaining the perimeter and area of ​​the contour graphic according to a preset graphic calculation function; Determining a graphic roundness value of the contour graphic according to the perimeter and area of ​​the contour graphic; Determining a roundness value of each of the fourth near-infrared images based on the graphic roundness value of the contour graphic; The fourth near infrared images having a roundness value greater than a preset roundness threshold are screened to obtain a plurality of first near infrared images.

5. A target recognition method for eliminating obstacle interference as claimed in claim 1, characterized in that: Generating a target fitting circle of each of the first near-infrared images based on the contour points, and identifying the target based on the radius of the target fitting circle, specifically includes: Fitting the contour points of each of the first near-infrared images according to a preset fitting algorithm, generating a target fitting circle of each of the first near-infrared images, and determining a target fitting circle set; wherein the target fitting circle includes: center coordinates and radius; Sorting the target fitting circles in the target fitting circle set according to the size of the radius, and obtaining a first target fitting circle with a maximum radius and a second target fitting circle with a minimum radius; After determining that the radius difference between the first target fitting circle and the second target fitting circle is less than a preset radius threshold, the first target fitting circle and the second target fitting circle are removed from the target fitting circle set to obtain a first target fitting circle set; Calculating the average values ​​of the center coordinates and the radius in the first target fitting circle set to obtain the target center coordinates and the target radius; The target position is determined according to the target center coordinates and the target radius to complete the identification of the target.

6. A target recognition device that eliminates obstacle interference, characterized in that: include: Contour point extraction module, roundness value calculation module and target recognition module; The contour point extraction module is used to collect a number of initial near-infrared images according to a preset time interval, and extract contour points of the initial near-infrared images; The roundness value calculation module is used to screen the initial near-infrared image according to a preset roundness value algorithm and the contour points to obtain a plurality of first near-infrared images; The target recognition module is used to generate a target fitting circle for each of the first near-infrared images based on the contour points, and to recognize a target based on the radius of the target fitting circle.

7. A target recognition device for eliminating obstacle interference as claimed in claim 6, characterized in that: The contour point extraction module comprises: a contour point extraction unit; The contour point extraction unit is used to collect a number of initial near-infrared images according to a preset time interval; Filter each of the initial near-infrared images according to a preset image filtering algorithm to determine a second near-infrared image; performing binarization processing on the second near infrared image to obtain a third near infrared image; Performing edge detection on the third near infrared image according to a preset edge detection algorithm to obtain an edge of each of the third near infrared images; Contour points are extracted from the edge of each of the third near-infrared images according to a preset contour point extraction algorithm to determine the contour points of the initial near-infrared image.

8. A target recognition device for eliminating obstacle interference as claimed in claim 6, characterized in that: The roundness value calculation module includes: a roundness value calculation unit; The roundness value calculation unit is used to screen the initial near infrared images whose number of contour points is greater than a preset contour point number threshold, to obtain a plurality of fourth near infrared images; The roundness value of each of the fourth near-infrared images is calculated according to a preset roundness value algorithm and the contour points, and the fourth near-infrared images are screened based on the roundness values ​​to obtain a plurality of first near-infrared images.

9. A target recognition device for eliminating obstacle interference as claimed in claim 8, characterized in that: The roundness value calculation unit includes: a roundness value calculation subunit; The roundness value calculation subunit is used to determine the contour graph of each of the fourth near infrared images according to the contour points of each of the fourth near infrared images; Obtaining the perimeter and area of ​​the contour graphic according to a preset graphic calculation function; Determining a graphic roundness value of the contour graphic according to the perimeter and area of ​​the contour graphic; Determining a roundness value of each of the fourth near-infrared images based on the graphic roundness value of the contour graphic; The fourth near infrared images having a roundness value greater than a preset roundness threshold are screened to obtain a plurality of first near infrared images.

10. A target recognition device for eliminating obstacle interference as claimed in claim 6, characterized in that: The target recognition module comprises: a target recognition unit; The target recognition unit is used to fit the contour points of each of the first near-infrared images according to a preset fitting algorithm, generate a target fitting circle of each of the first near-infrared images, and determine a target fitting circle set; wherein the target fitting circle includes: center coordinates and radius; Sorting the target fitting circles in the target fitting circle set according to the size of the radius, and obtaining a first target fitting circle with a maximum radius and a second target fitting circle with a minimum radius; After determining that the radius difference between the first target fitting circle and the second target fitting circle is less than a preset radius threshold, the first target fitting circle and the second target fitting circle are removed from the target fitting circle set to obtain a first target fitting circle set; Calculating the average values ​​of the center coordinates and the radius in the first target fitting circle set to obtain the target center coordinates and the target radius; The target position is determined according to the target center coordinates and the target radius to complete the identification of the target.

Images