A classification method for automotive seatbelt latch models based on visual detection

Abstract

This invention proposes a classification method for automotive seatbelt latch models based on visual detection. Considering that the basic characteristics of different seatbelt latch models are consistent, including length, width, positioning circle, and the shortest vertical distance from the positioning circle to the long side edge, the invention is characterized by identifying and locating the seatbelt latches based on the Hough transform circle detection algorithm, fitting the edge line, and counting the number of pixels at the intersection of the edge line and the edge detection image to distinguish different seatbelt latch models. It has significant advantages such as simple algorithm, simple operation, and easy expansion and implementation.

Description

Technical Field

[0001] This invention relates to a classification method for automotive seatbelt latch models based on visual detection. Background Technology

[0002] Car seat belts are designed to ensure passenger safety in emergencies, and the seat belt buckle needs to ensure a secure connection between the two ends of the seat belt. The seat belt latch is the core component of the seat belt buckle, requiring a tight connection between the spring and the pin to lock the latch when it is inserted and release it when it is unlocked. This component requires extremely precise manufacturing in the automotive industry. Because the latch directly determines whether the latch can be properly locked and unlocked, the quality inspection, classification, and installation of the seat belt latch all require extremely rigorous procedures.

[0003] Image processing technology is increasingly widely used in industrial production. Through preprocessing, segmentation, information extraction, and comparison of images, image processing technology greatly improves the efficiency of product inspection, classification, and transportation in industrial production. Many image difference comparison algorithms exist. Pixel-by-pixel comparison algorithms can set the allowed number of different pixels, the tolerable number of color differences and color difference tolerance, pixel transparency and opacity, and use masks for image comparison; the Hausdorff distance algorithm can compare the distance and offset of each point in the image, generally used to detect images with very small displacement deviations; color aggregation vectors classify pixels as coherent or incoherent based on whether each pixel in a given color bucket belongs to a large region of similar colors. However, these methods are relatively complex and computationally intensive, making them unsuitable for the rapid classification of bolt types.

[0004] Therefore, this invention proposes a method for distinguishing different models of car seat belt latches based on image processing technology. Based on the common positioning circle of different seat belt latch models and the differences in details on the long side edge, the method uses edge detection algorithm and Hough transform circle detection algorithm to locate with a coordinate point and distinguish details with a line to achieve model differentiation. It has significant advantages such as simple algorithm, simple operation, easy expansion and implementation. Summary of the Invention

[0005] The purpose of this invention is to propose a classification method for automotive seat belt latch models based on visual detection. Considering that the basic characteristics of different seat belt latch models are consistent, including length, width, positioning circle, and the shortest vertical distance from the positioning circle to the long side edge, this method uses the Hough transform circle detection algorithm to identify and locate the seat belt latches, fits the edge line, and counts the number of intersections between the edge line and the edge detection image to distinguish different seat belt latch models. It has significant advantages such as simple algorithm, simple operation, and easy expansion and implementation.

[0006] To achieve the above-mentioned objectives, this invention proposes a classification method for automotive seatbelt latch models based on visual detection, characterized by the following steps:

[0007] 1) Use vernier calipers to measure the length x and width y of the seat belt latch, the diameter e of the positioning circle, and the shortest vertical distance g from the edge of the positioning circle to the long side edge;

[0008] 2) Take a picture of the orthographic projection of the seat belt buckle at the same distance from the seat belt buckle with a camera to obtain an image with a resolution of a and a pixel size of b×c. Calculate the size of the image in mm by b÷a×25.4 and c÷a×25.4. Compare the image with the length x and width y of the seat belt buckle to obtain the ratio d between the image and the actual size.

[0009] 3) Apply the imresize method to compress the seat belt latch image by a factor of i to obtain a compressed image with a pixel size of b'×c';

[0010] 4) Apply the Canny operator to perform edge detection on the two compressed images respectively, set the sensitivity value to 0.7, and obtain the edge detection effect image of the bolt;

[0011] 5) Based on the features of the seat belt bolt positioning groove, the pixel value f corresponding to the diameter of the positioning circle is obtained as (e×d÷25.4×a×i), and set as the diameter detection interval. The Hough transform is applied to the edge detection effect image obtained in step 4) to perform circle detection, and the corresponding center pixel coordinates are obtained. These coordinates are then used as the center point of the image for positioning.

[0012] 6) Taking the center of the positioning circle as the starting point, determine the edge line parallel to the long side edge of the bolt in the direction perpendicular to the long side edge of the bolt, based on the shortest vertical distance g from the edge of the positioning circle to the long side edge;

[0013] 7) Count the number h of pixels where the edge line intersects with the image edge detection map;

[0014] 8) Repeat steps 2) to 7) to obtain the number of pixels hA, hB, etc., at the intersection of the edge line of different types of bolts and the edge detection map of the image;

[0015] 9) Compare the number of intersection pixels obtained in steps 7) and 8), and identify the bolt type by setting a reasonable threshold. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the implementation of the method of the present invention.

[0017] Figure 2 The images are orthographic projections of the seatbelt locking bolts. (a) shows model A, and (b) shows model B.

[0018] Figure 3 The images show the edge detection results. (a) is for model A, and (b) is for model B.

[0019] Figure 4 The images show the test results. (a) is for model A and (b) is for model B. Detailed Implementation

[0020] To further illustrate the effects and advantages of the present invention, a detailed description of the invention will be provided below with reference to simulation results of a specific embodiment. All results below were obtained through simulation using MATLAB R2016b software, and the implementation flowchart is attached. Figure 1 As shown: 1) Using vernier calipers, the length x of the seatbelt locking bolt is 44.6 mm, the width y is 27.1 mm, and the diameter e of the positioning circle is 8 mm.

[0021] The shortest vertical distance g from the edge of the positioning circle to the long side edge is 4.4mm;

[0022] 2) Take a frontal projection image of the seatbelt locking bolt of model A with a camera at a distance of 69mm. The resulting image has a resolution of 72 DPI (pixels per inch) and a pixel size of 3000×4000 (b×c). See attached image. Figure 2 As shown in (a), the image size is calculated to be 1088.57 × 1411.11 mm by calculating b ÷ a × 25.4 and c ÷ a × 25.4. This size is then compared with the length x of the seatbelt latch.

[0023] By comparing y, the ratio d between the image and the actual size is found to be 19:1;

[0024] 3) Apply the imresize method to compress the seatbelt bolt image by a factor of i (0.2), resulting in a pixel size of b'×c'.

[0025] A compressed image of 600×800 pixels;

[0026] 4) Apply the Canny operator to perform edge detection on the compressed image, setting the sensitivity value to 0.7, to obtain the edge detection result image of the Model A bolt, as shown in the attached image. Figure 3 As shown in (a);

[0027] 5) Based on the features of the seat belt bolt positioning groove, the pixel value f corresponding to the diameter of the positioning circle is (e×d÷25.4×a×i)86, and it is set as the diameter detection interval (80, 88). The Hough transform is used to perform circle detection on the edge detection effect image obtained in step 4), and the corresponding center pixel coordinates (317, 584) are obtained. This coordinate is used as the center point of the image for positioning.

[0028] 6) Taking the center of the positioning circle as the starting point, and based on the shortest vertical distance *g* from the edge of the positioning circle to the long side edge, determine the edge line parallel to the long side edge of the bolt, perpendicular to the direction of the long side edge of the bolt, as shown in the attached figure. Figure 4 As shown in (a);

[0029] 7) The number of pixels hA at the intersection of the edge line of the A-type bolt and the edge detection map of the image is 8;

[0030] 8) Repeat steps 2) to 7) to obtain the orthographic projection images of the seat belt latches for model B, as shown in the attached figures. Figure 2 (b) shows the edge detection image of the seatbelt latch for model B, as attached. Figure 3 As shown in (b), the circular positioning and edge line fitting diagram of the seat belt latch of model B are attached. Figure 4 As shown in (b), the number of pixels hB at the intersection of the edge line of the B-type bolt and the image edge detection map is 4.

[0031] 9) Compare the number of intersection pixels obtained in steps 7) and 8), and identify the bolt type by setting a reasonable threshold of 6.

[0032] The above embodiments demonstrate the effectiveness of the classification method for automotive seatbelt latch models proposed in this invention.

[0033] It should be noted that the above embodiments are not intended to limit the scope of protection of the present invention. Equivalent transformations or substitutions made on the basis of the above technical solutions, as well as several improvements such as extensions based on them, all fall within the scope of protection of the claims of the present invention.

Claims

1. A method of classifying a type of automobile seat belt buckle, characterized by, Comprising the following steps: 1) Measure the length x, width y of the seat belt lock, the diameter e of the positioning circle, and the vertical shortest distance g from the edge of the positioning circle to the long side edge with a vernier caliper; 2) Take a front projection image of the seat belt lock with a camera at the same distance as the seat belt lock, obtain an image with resolution a and pixel size b x c, calculate the size of the image in mm through b ÷ a x 25.4, c ÷ a x 25.4, and compare it with the length x and width y of the seat belt lock to obtain the size ratio d of the image and the real object; 3) Apply the imresize method to compress the seat belt lock image by i times to obtain a compressed image with pixel size b' x c'; 4) Apply the Canny operator to the two compressed images for edge detection, set the sensitivity value to 0.7, and obtain the edge detection effect diagram of the lock; 5) According to the positioning circle groove feature of the seat belt lock, obtain the pixel value f corresponding to the diameter of the positioning circle (e x d ÷ 25.4 x a x i), and set it as the diameter detection interval. Apply Hough transform to the edge detection effect diagram obtained in step 4) for circle detection to obtain the corresponding center pixel coordinates, and take this coordinate as the center point of the image for positioning; 6) Take the center of the positioning circle as the starting point, determine the edge line parallel to the long side edge of the lock according to the vertical shortest distance g from the edge of the positioning circle to the long side edge, and perpendicular to the direction of the long side edge of the lock; 7) Count the number of intersection pixels h between the edge line and the image edge detection diagram; 8) Repeat steps 2) to 7) to obtain the intersection pixel number hA, hB of the edge line and the image edge detection diagram of different types of locks; 9) Compare the intersection pixel numbers obtained in steps 7) and 8), set a reasonable threshold value, and identify the type of the lock.

Images