Automatic visual inspection system for defects of high-brightness reflective parts
By using 180-degree reflective dome and neural network technology in the high-reflective part detection system, combined with the active lighting system, the complete outline of the highly reflective parts is solved, and the detection efficiency and accuracy are improved and reduces system costs.
Patent Information
- Application Number
- CN202411776145.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-05
- Publication Date
- 2025-01-17
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The prior art is difficult to effectively detect surface defects of highly reflective parts, especially because the high-gloss mirror objects lack corresponding points during stereo imaging, which leads to difficulty in calculating depth and surface profiles, and the cost of active lighting systems is high.
Using an automatic vision detection system including a 180-degree reflective dome, a first camera, a projector and a second camera, the component with a plurality of phase shift sinusoidal patterns is illuminated by a projector using neural network technology and an active lighting system. Create a texture pattern to capture the complete outline of the part surface using the first and second cameras.
It improves the efficiency and accuracy of high-reflective parts detection, solves the problem of poor imaging effects, and reduces system costs.
Smart Images
Figure CN119322068A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of visual inspection technology, and in particular to an automatic visual inspection system for defects of highlight reflective parts. Background Art
[0002] Inspecting shiny, reflective parts for defects requires automated imaging systems that employ novel lighting and imaging techniques. Components such as automotive accessories and bathroom fixtures are often shiny and chrome-plated. To inspect these parts, manufacturers typically employ human operators to sample the consistency of the plating before these parts are shipped to customers for final assembly. Traditionally, a number of different methods have been used to measure the surface of three-dimensional parts. One of the simplest uses a passive stereo imaging method, which uses ambient light to illuminate the object and a pair of calibrated stereo cameras to image the surface of the object. Since the horizontal displacement of these cameras causes the images captured by the two cameras to overlap, corresponding points can be found between the two, and triangulation is used to determine the depth and surface profile of the object. If the object is a high-gloss, specular object, there may be no or very few corresponding points between the two images. In this case, the depth and surface profile cannot be calculated by capturing two stereo images. The system also suffers if the part contains multiple cavities or steep edges. In such applications, parts of the object may be visible to one camera but not to the other. Since some parts of the image will be occluded, it is difficult to generate surface profile information on the entire part. To overcome this problem to some extent, active lighting systems can be used. In the simplest form, a single structured light projection is used where a stripe of light is projected onto an object as it moves down a conveyor. Alternatively, the stripe of light moves over the object. The reflected structured light from the object is then digitized by a calibrated camera to determine the line profile through the object and construct a 3D surface profile. Here, light may not reflect to the camera from an object with multiple cavities and multiple lasers or laser / camera configurations may be required, increasing the cost of the system. Summary of the invention
[0003] The technical problem solved by the present invention is to provide a new active imaging system, which solves the problem of poor imaging effect of highly reflective objects, applies neural network technology to surface defect detection of highly reflective objects, and improves detection efficiency and accuracy.
[0004] The technical solution adopted by the present invention to solve its technical problem is: an automatic visual inspection system for defects of high-brightness reflective parts, including high-brightness reflective parts of car lights, the high-brightness reflective parts of car lights are covered with a 180-degree reflective dome on the outside, and a first camera, a projector and a second camera are arranged on the 180-degree reflective dome in sequence, the setting angle between the first camera and the projector is 35 degrees, and the setting angle between the projector and the second camera is 35 degrees.
[0005] Further steps are: Establishing algorithm model:
[0006] b x =σ(t x )+c x
[0007] b y =σ(t y )+c y
[0008]
[0009] Where: (c x , c y ) is the center coordinate of the preset bounding box on the feature map;
[0010] (p w , p y ) is the width and height of the preset bounding box on the feature map;
[0011] (t x , t y , t w , t h ) are the center offsets of the bounding box predicted by the network (t x , t y )as well as
[0012] Aspect ratio (t w , t h );
[0013] (b x , b y , b w , b h ) is the final predicted target bounding box;
[0014] σ(x) function is the sigmoid function;
[0015] The (c x , c y )、(p w , p y )、(t x , t y ) and (t w, t h ) are all obtained by the first camera (11), the projector (12) and the second camera (13) according to the specific detection photos. Specific result values.
[0016] Furthermore, the σ(x) function scales the prediction offset to between 0 and 1.
[0017] The beneficial effects of the present invention are as follows: a highly reflective part of a headlight is placed under a 180-degree reflective dome, a projector 12 is used to illuminate the part with multiple phase-shifted sinusoidal patterns, thereby creating a texture on the smooth surface of the object, and then the generated stripe pattern is reflected from the mirror surface of the part and deformed as the slope of the measured surface and the modulated stripe pattern changes, and then captured by the first camera and the second camera, so that the complete contour of the part surface is illuminated, solving the problem of poor imaging of highly reflective objects. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Figure 1 Schematic diagram of the structure of the automatic visual inspection system for defects in highlight reflective parts.
[0019] Marked in the figure are: 180-degree reflection dome 1, high-reflective part of the headlight 2, first camera 11, projector 12, second camera 13. DETAILED DESCRIPTION
[0020] The present invention is further described below in conjunction with the accompanying drawings and specific embodiments.
[0021] like Figure 1 As shown, an automatic visual inspection system for defects of high-brightness reflective parts includes a high-brightness reflective part 2 of a headlight, the high-brightness reflective part 2 of the headlight is covered with a 180-degree reflective dome 1, and a first camera 11, a projector 12, and a second camera 13 are sequentially arranged on the 180-degree reflective dome 1, the setting angle between the first camera 11 and the projector 12 is 35 degrees, and the setting angle between the projector 12 and the second camera 13 is 35 degrees; the model of the projector 12 is STAR-07, and the first camera 11 and the second camera 13 are Basler's 4096x 3000 CMOS-based ace acA411220um USB 3.0 camera, the highly reflective part 2 of the headlight is placed under the 180-degree reflective dome 1, and a projector 12 is used to illuminate the part with multiple phase-shifted sinusoidal patterns to create a texture on the smooth surface of the object. Then, the generated fringe pattern is reflected from the mirror surface of the part and deformed as the slope of the measured surface and the modulated fringe pattern changes, and then captured by the first camera 11 and the second camera 13. In this way, the complete outline of the part surface is illuminated, solving the problem of poor imaging of highly reflective objects.
[0022] On this basis, the algorithm model is established:
[0023] b x =σ(t x )+c x
[0024] b y =σ(t y )+c y
[0025]
[0026]
[0027] Where: (c x , c y ) is the center coordinate of the preset bounding box on the feature map;
[0028] (p w , p y ) is the width and height of the preset bounding box on the feature map;
[0029] (t x , t y , t w , t h ) are the center offsets of the bounding box predicted by the network (t x , t y )as well as
[0030] Aspect ratio (t w , t h );
[0031] (b x , b y , b w , b h ) is the final predicted target bounding box;
[0032] σ(x) function is the sigmoid function;
[0033] The (c x , c y )、(p w , p y )、(t x , t y ) and (t w , t h ) are all obtained by the first camera (11), the projector (12) and the second camera (13) according to the specific detection photos. Specific result values.
[0034] On this basis, the σ(x) function scales the predicted offset to between 0 and 1, thereby accelerating network convergence.
[0035] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. An automatic visual inspection system for defects in high-brightness reflective parts, characterized by: The invention comprises a highly reflective part (2) of a vehicle lamp, wherein the highly reflective part (2) of the vehicle lamp is externally covered with a 180-degree reflective dome (1), and a first camera (11), a projector (12) and a second camera (13) are sequentially arranged on the 180-degree reflective dome (1), wherein the setting angle between the first camera (11) and the projector (12) is 35 degrees, and the setting angle between the projector (12) and the second camera (13) is 35 degrees.
2. An automatic visual inspection system for defects in high-brightness reflective parts as claimed in claim 1, characterized in that: Building an algorithm model: b x =σ(t x )+c x b y =σ(t y )+c y Where: (c x , c y ) is the center coordinate of the preset bounding box on the feature map; (p w , p y ) is the width and height of the preset bounding box on the feature map; (t x , t y , t w , t h ) are the center offsets of the bounding box predicted by the network (t x , t y )as well as Aspect ratio (t w , t h ); (b x , b y , b w , b h ) is the final predicted target bounding box; σ(x) function is the sigmoid function; The x , c y )、(p w , p y )、(t x , t y ) and (t w , t h ) are all obtained by the first camera (11), the projector (12) and the second camera (13) according to the specific detection photos. Specific result values.
3. An automatic visual inspection system for defects in high-brightness reflective parts as claimed in claim 2, characterized in that: The σ(x) function scales the prediction offset to between 0 and 1.