A multi-spectral light source control device for visual recognition
Through the automated light source switching and deep learning algorithm of the multi-spectral light source control device, the problem of long-term and high labor intensity of light source selection in the existing technology is solved, and the optimal spectrum or light source combination is automatically determined, which improves the efficiency of defect detection and accuracy.
Patent Information
- Application Number
- CN202111010528.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-08-31
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2041-08-31
AI Technical Summary
In industrial production, when detecting defects in different target objects, existing light source selection methods require manual switching of different light sources, resulting in increased labor intensity and cost of workers, and it takes a long time to achieve the best lighting effect efficiently.
Design a multi-spectral light source control device, including a light source control unit, a visual recognition device and a data analysis unit, and automatically control the light source to switch within all spectral ranges, and combines a deep learning algorithm to generate a recognition success rate sorting table under all light sources combinations. Automatically determine the optimal spectrum or light source combination.
Automatic switching of light sources without manual intervention is achieved, saving manpower and time, reducing labor intensity and cost, and improving the efficiency and accuracy of defect detection.
Smart Images

Figure CN114494671B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of light sources, and in particular to a multi-spectral light source control device for visual recognition. Background Art
[0002] When inspecting defects of different target objects in industrial production, in addition to cameras and recognition algorithms, light source is also a key influencing factor. Generally, the target object is illuminated by light source to make the defects clearly visible. Therefore, selecting the light source with the best lighting effect in different defect detection scenarios becomes a difficult problem that needs to be solved. The current method of light source selection is to switch different light sources manually, and then compare the recognition effects, and adjust the light source according to the comparison results for further identification. This requires a lot of manpower and time, which increases the labor intensity of workers and the cost of the company. Summary of the invention
[0003] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a multi-spectral light source control device for visual recognition.
[0004] A multi-spectral light source control device for visual recognition, comprising a light source control unit, a light source, a visual recognition device and a data analysis unit; the visual recognition device is used to collect image data, submit it to a training library algorithm model to complete the recognition process, and generate recognition data; the data analysis unit stores all known sample feature types to be evaluated, is connected to the visual recognition device when working, is used to receive the recognition results and corresponding device parameters conveyed by the visual device, and completes the storage statistics of known samples to generate a visual recognition success rate ranking table of all sample features under various spectral segments and light source combinations; the light source control unit is electrically connected to the light source, the light source control unit is connected to the visual recognition device, and the visual recognition device shoots the light source to obtain light source irradiation image information.
[0005] Further, the light source may be one or more, and the light source may be a single-spectrum light source and / or a multi-spectrum light source.
[0006] Furthermore, the visual recognition device includes a camera, a camera gimbal and an image recognition workstation configured with a deep learning-based recognition algorithm. The camera can be configured with one or more according to the requirements of the recognition scene. The camera is set on the camera gimbal. The camera is connected to the image recognition workstation via LAN / USB or the like. The workstation can be a dedicated image server, a desktop computer or a high-performance industrial computer.
[0007] Furthermore, the light source control unit controls the light emission of different spectral segments of a single multi-spectral light source, as well as the sequential and random light emission of multiple different single-spectrum light sources, to obtain the light source lighting image results of the maximum number of different spectral lighting combinations, and after running for a period of time, obtains multiple rounds of recognition results and corresponding parameter light source working parameter data for all combinations of all light source illumination effects.
[0008] Furthermore, the data analysis unit provides the best light source illumination result ranking of the visual scene and displays the corresponding light source parameters through statistics and analysis of the data in the recognition result record table.
[0009] Furthermore, the recognition workstation controls and ensures full coverage of the image recognition effects of all spectral combinations and lighting combinations of all light sources through a loop control program.
[0010] Furthermore, the data analysis unit includes a sample image feature classification unit, a comparison unit for comparing image recognition features with sample features, and a data information processing unit for storing statistics and displaying comparison results.
[0011] The present invention has the following beneficial effects:
[0012] 1. There is no need for manual intervention in light source switching. The light source is continuously switched within all spectral ranges supported by the light source under the automatic control of the light source control unit, completing all possible spectrum and light source working combinations, and automatically recording the working parameters of the corresponding light source, thus realizing the parameter determination of the optimal spectrum or light source combination lighting effect required for visual recognition of a specific sample.
[0013] 2. Save manpower and time, reduce workers' labor intensity and company costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 It is a structural block diagram of the present invention. DETAILED DESCRIPTION
[0015] 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.
[0016] In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as limiting the present invention; the terms "first", "second", and "third" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance. In addition, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.
[0017] A multi-spectral light source control device for visual recognition, comprising a light source control unit, a light source, a visual recognition device and a data analysis unit; the visual recognition device is used to collect image data, submit it to a training library algorithm model to complete the recognition process, and generate recognition data; the data analysis unit stores all known sample feature types to be evaluated, is connected to the visual recognition device when working, is used to receive the recognition results and corresponding device parameters conveyed by the visual device, and completes the storage statistics of known samples to generate a visual recognition success rate ranking table of all sample features under various spectral segments and light source combinations; the light source control unit is electrically connected to the light source, the light source control unit is connected to the visual recognition device, and the visual recognition device shoots the light source to obtain light source irradiation image information.
[0018] Specifically, the light source may be one or more, and the light source may be a single-spectrum light source and / or a multi-spectrum light source.
[0019] Specifically, the visual recognition device includes a camera, a camera gimbal, and an image recognition workstation configured with a deep learning-based recognition algorithm. One or more cameras can be configured according to the requirements of the recognition scene. The camera is set on the camera gimbal. The camera is connected to the image recognition workstation via LAN / USB or the like. The workstation can be a dedicated image server, a desktop computer, or a high-performance industrial computer.
[0020] Specifically, the light source control unit controls the light emission of different spectral segments of a single multi-spectral light source, as well as the sequential and random light emission of multiple different single-spectrum light sources, to obtain the light source lighting image results of the maximum number of different spectral lighting combinations, and after running for a period of time, obtains multiple rounds of recognition results and corresponding parameter light source working parameter data for all combinations of all light source illumination effects.
[0021] Specifically, the data analysis unit provides the best light source illumination result ranking of the visual scene and displays the corresponding light source parameters through data statistics and analysis of the recognition result record table.
[0022] Specifically, the recognition workstation controls and ensures full coverage of the image recognition effects of the spectrum combinations and lighting combinations of all light sources through a loop control program.
[0023] Specifically, the data analysis unit includes a sample image feature classification unit, a comparison unit for comparing image recognition features with sample features, and a data information processing unit for storing statistics and displaying comparison results.
[0024] The data analysis unit will store a large amount of sample image data of the same type, and their features are similar. Then start the light source test process, run for 1 full day, 2 full days or even longer as needed, and the number of cycle tests of all light sources and spectra can be hundreds, thousands, tens of thousands or even hundreds of thousands of times, etc., to see the ranking of recognition results under all spectrum combinations and light source combination effects. The light source combination and corresponding parameters with the highest recognition success rate are the best light source parameters for feature extraction of this type of samples.
[0025] The light source control unit controls the light emission of different spectral segments of the same multi-spectral light source, as well as the sequential and random light emission of multiple different single-spectrum light sources, to obtain the light source illumination image results of the maximum number of different spectral illumination combinations. After a period of operation, all combinations of all light source illumination effects are controlled to obtain multiple rounds of data, combined with artificial intelligence and deep learning algorithms, and the recognition effect of the light source control unit is fed back to the multi-spectral light source control. This process is an automated process that does not require manual intervention. The results are automatically recorded, and finally the spectral value of the light source is selected using this result. For example, if the spectral value is 360nm, a 360nm light source is purchased.
[0026] For example, if there are defects on the surface of a target object, we can run the device for one night for different defects, observe the results the next day, and determine which light source is the best. We then choose that light source without having to do the test manually. Then we use this solution as a basis and let the customer purchase the corresponding light source, thus completing the corresponding light source procurement work.
[0027] Visual recognition defect detection is used to detect surface defects of translucent plastic products, needle tubes of medical devices, needle caps of syringe needles, surface electroplating, and pharmaceutical packaging containers.
[0028] The illumination control of various spectra of the multi-spectral light source is associated with artificial visual recognition and recognition results based on deep learning algorithms, and then a closed loop is formed, and finally the results are output and recorded. For example, if a scene uses red light, the contrast of the defect is very low and the features are hidden. If blue light or white light is used, the effect is very good. The greater the contrast between the defect and the background, the better. This requires adjusting the light source to achieve the best effect.
[0029] Visual recognition equipment uses industrial cameras. Different cameras are used for different scenes. The recognition workstation has strong performance and can judge the effect. For example, if we set ten target objects with defects and run them for more than ten hours, we find that one object without defects is not recognized. We can check which spectrum caused the result, and then look at the pictures saved at that time to conclude that this spectrum is not applicable in this scene.
[0030] The data analysis unit simply lists and sorts these data. The auxiliary program of the workstation recognizes this table, such as queuing it up. It needs to run many rounds, and may run tens of thousands of times in more than ten hours. Then it sorts the data and observes which spectrum has the highest success rate and which spectrum has the lowest lighting success rate. Finally, the result is given. The success rate within this range is 100%, and within this range it becomes 90%. This is the analysis result. Through this process, the light source is selected. For example, a light source may support multiple spectral range working modes, which can be red, green, blue and purple. This is one support method, or another support method. For example, prepare a purple light source, a red light source, a white light source, and random signal quantities in disorder. Each quantity is known. Which quantity is controlled, and then its parameters are recorded to form a multi-spectral light source with a single light source. This is a comprehensive multi-spectral light source. The other is that each light source is a single spectrum. Combine them to see which combination works best. There are mainly two such methods.
[0031] See also Figure 1 , Figure 1The structural block diagram of the present invention is introduced, including a light source control unit, a light source, a visual recognition device and a data analysis unit; the visual recognition device is used to collect image data, submit it to a training library algorithm model to complete the recognition process, and generate recognition data; the data analysis unit stores all known sample feature types to be evaluated, is connected to the visual recognition device when working, is used to receive the recognition results and corresponding device parameters conveyed by the visual device, and completes the storage statistics of known samples to generate a visual recognition success rate ranking table of all sample features under various spectral bands and light source combinations; the light source control unit is electrically connected to the light source, the light source control unit is connected to the visual recognition device, and the visual recognition device shoots the light source to obtain light source irradiation image information.
[0032] Working principle: The light source control unit controls the light emission of different spectral segments of the same multi-spectral light source, as well as the sequential and random light emission of multiple different single-spectrum light sources. The visual recognition device photographs the light source to obtain the light source illumination image information, and obtains the light source illumination image results of the maximum number of different spectral illumination combinations. After a period of operation, all combinations of all light source illumination effects are controlled to obtain multiple rounds of data. The visual recognition device transmits the data to the data analysis unit. The data analysis unit receives the image data and compares the image data with the model data, and generates visual recognition data to feed back to the visual recognition device. The data recognition device feeds back to the light source control unit to finally determine the best light source.
[0033] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments or to make equivalent substitutions for some of the technical features therein. 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. A multi-spectral light source control device for visual recognition, Features: It includes a light source control unit, a light source, a visual recognition device and a data analysis unit; The visual recognition device is used to collect image data, submit it to the training library algorithm model to complete the recognition process, and generate recognition data; The data analysis unit stores all known sample feature types to be evaluated, is connected to the visual recognition device when working, is used to receive the recognition results and corresponding device parameters transmitted by the visual device, and completes the storage statistics of known samples, and generates a visual recognition success rate ranking table of all sample features under various spectral bands and light source combinations, wherein, for each type of sample feature, the light source combination with the highest recognition success rate and the corresponding parameters are the optimal light source parameters for extracting the sample features of this type; the data analysis unit gives the optimal light source illumination result ranking of the visual scene and displays the corresponding light source parameters through data statistics and analysis of the recognition result record table; The light source control unit is electrically connected to the light source, and the light source control unit is connected to the visual recognition device, and the visual recognition device photographs the target object illuminated by the light source to obtain the light source illumination image information; the light source may be one or more, and the light source may be a single-spectrum light source and / or a multi-spectrum light source; The light source control unit controls the light emission of different spectral segments of a single multi-spectral light source, and controls the sequential and random light emission of multiple different single-spectrum light sources, obtains the light source illumination image result of the maximum number of different spectral illumination combinations, and after a period of operation, obtains multiple rounds of recognition results and corresponding light source working parameter data for all combinations of all light source illumination effects; The visual recognition device includes a camera, a camera gimbal, and an image recognition workstation configured with a recognition algorithm based on deep learning. The camera can be configured with one or more cameras according to the requirements of the recognition scene. The camera is set on the camera gimbal, and the camera is connected to the image recognition workstation via LAN / USB; The image recognition workstation controls and ensures full coverage of the image recognition effects of all light source spectrum combinations and lighting combinations through a loop control program.
2. The multi-spectral light source control device for visual recognition according to claim 1, Features: Image recognition workstations include dedicated image servers, desktop computers or high-performance industrial computers.
