Plastic product quality detection system and method based on artificial intelligence

Abstract

The invention discloses a plastic product quality detection system and method based on artificial intelligence, and relates to the technical field of plastic product quality detection.The method comprises the following steps that in the scanning process, the collection condition of point cloud data is monitored in real time, and the collection process is divided into normal collection and abnormal collection based on the collection condition of the point cloud data; and an early warning prompt is given out for abnormal acquisition. According to the method, the acquisition process can be divided into normal acquisition and abnormal acquisition in real time by monitoring the acquisition condition of the point cloud data, an early warning prompt is sent to inform related personnel to know the abnormal condition under the abnormal condition, and the abnormal condition possibly existing in the point cloud generation process can be found in time through real-time monitoring and intelligent analysis under the normal acquisition condition; the situation that the generated three-dimensional model is not matched with the actual plastic product in shape is effectively avoided, it can be guaranteed that the product quality meets the standard, defective products are prevented from flowing into the market, and therefore the product quality and safety are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of plastic product quality detection, and in particular to a plastic product quality detection system and method based on artificial intelligence. Background Art

[0002] Artificial intelligence-based plastic product quality inspection refers to the process of using artificial intelligence technologies, such as machine learning and computer vision, to inspect and evaluate the quality of plastic products. By scanning and analyzing the surface, structure, size and other characteristics of plastic products, the artificial intelligence system can identify possible defects, flaws or non-conformities. This method can improve the efficiency and accuracy of inspection, reduce the errors caused by human intervention, and realize the rapid inspection and screening of large quantities of products, thereby improving the quality and efficiency of plastic product production.

[0003] In AI-based plastic product quality inspection, the role of 3D scanners is to obtain three-dimensional models of plastic products. By scanning plastic products in three dimensions, accurate and comprehensive digital representations of their surfaces and structures can be obtained, including surface textures, shape details, etc. These three-dimensional models can provide basic data for subsequent quality inspection and analysis, allowing AI systems to more accurately identify defects, flaws or non-conformities, and to conduct more comprehensive comparisons and evaluations, improving the accuracy and efficiency of inspections.

[0004] When a 3D scanner is used to obtain a three-dimensional model of a plastic product, the 3D scanner releases or emits energy (such as light or laser) in some way. The emitted energy interacts with the surface of the plastic product and is reflected or scattered back to the 3D scanner. The 3D scanner records the returned energy data and accurately measures their relationship with the position and direction of the scanner. The recorded data is converted into a point cloud (a collection of a series of three-dimensional points) to represent the geometric shape of the surface of the plastic product. The density and position of these points are based on the energy data collected by the scanner to reflect the morphology and texture of the surface of the plastic product.

[0005] The prior art has the following deficiencies: When obtaining a three-dimensional model of a plastic product through a 3D scanner, if there are abnormalities in the point cloud generation process, the generated three-dimensional model may not match the actual shape of the plastic product, which may cause the quality inspection algorithm to be unable to accurately identify defects. When this happens, products with defects that have been missed may enter the market, thereby affecting product quality and safety.

[0006] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0007] The purpose of the present invention is to provide a plastic product quality detection system and method based on artificial intelligence. By monitoring the collection of point cloud data, the collection process can be divided into normal collection and abnormal collection in real time. In abnormal situations, an early warning prompt can be issued to notify relevant personnel. Under normal collection conditions, through real-time monitoring and intelligent analysis, abnormal situations that may exist in the point cloud generation process can be discovered in time, effectively avoiding the situation where the generated three-dimensional model does not match the actual shape of the plastic product. This helps to ensure that product quality meets standards and avoids defective products from entering the market, thereby improving product quality and safety to solve the problems in the above-mentioned background technology.

[0008] In order to achieve the above object, the present invention provides the following technical solution: a plastic product quality detection method based on artificial intelligence, comprising the following steps: During the scanning process, the point cloud data collection is monitored in real time, and the collection process is divided into normal collection and abnormal collection based on the point cloud data collection situation, and an early warning prompt is issued for abnormal collection; Under normal acquisition conditions, the point cloud is processed and optimized to improve the accuracy and completeness of the 3D model; Establish a real-time monitoring mechanism for the point cloud generation process and intelligently perceive abnormal situations in the point cloud generation process; After sensing abnormal hidden dangers, further judge the abnormalities and divide the abnormal hidden dangers into sudden abnormal hidden dangers and non-sudden abnormal hidden dangers. For sudden abnormal hidden dangers, continue to generate point clouds, and for non-sudden abnormal hidden dangers, issue early warning prompts.

[0009] Preferably, the collection process is divided into normal collection and abnormal collection based on the collection situation of point cloud data as follows: The process of point cloud data collection is divided into several fixed time windows, and the actual data collection speed at different time periods in each fixed time window is obtained; During the scanning process, the optimal data acquisition speed range for point cloud data acquisition is obtained; After comparing the actual data collection speeds obtained in different time periods with the optimal data collection speed range, the actual data collection speeds that are not within the optimal data collection speed range are screened out, and the abnormal collection speed index is calculated by summing the screened actual data collection speeds; The abnormal acquisition speed index is compared and analyzed with the pre-set abnormal acquisition speed index reference threshold. If the abnormal acquisition speed index is greater than or equal to the abnormal acquisition speed index reference threshold, the fixed time window is marked as abnormal acquisition; if the abnormal acquisition speed index is less than the abnormal acquisition speed index reference threshold, the fixed time window is marked as normal acquisition.

[0010] Preferably, during the point cloud generation process, the scanning interval information and exposure time information are obtained during the point cloud generation. The scanning interval refers to the distance between two adjacent scanning points during the scanning process, and the exposure time refers to the time that the scanning device is exposed to the light source during the scanning process. After processing the scanning interval information and the exposure time information, a scanning interval non-uniformity index and an exposure time variation index are generated.

[0011] Preferably, the logic of obtaining the scanning interval unevenness index is as follows: The point cloud topology is constructed using the Euclidean nearest neighbor search algorithm. , find the nearest neighbor point within its fixed time window; For each point , calculate the uneven value of its scanning interval , the specific calculation formula is: , where N is the point The number of nearest neighbors of Yes The average distance to its nearest neighbor, Yes Its nearest neighbor The distance between Through each point The scanning interval uneven value Calculate the scanning interval uneven index, the calculation expression is: , where represents the scanning interval unevenness index, and q represents the total number of scanning points.

[0012] Preferably, the specific steps of constructing the point cloud topology structure by using the Euclidean nearest neighbor search algorithm are: Step 1: Each point in the point cloud As a point to be queried; Step 2: Select the kd-tree data structure to store the point cloud, and organize the points in the point cloud according to the data structure to facilitate efficient nearest neighbor search; Step 3. For each point, perform the following steps: Choose a starting point ; Searching in a data structure with The nearest m points, where m is a pre-set parameter representing the number of nearest neighbors.

[0013] These nearest neighbor points are The nearest neighbor point is recorded; Step 4: For each point in the entire point cloud , repeat the nearest neighbor search process in step 3 to obtain the nearest neighbor point list for each point; Step 5: After completing the nearest neighbor search for all points, the point cloud topology structure is established, so that each point has its nearest neighbor point within a fixed time window.

[0014] Preferably, the logic for obtaining the exposure duration change index is as follows: Get the actual exposure time when performing several exposures in a fixed time window during the point cloud generation process, and calibrate the actual exposure time according to the time series as , y represents the number of actual exposure durations obtained in a time series under a fixed duration window during the point cloud generation process, y=1, 2, 3, 4, ..., Q, Q is a positive integer; The actual exposure time obtained Calculate the exposure duration change index. The calculation expression is: , where Represents the exposure duration change index.

[0015] Preferably, the scanning interval unevenness index generated in the fixed time window during the point cloud generation process is and exposure duration change index Establish an analysis model for comprehensive analysis and generate point cloud to generate anomaly assessment coefficients , generate anomaly assessment coefficients through point clouds Intelligent perception of abnormal situations in the point cloud generation process.

[0016] Preferably, the point cloud generation anomaly assessment coefficient generated in the point cloud generation process within a fixed time window is compared and analyzed with a pre-set point cloud generation anomaly assessment coefficient reference threshold; if the anomaly assessment coefficient is greater than or equal to the point cloud generation anomaly assessment coefficient reference threshold, a point cloud generation hidden danger signal is generated; if the anomaly assessment coefficient is less than the point cloud generation anomaly assessment coefficient reference threshold, a point cloud generation normal signal is generated.

[0017] Preferably, when a hidden danger signal is generated during the point cloud generation process, several subsequently generated point cloud generation anomaly assessment coefficients are obtained to establish an analysis set, and the point cloud generation anomaly assessment coefficient average value and the point cloud generation anomaly assessment coefficient standard deviation are calculated by analyzing the point cloud generation anomaly assessment coefficients in the set, and the point cloud generation anomaly assessment coefficient average value and the point cloud generation anomaly assessment coefficient standard deviation are compared and analyzed with the pre-set point cloud generation anomaly assessment coefficient reference threshold and standard deviation reference threshold, respectively, and the analysis results are as follows: If the average value of the point cloud generation anomaly assessment coefficient is less than the reference threshold value of the point cloud generation anomaly assessment coefficient and the standard deviation of the point cloud generation anomaly assessment coefficient is less than the reference threshold value of the standard deviation, the hidden danger is classified as a sudden abnormal hidden danger, and point cloud generation continues for the sudden abnormal hidden danger; If the average value of the point cloud generation anomaly assessment coefficient is less than the reference threshold of the point cloud generation anomaly assessment coefficient and the standard deviation of the point cloud generation anomaly assessment coefficient is less than the reference threshold of the standard deviation, the hidden danger is classified as a non-sudden abnormal hidden danger, and an early warning prompt is issued for the non-sudden abnormal hidden danger.

[0018] The plastic product quality inspection system based on artificial intelligence includes a real-time monitoring and abnormal warning module, a point cloud processing and optimization module, a real-time monitoring abnormal perception module, and an abnormal hidden danger judgment and processing module; The real-time monitoring and abnormal warning module monitors the acquisition of point cloud data in real time during the scanning process, divides the acquisition process into normal acquisition and abnormal acquisition based on the acquisition of point cloud data, and issues an early warning for abnormal acquisition; Point cloud processing and optimization module: Under normal acquisition conditions, point clouds are processed and optimized to improve the accuracy and completeness of the 3D model; Real-time monitoring of the abnormal perception module, establishes a real-time monitoring mechanism for the point cloud generation process, and intelligently perceives abnormal situations in the point cloud generation process; The abnormal hidden danger judgment and processing module, after sensing the abnormal hidden danger, further judges the abnormality and divides the abnormal hidden danger into sudden abnormal hidden danger and non-sudden abnormal hidden danger. For sudden abnormal hidden danger, it continues to generate point cloud, and for non-sudden abnormal hidden danger, it issues early warning prompts.

[0019] In the above technical solution, the technical effects and advantages provided by the present invention are: The present invention can monitor the collection of point cloud data and divide the collection process into normal collection and abnormal collection in real time. In abnormal situations, an early warning prompt can be issued to inform relevant personnel. In normal collection situations, through real-time monitoring and intelligent analysis, abnormal situations that may exist in the point cloud generation process can be discovered in time, effectively avoiding the situation where the generated three-dimensional model does not match the actual shape of the plastic product. This helps to ensure that product quality meets standards and avoid defective products from entering the market, thereby improving product quality and safety.

[0020] After discovering an abnormal situation, the present invention further judges and processes the abnormality, divides the abnormal hidden dangers into sudden abnormalities and non-sudden abnormalities, and issues early warning prompts of corresponding levels accordingly. For sudden abnormalities, point cloud generation continues, and for non-sudden abnormalities, early warning prompts are issued and relevant personnel are notified to avoid possible quality problems. This hierarchical processing can enable production management personnel to better respond to abnormal situations and take timely measures to ensure product quality and safety. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0022] Figure 1 The figure is a flow chart of the artificial intelligence-based plastic product quality inspection system and method of the present invention.

[0023] Figure 2 Schematic diagram of the modules of the artificial intelligence-based plastic product quality inspection system and method of the present invention. DETAILED DESCRIPTION

[0024] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0025] The present invention provides Figure 1 The artificial intelligence-based plastic product quality detection method shown includes the following steps: During the scanning process, the point cloud data collection is monitored in real time, and the collection process is divided into normal collection and abnormal collection based on the point cloud data collection situation, and an early warning prompt is issued for abnormal collection; Based on the collection of point cloud data, the collection process is divided into normal collection and abnormal collection as follows: The process of point cloud data collection is divided into several fixed time windows, and the actual data collection speed at different time periods in each fixed time window is obtained; It should be noted that 3D scanners are usually equipped with sensors that can monitor the movement speed and direction of the scanning head. These sensors can provide real-time scanning speed information, and the data acquisition speed can be obtained through the output data of the sensors.

[0026] During the scanning process, the optimal data acquisition speed range for point cloud data acquisition is obtained; It should be noted that consulting the technical specifications of the 3D scanner and the documentation provided by the manufacturer will usually include a recommended data acquisition speed range, and the manufacturer will usually provide optimal operating guidelines based on the performance and design characteristics of the equipment.

[0027] After comparing the actual data collection speeds obtained in different time periods with the optimal data collection speed range, the actual data collection speeds that are not within the optimal data collection speed range are screened out, and the abnormal collection speed index is calculated by summing the screened actual data collection speeds; The abnormal acquisition speed index is compared and analyzed with the pre-set abnormal acquisition speed index reference threshold. If the abnormal acquisition speed index is greater than or equal to the abnormal acquisition speed index reference threshold, the fixed time window is marked as abnormal acquisition; if the abnormal acquisition speed index is less than the abnormal acquisition speed index reference threshold, the fixed time window is marked as normal acquisition.

[0028] Under normal acquisition conditions, the point cloud is processed and optimized to improve the accuracy and completeness of the 3D model; Processing and optimizing point clouds to improve the accuracy and completeness of the 3D model are very important steps. These processes involve multiple steps, including: Noise removal: Remove noise points introduced by scanner errors, environmental interference or equipment failure to improve the quality and accuracy of point cloud data. By applying filters or mathematical algorithms, identify and eliminate points that are inconsistent or abnormal with surrounding points. Common noise removal methods include Gaussian filtering, median filtering, statistical filtering, etc.

[0029] Outlier removal: Eliminate outliers in the point cloud that do not conform to the true features of the plastic product surface to ensure that the generated model accurately reflects the actual shape. Use algorithms or rules to identify and remove outliers in the point cloud. These points may be introduced due to sensor failure, abnormal surface reflectivity, poor scanning conditions, etc.

[0030] Duplicate point removal: Eliminate overlapping or repeated points during scanning to reduce data volume and improve the stability and accuracy of the point cloud. Identify and merge or delete duplicate points in adjacent or similar locations in the point cloud. This can be achieved through spatial clustering algorithms or point cloud matching methods.

[0031] Point cloud alignment and registration: Multiple point cloud data sets collected from different angles or positions are fused into a complete, continuous model. Point clouds from different angles or positions are aligned and registered through techniques such as feature matching and iterative closest point (ICP) algorithm to eliminate discontinuities and gaps introduced by posture changes during scanning.

[0032] Surface fitting and smoothing: Surface fitting and smoothing are performed on point cloud data to generate smooth, continuous surfaces. The point cloud data is converted into smooth surfaces or polygonal meshes using mathematical models or curve fitting algorithms. At the same time, smoothing algorithms are used to remove sharp angles and irregularities in the data.

[0033] Edge and feature extraction: Identify and extract edges and features in point clouds to aid in subsequent analysis and processing. Apply edge detection algorithms and feature extraction techniques to identify areas with significant shape and structural features, such as holes, edges, corners, etc., from point clouds.

[0034] Establish a real-time monitoring mechanism for the point cloud generation process and intelligently perceive abnormal situations in the point cloud generation process; During the point cloud generation process, the scanning interval information and exposure time information are obtained. The scanning interval refers to the distance between two adjacent scanning points during the scanning process, and the exposure time refers to the time during which the scanning device (camera or sensor) is exposed to the light source during the scanning process. The scanning interval non-uniformity index and exposure time variation index are generated after the scanning interval information and exposure time information are processed. Uneven scanning intervals during point cloud generation may cause the generated 3D model to not match the actual shape of the plastic product. First, when the scanning interval is uneven, some areas may be scanned too densely, while other areas may be scanned too sparsely. In this case, densely scanned areas can capture more detailed information, while sparsely scanned areas may miss some subtle features or curvature changes. Therefore, the generated point cloud data may not fully reflect the actual shape of the plastic product surface, resulting in a mismatch between the 3D model and the actual shape. Second, when the scanning interval is uneven, the generated point cloud data may have overly dense point cloud sets in some areas and sparse point cloud sets in other areas. This uneven point cloud distribution may cause the generated 3D model to show discontinuity or angle discontinuity in some areas, thereby affecting the smoothness of the model surface and not matching the actual shape of the plastic product. Finally, uneven scanning intervals may cause the scanning points to be too densely distributed in some areas and too sparsely distributed in other areas. This may introduce some errors in the point cloud data, resulting in shape distortion or misalignment in some areas of the generated 3D model. Especially in areas with curved surfaces or complex geometries, uneven point cloud distribution may cause more severe shape distortion, making the model not match the actual shape.

[0035] The logic of obtaining the scanning interval uneven index is as follows: The point cloud topology is constructed using the Euclidean nearest neighbor search algorithm. , find the nearest neighbor point within its fixed time window; The specific steps of constructing the point cloud topology structure through the Euclidean nearest neighbor search algorithm are: Step 1: Each point in the point cloud As a point to be queried; Step 2: Select the kd-tree data structure to store the point cloud, and organize the points in the point cloud according to the data structure to facilitate efficient nearest neighbor search; Step 3. For each point, perform the following steps: Choose a starting point ; Searching in a data structure with The nearest m points, where m is a pre-set parameter representing the number of nearest neighbors.

[0036] These nearest neighbor points are The nearest neighbor point is recorded; Step 4: For each point in the entire point cloud , repeat the nearest neighbor search process in step 3 to obtain the nearest neighbor point list for each point; Step 5: After completing the nearest neighbor search for all points, the point cloud topology structure is established, so that each point has its nearest neighbor point within a fixed time window.

[0037] For each point , calculate the uneven value of its scanning interval , the specific calculation formula is: , where N is the point The number of nearest neighbors of Yes The average distance to its nearest neighbor, Yes Its nearest neighbor The distance between Through each point The scanning interval uneven value Calculate the scanning interval uneven index, the calculation expression is: , where represents the scanning interval unevenness index, and q represents the total number of scanning points.

[0038] From the calculation expression of the scanning interval unevenness index, it can be seen that the larger the performance value of the scanning interval unevenness index generated by the point cloud generation process within a fixed time window, the greater the hidden danger that the three-dimensional model generated by the point cloud does not match the actual shape of the plastic product, resulting in the quality inspection algorithm being unable to accurately identify defects; conversely, it indicates that the three-dimensional model generated by the point cloud does not match the actual shape of the plastic product, resulting in the smaller the hidden danger that the quality inspection algorithm is unable to accurately identify defects.

[0039] Large changes in exposure time during point cloud generation may cause the generated 3D model to not match the actual shape of the plastic product, thus affecting the accuracy of the quality inspection algorithm. First, large changes in exposure time may lead to inconsistent data quality collected in different parts of the point cloud. In areas where the exposure time is shorter, due to less light collected, the point cloud data may be sparse or some details may be missing. On the contrary, in areas where the exposure time is longer, too much light information may be collected, resulting in too dense point cloud data or even overexposure. This shape distortion will cause the generated 3D model to not match the actual shape of the plastic product, making it impossible for the quality inspection algorithm to accurately identify defects. Secondly, changes in exposure time may cause large brightness differences between the collected point cloud data. This brightness difference may affect the feature extraction and matching process of the point cloud data, making it difficult to align or fuse the point clouds collected under different exposure conditions. Due to the inconsistency of the point cloud data, the quality inspection algorithm may make misjudgments or miss detections and fail to accurately identify defects in plastic products. Finally, changes in exposure time will increase the complexity of data preprocessing and correction for the quality inspection algorithm. Additional correction and calibration work is required on the collected point cloud data to minimize the effects of exposure time variation. However, even with correction, the robustness of the quality inspection algorithm will still be affected, and large errors may be generated in abnormally exposed areas, making the inspection results unreliable.

[0040] The logic for obtaining the exposure duration change index is as follows: Get the actual exposure time when performing several exposures in a fixed time window during the point cloud generation process, and calibrate the actual exposure time according to the time series as , y represents the number of actual exposure durations obtained in a time series under a fixed duration window during the point cloud generation process, y=1, 2, 3, 4, ..., Q, Q is a positive integer; It should be noted that special exposure time measuring instruments, such as photocells, exposure meters, etc., are used to accurately measure the actual duration of each exposure. These instruments can detect the intensity and duration of light to determine the exposure duration.

[0041] The actual exposure time obtained Calculate the exposure duration change index. The calculation expression is: , where Represents the exposure duration change index.

[0042] From the calculation expression of the exposure time variation index, it can be seen that the larger the performance value of the exposure time variation index generated by the point cloud generation process within a fixed time window, the greater the risk that the three-dimensional model generated by the point cloud does not match the actual shape of the plastic product, resulting in the quality inspection algorithm being unable to accurately identify defects. Conversely, the smaller the risk that the three-dimensional model generated by the point cloud does not match the actual shape of the plastic product, resulting in the quality inspection algorithm being unable to accurately identify defects.

[0043] The scanning interval unevenness index generated in a fixed time window during the point cloud generation process and exposure duration change index Establish an analysis model for comprehensive analysis and generate point cloud to generate anomaly assessment coefficients , generate anomaly assessment coefficients through point clouds Intelligent perception of abnormal situations in the point cloud generation process.

[0044] The above analysis model is not specifically limited here, and can realize the scanning interval unevenness index and exposure duration change index Perform comprehensive analysis and generate point cloud to generate anomaly assessment coefficients In order to realize the technical solution of the present invention, the present invention provides a specific implementation method; Point cloud generation anomaly assessment coefficient The resulting calculation formula is: , where , The scanning interval non-uniformity index and exposure duration change index The preset scaling factor of , Both are greater than 0.

[0045] From the calculation formula, it can be seen that the larger the performance value of the scanning interval unevenness index generated in the fixed time window during the point cloud generation process, the larger the performance value of the exposure time variation index, that is, the point cloud generation process generated in the fixed time window generates an abnormal evaluation coefficient. The larger the performance value, the greater the risk that the three-dimensional model generated by the point cloud does not match the actual shape of the plastic product, resulting in the quality inspection algorithm being unable to accurately identify defects. Conversely, the smaller the risk that the three-dimensional model generated by the point cloud does not match the actual shape of the plastic product, resulting in the quality inspection algorithm being unable to accurately identify defects.

[0046] The point cloud generation anomaly assessment coefficient generated in the point cloud generation process within a fixed time window is compared and analyzed with the pre-set point cloud generation anomaly assessment coefficient reference threshold. If the anomaly assessment coefficient is greater than or equal to the point cloud generation anomaly assessment coefficient reference threshold, a point cloud generation hidden danger signal is generated, indicating that there are abnormal hidden dangers in the point cloud generation process within the fixed time window. If the anomaly assessment coefficient is less than the point cloud generation anomaly assessment coefficient reference threshold, a point cloud generation normal signal is generated, indicating that the three-dimensional model can be efficiently generated through cloud generation within the fixed time window.

[0047] After sensing abnormal hidden dangers, further judge the abnormalities and divide the abnormal hidden dangers into sudden abnormal hidden dangers and non-sudden abnormal hidden dangers. For sudden abnormal hidden dangers, continue to generate point clouds, and for non-sudden abnormal hidden dangers, issue early warning prompts; When a hidden danger signal is generated during the point cloud generation process, several subsequently generated point cloud generation anomaly assessment coefficients are obtained to establish an analysis set, and the point cloud generation anomaly assessment coefficient average value and the point cloud generation anomaly assessment coefficient standard deviation are calculated by analyzing the point cloud generation anomaly assessment coefficients in the set, and the point cloud generation anomaly assessment coefficient average value and the point cloud generation anomaly assessment coefficient standard deviation are compared and analyzed with the pre-set point cloud generation anomaly assessment coefficient reference threshold and standard deviation reference threshold, respectively. The analysis results are as follows: If the average value of the point cloud generation anomaly assessment coefficient is less than the reference threshold value of the point cloud generation anomaly assessment coefficient and the standard deviation of the point cloud generation anomaly assessment coefficient is less than the reference threshold value of the standard deviation, the hidden danger is classified as a sudden abnormal hidden danger, and point cloud generation continues for the sudden abnormal hidden danger; When the average value of the point cloud generation anomaly assessment coefficient is less than the reference threshold of the point cloud generation anomaly assessment coefficient and the standard deviation of the point cloud generation anomaly assessment coefficient is less than the standard deviation reference threshold, it indicates that the point cloud generation anomaly assessment coefficients in the analysis set are generally less than the reference threshold of the point cloud generation anomaly assessment coefficient. This indicates that the point cloud generation anomaly is an emergency and the impact is relatively small. There is no need to issue an early warning for this situation and the point cloud generation can continue.

[0048] If the point cloud generation anomaly assessment coefficient average value is less than the point cloud generation anomaly assessment coefficient reference threshold and the point cloud generation anomaly assessment coefficient standard deviation is less than the standard deviation reference threshold, the hidden danger is classified as a non-sudden abnormal hidden danger, and an early warning is issued for the non-sudden abnormal hidden danger; When the point cloud generation anomaly assessment coefficient average value is less than the point cloud generation anomaly assessment coefficient reference threshold and the point cloud generation anomaly assessment coefficient standard deviation is less than the standard deviation reference threshold, it indicates that the point cloud generation anomaly assessment coefficient in the analysis set is not generally less than the point cloud generation anomaly assessment coefficient reference threshold. This indicates that the point cloud generation anomaly is not a sudden situation, and this impact is relatively large. For non-sudden abnormal hidden dangers, an early warning prompt is issued to inform relevant personnel that if this state continues, the three-dimensional model generated by the point cloud may not match the actual shape of the plastic product, resulting in the quality inspection algorithm being unable to accurately identify defects.

[0049] The present invention can divide the collection process into normal collection and abnormal collection in real time by monitoring the collection of point cloud data, and issue an early warning prompt to notify relevant personnel under abnormal circumstances. Under normal collection conditions, through real-time monitoring and intelligent analysis, it can promptly discover possible abnormalities in the point cloud generation process, effectively avoiding the situation where the generated three-dimensional model does not match the actual shape of the plastic product, which helps to ensure that product quality meets standards and avoids defective products from entering the market, thereby improving product quality and safety.

[0050] After discovering an abnormal situation, the present invention further judges and processes the abnormality, divides the abnormal hidden dangers into sudden abnormalities and non-sudden abnormalities, and issues early warning prompts of corresponding levels accordingly. For sudden abnormalities, point cloud generation continues, and for non-sudden abnormalities, early warning prompts are issued and relevant personnel are notified to avoid possible quality problems. This hierarchical processing can enable production management personnel to better respond to abnormal situations and take timely measures to ensure product quality and safety.

[0051] The present invention provides Figure 2 The artificial intelligence-based plastic product quality inspection system shown includes a real-time monitoring and abnormal warning module, a point cloud processing and optimization module, a real-time monitoring abnormal perception module, and an abnormal hidden danger judgment and processing module; The real-time monitoring and abnormal warning module monitors the acquisition of point cloud data in real time during the scanning process, divides the acquisition process into normal acquisition and abnormal acquisition based on the acquisition of point cloud data, and issues an early warning for abnormal acquisition; Point cloud processing and optimization module: Under normal acquisition conditions, point clouds are processed and optimized to improve the accuracy and completeness of the 3D model; Real-time monitoring of the abnormal perception module, establishes a real-time monitoring mechanism for the point cloud generation process, and intelligently perceives abnormal situations in the point cloud generation process; The abnormal hidden danger judgment and processing module, after sensing the abnormal hidden danger, further judges the abnormality and divides the abnormal hidden danger into sudden abnormal hidden danger and non-sudden abnormal hidden danger. For sudden abnormal hidden danger, point cloud generation continues, and for non-sudden abnormal hidden danger, early warning prompts are issued; The artificial intelligence-based plastic product quality detection method provided in the embodiment of the present invention is realized by the above-mentioned artificial intelligence-based plastic product quality detection system. The specific methods and processes of the artificial intelligence-based plastic product quality detection system are detailed in the embodiment of the above-mentioned artificial intelligence-based plastic product quality detection method, which will not be repeated here.

[0052] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0053] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired or wireless (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0054] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0055] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0056] In the several embodiments provided in the present application, it should be understood that the disclosed systems and methods can be implemented in other ways. For example, the embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0057] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0058] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0059] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A plastic product quality detection method based on artificial intelligence, characterized in that: The following steps are involved: During the scanning process, the point cloud data collection is monitored in real time, and the collection process is divided into normal collection and abnormal collection based on the point cloud data collection situation, and an early warning prompt is issued for abnormal collection; Under normal acquisition conditions, the point cloud is processed and optimized to improve the accuracy and completeness of the 3D model; Establish a real-time monitoring mechanism for the point cloud generation process and intelligently perceive abnormal situations in the point cloud generation process; After sensing abnormal hidden dangers, further judge the abnormalities and divide the abnormal hidden dangers into sudden abnormal hidden dangers and non-sudden abnormal hidden dangers. For sudden abnormal hidden dangers, continue to generate point clouds, and for non-sudden abnormal hidden dangers, issue early warning prompts.

2. The artificial intelligence-based plastic product quality detection method according to claim 1, characterized in that: Based on the collection of point cloud data, the collection process is divided into normal collection and abnormal collection as follows: The process of point cloud data collection is divided into several fixed time windows, and the actual data collection speed at different time periods in each fixed time window is obtained; During the scanning process, the optimal data acquisition speed range for point cloud data acquisition is obtained; After comparing the actual data collection speeds obtained in different time periods with the optimal data collection speed range, the actual data collection speeds that are not within the optimal data collection speed range are screened out, and the abnormal collection speed index is calculated by summing the screened actual data collection speeds; The abnormal acquisition speed index is compared and analyzed with the pre-set abnormal acquisition speed index reference threshold. If the abnormal acquisition speed index is greater than or equal to the abnormal acquisition speed index reference threshold, the fixed time window is marked as abnormal acquisition; if the abnormal acquisition speed index is less than the abnormal acquisition speed index reference threshold, the fixed time window is marked as normal acquisition.

3. The artificial intelligence-based plastic product quality detection method according to claim 1, characterized in that: During the point cloud generation process, the scanning interval information and exposure time information are obtained. The scanning interval refers to the distance between two adjacent scanning points during the scanning process, and the exposure time refers to the time the scanning device is exposed to the light source during the scanning process. After processing the scanning interval information and exposure time information, the scanning interval unevenness index and exposure time variation index are generated.

4. The artificial intelligence-based plastic product quality detection method according to claim 3, characterized in that: The logic of obtaining the scanning interval uneven index is as follows: The point cloud topology is constructed using the Euclidean nearest neighbor search algorithm. , find the nearest neighbor point within its fixed time window; For each point , calculate the uneven value of its scanning interval , the specific calculation formula is: , where N is the point The number of nearest neighbors of Yes The average distance to its nearest neighbor, Yes Its nearest neighbor The distance between Through each point The scanning interval uneven value Calculate the scanning interval uneven index, the calculation expression is: , where represents the scanning interval unevenness index, and q represents the total number of scanning points.

5. The artificial intelligence-based plastic product quality detection method according to claim 4, characterized in that: The specific steps of constructing the point cloud topology structure through the Euclidean nearest neighbor search algorithm are: Step 1: Each point in the point cloud As a point to be queried; Step 2: Select the kd-tree data structure to store the point cloud, and organize the points in the point cloud according to the data structure to facilitate efficient nearest neighbor search; Step 3. For each point, perform the following steps: Choose a starting point ; Searching in a data structure with The nearest m points, where m is a pre-set parameter representing the number of nearest neighbors; These nearest neighbor points are The nearest neighbor point is recorded; Step 4: For each point in the entire point cloud , repeat the nearest neighbor search process in step 3 to obtain the nearest neighbor point list for each point; Step 5: After completing the nearest neighbor search for all points, the point cloud topology structure is established, so that each point has its nearest neighbor point within a fixed time window.

6. The artificial intelligence-based plastic product quality detection method according to claim 5, characterized in that: The logic for obtaining the exposure duration change index is as follows: Get the actual exposure time when performing several exposures in a fixed time window during the point cloud generation process, and calibrate the actual exposure time according to the time series as , y represents the number of actual exposure durations obtained in a time series under a fixed duration window during the point cloud generation process, y=1, 2, 3, 4, ..., Q, Q is a positive integer; The actual exposure time obtained Calculate the exposure duration change index. The calculation expression is: , where Represents the exposure duration change index.

7. The artificial intelligence-based plastic product quality detection method according to claim 6, characterized in that: The scanning interval unevenness index generated in a fixed time window during the point cloud generation process and exposure duration change index Establish an analysis model for comprehensive analysis and generate point cloud to generate anomaly assessment coefficients , generate anomaly assessment coefficients through point clouds Intelligent perception of abnormal situations in the point cloud generation process.

8. The artificial intelligence-based plastic product quality detection method according to claim 7, characterized in that: The point cloud generation anomaly assessment coefficient generated by the point cloud generation process within a fixed time window is compared and analyzed with the pre-set point cloud generation anomaly assessment coefficient reference threshold. If the anomaly assessment coefficient is greater than or equal to the point cloud generation anomaly assessment coefficient reference threshold, a point cloud generation hidden danger signal is generated; if the anomaly assessment coefficient is less than the point cloud generation anomaly assessment coefficient reference threshold, a point cloud generation normal signal is generated.

9. The artificial intelligence-based plastic product quality detection method according to claim 7, characterized in that: When a hidden danger signal is generated during the point cloud generation process, several subsequently generated point cloud generation anomaly assessment coefficients are obtained to establish an analysis set, and the point cloud generation anomaly assessment coefficient average value and the point cloud generation anomaly assessment coefficient standard deviation are calculated by analyzing the point cloud generation anomaly assessment coefficients in the set, and the point cloud generation anomaly assessment coefficient average value and the point cloud generation anomaly assessment coefficient standard deviation are compared and analyzed with the pre-set point cloud generation anomaly assessment coefficient reference threshold and standard deviation reference threshold, respectively. The analysis results are as follows: If the average value of the point cloud generation anomaly assessment coefficient is less than the reference threshold of the point cloud generation anomaly assessment coefficient and the standard deviation of the point cloud generation anomaly assessment coefficient is less than the reference threshold of the standard deviation, the hidden danger is classified as a sudden abnormal hidden danger, and point cloud generation continues for the sudden abnormal hidden danger; If the average value of the point cloud generation anomaly assessment coefficient is less than the reference threshold of the point cloud generation anomaly assessment coefficient and the standard deviation of the point cloud generation anomaly assessment coefficient is less than the standard deviation reference threshold, the hidden danger is classified as a non-sudden abnormal hidden danger, and an early warning prompt is issued for the non-sudden abnormal hidden danger.

10. A plastic product quality detection system based on artificial intelligence, used to implement the plastic product quality detection method based on artificial intelligence as described in any one of claims 1 to 9, characterized in that: It includes real-time monitoring and abnormal warning module, point cloud processing and optimization module, real-time monitoring abnormal perception module and abnormal hidden danger judgment and processing module; The real-time monitoring and abnormal warning module monitors the acquisition of point cloud data in real time during the scanning process, divides the acquisition process into normal acquisition and abnormal acquisition based on the acquisition of point cloud data, and issues an early warning for abnormal acquisition; Point cloud processing and optimization module: Under normal acquisition conditions, point clouds are processed and optimized to improve the accuracy and completeness of the 3D model; Real-time monitoring of the abnormal perception module, establishes a real-time monitoring mechanism for the point cloud generation process, and intelligently perceives abnormal situations in the point cloud generation process; The abnormal hidden danger judgment and processing module, after sensing the abnormal hidden danger, further judges the abnormality and divides the abnormal hidden danger into sudden abnormal hidden danger and non-sudden abnormal hidden danger. For sudden abnormal hidden danger, point cloud generation continues, and for non-sudden abnormal hidden danger, early warning prompts are issued.

Images