Object Detection Method Based on Computer Vision
Through the object detection method of computer vision, OCR technology is used to collect and analyze image data, and an in-house and out-of-house scenario analysis model is established, which solves the problem that the device status cannot be updated in time in the intelligent equipment management system, and realizes the light, warehouse layout and QR code Real-time monitoring and adjustment of quality improves the safety and efficiency of in-store operation.
Patent Information
- Application Number
- CN202411171158.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-26
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2044-08-26
AI Technical Summary
In the intelligent device management system, the problem is that the device status cannot be changed in time due to network or environmental factors during the QR code standard box management process.
Using object detection method based on computer vision, OCR technology collects and recognizes scene and standard box image data, establishes a scenario analysis model for inlet and exit database, generates prediction evaluation coefficients for inlet and exit databases, classifies scene categories and implements corresponding control methods, and dynamically adjusts. The plan is to deal with environmental factors such as light accumulation, warehouse layout and QR code quality.
It improves the safety and efficiency of operation, reduces the risk of delayed update of equipment status information, and ensures the smooth and accurate process of in-store access.
Smart Images

Figure CN118692094B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of computer vision, and more specifically, to a target detection method based on computer vision. Background Art
[0002] With the rapid development of computer vision technology, object detection has become increasingly important in the fields of image processing and artificial intelligence. Object detection refers to the process of identifying and locating specific target objects in images or videos. It is a core component of many computer vision applications, including intelligent monitoring, autonomous driving, and medical image analysis. Traditional object detection methods usually rely on manually designed features and classifiers, and their performance is limited by the expressiveness of features and the generalization ability of classifiers. With the rise of deep learning technology, object detection methods based on neural networks have made great breakthroughs.
[0003] Computer vision is an interdisciplinary field that involves the interaction between computer systems and advanced image recognition and understanding. Its goal is to enable computer systems to mimic the functions of the human visual system, that is, to obtain information from images or videos and to understand and analyze them. Computer vision generally involves knowledge in areas such as image processing, pattern recognition, machine learning, and artificial intelligence.
[0004] Object detection is an important task in the field of computer vision. Its goal is to accurately locate and identify target objects of a specific category in images or videos. Unlike simple image classification, object detection not only needs to identify the target in the image, but also needs to accurately locate the position of the target in the image, usually represented by a rectangular bounding box.
[0005] For example, the invention patent with publication number: CN105069466B discloses a method for identifying the color of pedestrian clothing based on digital image processing. Pedestrian images are collected by adopting the pedestrian detection method combining the HOG feature description operator with the SVM classifier; the Sobel operator is used to detect the edge contour shape of the pedestrian to obtain the image to be searched; a pedestrian contour shape template is made, and the pedestrian contour shape template is matched with the corresponding area in the image to be searched to obtain the upper and lower body images of the pedestrian; the seed filling method is used to mark the connected areas of the clothing colors of the upper and lower body areas of the pedestrian respectively; color features are extracted from the color connected areas; the SVM classifier is used to perform color classification and discrimination to obtain the color of the pedestrian's clothing, and the final result is output, which improves the recognition accuracy of the color of the pedestrian's clothing, brings great guarantees for the implementation of standardized clothing in dangerous areas, and eliminates safety hazards.
[0006] For example, the invention patent with publication number: CN117057377B discloses a code recognition and matching method, which obtains a grayscale code image by collecting the code image, filtering, stripping the background, position correction and grayscale processing, and extracting the image features on the grayscale code image. When extracting the image features, the grayscale code image is divided into blocks, and two features are extracted from each image block: pixel gray value distribution features and pixel gradient features. Through this block processing and feature extraction on each block, the features on each image block are accurately captured, so that the accuracy of recognition and matching is improved when feature matching is performed later.
[0007] The prior art has the following problems: In the daily use of the intelligent equipment management system, it is found that when the standard boxes are put in and out of the warehouse as a whole, the status of some equipment cannot be changed in time due to the influence of network or other environmental factors. In view of the above problems, the present invention proposes a solution. Summary of the invention
[0008] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a target detection method based on computer vision, which solves the problem that the status cannot be changed in time during the management of two-dimensional code standard boxes through computer vision.
[0009] To achieve the above object, the present invention provides the following technical solutions:
[0010] 1. A target detection method based on computer vision, comprising the following steps: when a standard box enters and leaves the warehouse as a whole, scene image data is collected and identified through OCR, and standard box image data is collected during the QR code scanning stage; an entry and exit scene analysis model is established based on the scene image data and the standard box image data, and an entry and exit prediction evaluation coefficient is generated; the entry and exit scenes are classified into scene categories, the scene categories include normal scenes, low interference and excessive interference scenes, the expected scene categories are classified according to the entry and exit prediction evaluation coefficients, and corresponding control measures are implemented; when the scene category is insufficient signal, the control measures are tracked, the control effect is evaluated, and a feedback signal is issued.
[0011] In a preferred embodiment, the scene image data includes a light gathering influence coefficient and a warehouse layout influence coefficient; the standard box image data includes a QR code guided evaluation coefficient.
[0012] In a preferred embodiment, the method for obtaining the light gathering influence coefficient is as follows: obtain the light structure, light source distribution, and light gathering of the warehouse, and draw a three-dimensional image through a computer; filter the three-dimensional image; extract features from the processed three-dimensional image, and calculate the light intensity evaluation coefficient based on the conservation of mass and momentum combined with the diffusion coefficient and boundary conditions; simulate the light path conditions inside the warehouse through images, and establish a simulation equation to obtain the light divergence effect coefficient; and calculate the light gathering influence coefficient in combination with the light intensity evaluation coefficient.
[0013] In a preferred embodiment, the method for obtaining the warehouse layout influence coefficient is as follows: voxelize the three-dimensional model of the warehouse, and divide the warehouse into multiple cubes, label and obtain a voxel set; count the number of voxels occupied by obstacles to obtain the obstacle volume; identify and analyze the distribution of obstacle paths in the voxelized warehouse model, and calculate the curvature of the obstacle path; combine the curvature of the obstacle path and the obstacle volume analysis to calculate the warehouse layout influence coefficient.
[0014] In a preferred embodiment, the method for obtaining the QR code guidance evaluation coefficient is as follows: obtaining the guidance information entropy of the QR code by solving the QR code recognition model of the finite element model; calculating the maximum main information entropy value identified; and calculating the QR code guidance evaluation coefficient by combining the maximum main information entropy value identified and the network signal interference degree.
[0015] In a preferred embodiment, the predicted scene categories are classified according to the entry and exit prediction evaluation coefficients, and corresponding management and control measures are implemented, specifically: the entry and exit prediction evaluation coefficients are compared with the preset entry and exit hidden danger thresholds and entry and exit failure thresholds; the three corresponding results after the comparison are divided into normal scenes, low-interference scenes, and excessive interference scenes; for low-interference scenes, the light and signal are adjusted; for scenes with excessive interference, it is necessary to add signal amplifiers and re-plan the path.
[0016] In a preferred embodiment, the control measures are tracked, the control effects are evaluated, and feedback signals are issued, specifically: multiple inbound and outbound forecast evaluation coefficients within the G time during the control measures are obtained to construct a control set; the standard deviation and average value of the inbound and outbound forecast evaluation coefficients in the control set are calculated, and a control standard deviation reference threshold and a control average value reference threshold are set; the standard deviation of the inbound and outbound forecast evaluation coefficients in the control set is compared with the control standard deviation reference threshold, and the average value of the calculated inbound and outbound forecast evaluation coefficients in the control set is compared with the control average value reference threshold; the control effect is divided into three effects: control success, control failure, and control fluctuation according to the comparison results; and feedback signals are issued to relevant personnel after classification.
[0017] The technical effects and advantages of the target detection method based on computer vision of the present invention are as follows:
[0018] 1. The present invention uses a computer vision target detection method to collect and identify scene and standard box image data using OCR technology during the overall storage and warehousing process of standard boxes, thereby achieving accurate prediction and evaluation. This method can timely detect and respond to potential hidden dangers caused by environmental factors such as light concentration, warehouse layout, and QR code quality, thereby improving the safety and efficiency of operations. By dynamically adjusting the plan and optimizing control measures, the system effectively reduces the risk of delayed updates of equipment status information, ensuring the smoothness and accuracy of the storage and warehousing process.
[0019] 2. The present invention establishes an in-and-out scene analysis model by integrating scene image data and standard box image data, and generates an in-and-out prediction evaluation coefficient to accurately manage environmental variables in in-and-out operations; first, by integrating the light aggregation influence coefficient, warehouse layout influence coefficient and QR code guidance evaluation coefficient, an in-and-out prediction evaluation coefficient is formed, and the model can fully reflect the environmental influencing factors in the in-and-out operation; this method can classify scenes into normal scenes, low-interference scenes and scenes with excessive interference, so as to implement targeted management and control measures according to different environmental conditions. For normal scenes, the operating process can maintain the standard, and for low-interference scenes, the system will make appropriate adjustments, such as optimizing light and signals; for scenes with excessive interference, more complex adjustments may be required, such as adding signal amplifiers or replanning paths. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 It is a structural schematic diagram of the target detection method based on computer vision of the present invention. DETAILED DESCRIPTION
[0021] 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.
[0022] Embodiment 1, Figure 1 The present invention provides a target detection method based on computer vision, which comprises the following steps:
[0023] S1, when the standard box is put in and out of the warehouse as a whole, the scene image data is collected and recognized through OCR, and the standard box image data is collected during the QR code scanning stage.
[0024] In intelligent equipment management, the overall entry and exit of standard containers is a key link, which directly affects the safety and management efficiency of equipment; however, due to network instability or interference from environmental factors, the status of some equipment may not be updated in time, which leads to inaccurate information in the management system and even potential safety risks; by using OCR technology to collect scene image data and standard container image data during the overall entry and exit of standard containers, a comprehensive analysis model can be established to accurately predict and evaluate the entry and exit of standard containers; this process can help managers take targeted control measures in equipment entry and exit operations to ensure the safety and effectiveness of operations. By tracking and providing timely feedback on the control effect, the system can dynamically adjust the plan to minimize the risks that may arise in equipment management.
[0025] The management of the entry and exit of standard containers is a complex operation in the intelligent equipment management system. If it is not effectively monitored and predicted, it may lead to delayed updates of equipment status information, thus affecting the accuracy of overall management. In order to deal with these potential problems, OCR technology can be introduced in the process of standard container entry and exit to collect and identify the image data of the entry and exit scenes and the image data of the standard containers, and use these data to establish an entry and exit scene analysis model. This model can generate an entry and exit prediction evaluation coefficient, and classify and control the entry and exit scenes, thereby improving the safety and controllability of operations. Especially in the case of poor signal or large environmental interference, the system can track the effectiveness of control measures, provide timely feedback and adjustments, ensure the smooth progress of entry and exit operations, and reduce possible safety hazards.
[0026] During the overall entry and exit process of standard containers, the OCR technology is used to collect and identify scene (environment) image data, which can monitor and analyze the actual environmental conditions at the entry and exit site in real time. This can effectively capture key factors such as the signal conditions, light conditions, and spatial layout of the site, ensuring that the system can accurately judge and adapt to different environmental changes. By analyzing these image data, the system can predict possible abnormal situations and provide corresponding management and control measures, thereby improving the safety and reliability of entry and exit operations.
[0027] During the QR code scanning phase, by collecting the image data of the standard box, the accurate identification and recording of the standard box and its contents can be ensured; the image data can not only verify the integrity and accuracy of the QR code information, but also capture the appearance, status and potential damage of the standard box. By analyzing these image data, the system can detect possible abnormalities, such as QR code damage, label errors or box damage, and issue warnings in time to prevent erroneous information from entering the system; this can improve the accuracy and reliability of storage operations and ensure the effectiveness of equipment management.
[0028] The scene image data includes a light gathering influence coefficient and a warehouse layout influence coefficient;
[0029] The standard box image data includes a two-dimensional code to guide the evaluation coefficient.
[0030] The light concentration coefficient plays an important role in the process of standard box in and out of the warehouse, because it directly affects the recognition accuracy of the QR code and scene image. If the light is too concentrated or the reflection is strong, it may cause overexposure, glare or reflection problems in the image captured by the camera equipment, which will affect the scanning effect of the QR code and the accuracy of the image data. On the contrary, insufficient or uneven light distribution may also cause the image to be too dark or blurred, affecting the system's correct judgment of the status of the standard box. Therefore, the light concentration coefficient is a key environmental factor that affects the data collection and equipment recognition accuracy during the process of in and out of the warehouse, and must be controlled and optimized to ensure the smooth operation.
[0031] The light gathering influence coefficient has the following effects and functions on predicting and identifying hidden dangers of standard boxes entering and leaving the warehouse:
[0032] Image recognition accuracy: Excessive or insufficient light concentration will affect the image acquisition quality of the camera equipment, resulting in blurred, overexposed or dim images, thereby reducing the OCR system's recognition accuracy of QR codes and scenes. Accurate image recognition is the basis for predicting and discovering hidden dangers, so the impact of light concentration is directly related to the effectiveness of hidden danger identification.
[0033] Abnormal situation detection: By analyzing the light concentration influence coefficient, recognition errors caused by light problems can be predicted in advance; for example, if the light is too strong or uneven, the system may issue an alarm to prompt the operator to adjust the light source or rescan, thereby avoiding incorrect device status updates caused by light problems;
[0034] Prevent safety hazards: The light concentration influence coefficient plays an early warning role in the identification process, ensuring timely remedial measures under adverse light conditions to avoid operational errors. By real-time monitoring and adjusting light conditions, the hidden dangers of inbound and outbound operations caused by identification errors can be greatly reduced, improving the safety of the overall process.
[0035] Data collection optimization: Monitoring and adjusting the light gathering coefficient helps optimize the image data collection quality and ensures that the system can work stably under different light conditions. This optimization not only improves the efficiency of warehousing operations, but also can timely identify and predict potential risks through more accurate data support.
[0036] The method for obtaining the light gathering influence coefficient is as follows:
[0037] Obtain the light structure, light source distribution, and light gathering of the warehouse, and draw a three-dimensional image through a computer; filter the image; extract features from the processed image, and calculate the light intensity evaluation coefficient based on mass conservation and momentum conservation combined with diffusion coefficient and boundary conditions;
[0038] The light path inside the warehouse is simulated through images, and a simulation equation is established to obtain the light divergence effect coefficient; the light concentration influence coefficient is calculated by combining the light intensity evaluation coefficient;
[0039] The specific calculation formula of the light intensity evaluation coefficient is as follows:
[0040]
[0041] The specific calculation formula of the light divergence effect coefficient is as follows:
[0042]
[0043] The specific calculation formula of the light gathering influence coefficient is as follows:
[0044]
[0045] In the formula, is the light intensity evaluation coefficient, is the light divergence effect coefficient, is the light gathering influence coefficient, is the error term, is the volume of the warehouse, is the light source volume of the warehouse, is the air density, is the diffusion coefficient of light in air, is the Laplace operator of light intensity, is the light energy scattering rate, is the average light energy, is the nonlinear term of light energy divergence.
[0046] From the calculation formula of the light gathering influence coefficient, it can be seen that when the performance value of the light gathering influence coefficient is larger, it indicates that the hidden danger of the standard container being unable to change the status in time when entering and leaving the warehouse is more likely to occur; when the performance value of the light gathering influence coefficient is smaller, it indicates that the hidden danger of the standard container being unable to change the status in time when entering and leaving the warehouse is less likely to occur.
[0047] The warehouse layout influence factor affects the path planning, equipment movement and positioning accuracy of the standard container in the warehouse, which may cause collisions, delays or signal interference during operation; complex or narrow warehouse layouts increase the difficulty of equipment movement and the risk of inbound and outbound operations, thereby affecting information transmission and equipment status updates. By monitoring the warehouse layout influence factor, the system can predict these potential risks and optimize path planning and signal transmission in advance to ensure the smooth and safe inbound and outbound processes.
[0048] The warehouse layout influence coefficient has the following effects and functions on predicting and identifying hidden dangers of standard container in and out of the warehouse:
[0049] Path planning and equipment movement: The warehouse layout influence coefficient is directly related to the path planning and equipment movement of standard containers during the in-and-out process. If the warehouse layout is complex or the space is narrow, it may make it difficult for the equipment to move smoothly, increasing the risk of collision, jamming or delay. By analyzing the warehouse layout influence coefficient, the system can predict potential path obstacles and adjust the path planning in advance to avoid hidden dangers caused by warehouse layout problems.
[0050] Signal reception and transmission: Warehouse layout affects the propagation of wireless signals, especially when there are a lot of metal shelves or confined spaces, where signals may be shielded or interfered with. The warehouse layout influence factor helps identify signal weaknesses, so the system can provide early warning and take measures (such as adding signal amplifiers) to ensure that equipment status information can be smoothly transmitted and updated during the storage process;
[0051] Equipment positioning and management: Complex or irregular warehouse layouts may result in reduced positioning accuracy of standard containers and equipment, especially when using automatic navigation or positioning systems; by monitoring the warehouse layout influence coefficient, the system can optimize the positioning algorithm to ensure accurate identification and management of standard containers in the warehouse, reducing misoperation or inbound and outbound delays caused by inaccurate positioning;
[0052] Contingency plans and safety assurance: The warehouse layout impact factor can help identify potential safety hazards in warehousing and outbound operations, such as the accessibility of emergency exits or the reliability of fire-fighting equipment. Based on this factor, the system can predict possible safety issues that may arise during the warehousing and outbound process, and formulate emergency plans in advance to ensure the safety and reliability of warehousing and outbound operations.
[0053] The method for obtaining the warehouse layout influence coefficient is as follows:
[0054] The three-dimensional model of the warehouse is voxelized, and the warehouse is divided into multiple cubes, which are labeled and obtained as a voxel set; the number of voxels occupied by obstacles is counted to obtain the obstacle volume; the distribution of obstacle paths is identified and analyzed in the voxelized warehouse model, and the curvature of the obstacle path is calculated; the warehouse layout influence coefficient is calculated by combining the curvature of the obstacle path and the obstacle volume analysis;
[0055] The calculation formula of the obstacle volume is:
[0056]
[0057] The calculation formula of the curvature of the obstacle path is:
[0058]
[0059] The calculation formula of the warehouse layout influence coefficient is:
[0060]
[0061] In the formula, is the warehouse layout influence coefficient, is the total number of voxels in the voxel set, is the voxel label; is the volume of the ith voxel, is the blocking length of the i-th obstacle path, and e is a natural number.
[0062] From the calculation formula of the warehouse layout influence coefficient, it can be seen that when the performance value of the warehouse layout influence coefficient is larger, it indicates that the hidden danger of the standard container entering and leaving the warehouse is more likely to occur that the status cannot be changed in time; when the performance value of the warehouse layout influence coefficient is smaller, it indicates that the hidden danger of the standard container entering and leaving the warehouse is less likely to occur that the status cannot be changed in time.
[0063] The impact of the QR code guidance evaluation coefficient on the inbound and outbound process lies in its ability to accurately evaluate the readability and information integrity of the QR code; a high evaluation coefficient indicates that the QR code has good scanning quality and accurate data reading, which helps the system to smoothly record the inbound and outbound information of the standard box; while a low evaluation coefficient may indicate that the QR code is blurred, damaged, or has poor lighting, resulting in data recognition errors or delayed updates. By real-time monitoring of the QR code guidance evaluation coefficient, the system can promptly detect and correct scanning problems, thereby improving the accuracy and efficiency of inbound and outbound operations and reducing management risks caused by QR code problems.
[0064] The QR code guided evaluation coefficient has the following effects and functions on predicting and identifying hidden dangers of standard container in and out of the warehouse:
[0065] Scanning accuracy: The QR code guide evaluation coefficient helps evaluate the clarity and scanning quality of the QR code. A high evaluation coefficient means that the QR code is easy to read and can accurately identify the standard box information, thereby reducing the problem of inaccurate inbound and outbound records caused by scanning errors; a low evaluation coefficient indicates that the QR code may be blurred or damaged, increasing the risk of data reading errors;
[0066] Data integrity: The evaluation coefficient reflects whether the data stored in the QR code is complete and valid; a good evaluation coefficient ensures that the data in the QR code is correctly decoded, which helps the system to accurately record the status and location of the standard box, thereby preventing information loss or incorrect update;
[0067] Operational efficiency: By monitoring the QR code guidance evaluation coefficient, the system can identify problems that may affect scanning efficiency in advance and take measures to optimize the readability of the QR code; this helps to reduce operational delays caused by QR code problems and improve the overall efficiency of the warehousing process;
[0068] Hidden danger warning: A low QR code guidance evaluation coefficient can serve as a warning signal for potential hidden dangers, prompting operators to check and repair QR code problems or adjust the scanning environment; timely remedial measures can avoid potential errors in the warehousing process and ensure the accuracy and smooth progress of operations.
[0069] The method for obtaining the QR code guided evaluation coefficient is as follows:
[0070] The guidance information entropy of the QR code is obtained by solving the QR code recognition model of the finite element model; the maximum main information entropy value recognized is calculated; and the guidance evaluation coefficient of the QR code is calculated by combining the maximum main information entropy value recognized and the network signal interference degree;
[0071] The calculation formula of the guided information entropy of the two-dimensional code is:
[0072]
[0073] The calculation formula of the maximum identified primary information entropy value is:
[0074]
[0075] The calculation formula of the QR code guidance evaluation coefficient is as follows:
[0076]
[0077] In the formula, is the guided information entropy of the QR code, is the maximum primary information entropy in the QR code, is the intermediate main information entropy in the QR code, is the adjustment item; is the maximum identified primary information entropy value; The area of the QR code recognized by OCR; is the QR code guided evaluation coefficient, It is the network signal interference degree.
[0078] It can be seen from the calculation formula of the QR code guidance evaluation coefficient that the larger the performance value of the QR code guidance evaluation coefficient, the more likely it is that the standard containers entering and leaving the warehouse will have the hidden danger of not being able to change the status in time; when the performance value of the QR code guidance evaluation coefficient is smaller, it means that the standard containers entering and leaving the warehouse are more likely to have the hidden danger of not being able to change the status in time.
[0079] This embodiment uses computer vision target detection methods to collect and identify scene and standard box image data during the overall storage and retrieval process of standard boxes using OCR technology, thereby achieving accurate prediction and evaluation. This method can promptly detect and respond to potential hidden dangers caused by environmental factors such as light concentration, warehouse layout, and QR code quality, thereby improving the safety and efficiency of operations. By dynamically adjusting the plan and optimizing control measures, the system effectively reduces the risk of delayed updates of equipment status information and ensures a smooth and accurate storage and retrieval process.
[0080] Embodiment 2, S2, establishes a warehouse entry and exit scene analysis model based on scene image data and standard box image data, and generates a warehouse entry and exit prediction evaluation coefficient.
[0081] Specifically, the acquired light gathering influence coefficient, warehouse layout influence coefficient and QR code guidance evaluation coefficient are used to establish a warehouse entry and exit scenario analysis model to generate a warehouse entry and exit prediction evaluation coefficient;
[0082] The calculation formula of the inventory in and out forecast evaluation coefficient is as follows:
[0083]
[0084] In the formula, is the inventory forecast evaluation coefficient, It is the preset proportional coefficient of the light gathering influence coefficient. It is the preset proportional coefficient of the warehouse layout influence coefficient. It is the preset proportional coefficient of the QR code guided evaluation coefficient. is the light gathering influence coefficient, is the warehouse layout influence coefficient, is the QR code guided evaluation coefficient.
[0085] S3, classify the in-and-out scenes into scene categories, which include normal scenes, low interference scenes, and excessive interference scenes. Classify the expected scene categories according to the in-and-out prediction evaluation coefficients, and implement corresponding management and control measures.
[0086] Specifically, during the process of standard containers entering and leaving the warehouse, classifying the entry and exit scenes is a key step to ensure smooth operation. First, the system divides the entry and exit scenes into three main categories based on the collected scene image data: normal scenes, low-interference scenes, and excessive interference scenes. Normal scenes refer to stable environmental conditions, good light, and no significant signal interference or other influencing factors. Low-interference scenes refer to a certain degree of insufficient light or slight interference in the environment, but the impact on operations is small. Excessive interference scenes refer to excessive light or serious signal interference in the environment, which may cause problems with device recognition and data collection.
[0087] Based on the scene classification, the system uses the in-and-out prediction evaluation coefficient to further evaluate and classify these scene categories; the prediction evaluation coefficient is based on the analysis of environmental data and is used to measure the impact of various scenes on in-and-out operations; for each scene category, the system will implement corresponding management and control measures; for example, for normal scenes, the system can maintain standard operating procedures; for low-interference scenes, it may be necessary to make appropriate adjustments to the light or signal; and for scenes with excessive interference, more stringent measures need to be taken, such as adjusting the light source, adding signal amplifiers or re-planning the path, etc., to reduce the impact of interference on operations; through this classification and control method, the system can flexibly respond to different environmental conditions and improve the accuracy and safety of in-and-out operations.
[0088] The entry and exit prediction evaluation coefficient is compared with the preset entry and exit hidden danger threshold and entry and exit failure threshold. The three corresponding results after comparison are divided into normal scenarios, low-interference scenarios, and excessive interference scenarios. For low-interference scenarios, appropriate adjustments are made to the light and signal. For scenarios with excessive interference, it is necessary to add signal amplifiers or re-plan the path.
[0089] It should be noted that the threshold of entry and exit hazards requires a set of proportional coefficient standards to be set in the blasting prediction assessment. These standards can be determined based on historical data, engineering experience, engineering technical specifications and other factors. For example, a standard proportional coefficient range can be set, including the possibility of normal scenarios, low interference scenarios, and excessive interference scenarios; in the work, an entry and exit hazard assessment is performed, and specific entry and exit prediction assessment coefficients are obtained. Then, these coefficients are compared with the preset proportional coefficient standards. By comparison, it can be determined within which standard range the actual assessment coefficient is, so as to preliminarily determine which category the scene category may belong to.
[0090] S4, when the scenario category is insufficient signal, the control measures are tracked, the control effects are evaluated, and feedback signals are issued.
[0091] Obtain multiple inbound and outbound forecast evaluation coefficients within the G time period during which the control measures are in place, synthesize the control set, calculate the standard deviation and average value of the inbound and outbound forecast evaluation coefficients within the control set, set the control standard deviation reference threshold and the control average value reference threshold; compare the standard deviation of the inbound and outbound forecast evaluation coefficients within the control set with the control standard deviation reference threshold, and compare the average value of the inbound and outbound forecast evaluation coefficients within the control set with the control average value reference threshold; divide the control effect into three types of control success, control failure, and control fluctuation based on the three comparison results; send feedback signals to relevant personnel after classification.
[0092] This embodiment establishes a warehouse entry and exit scene analysis model by integrating scene image data and standard box image data, and generates a warehouse entry and exit prediction evaluation coefficient to accurately manage environmental variables in warehouse entry and exit operations; first, by integrating the light aggregation influence coefficient, warehouse layout influence coefficient and QR code guidance evaluation coefficient to form a warehouse entry and exit prediction evaluation coefficient, the model can fully reflect the environmental influencing factors in warehouse entry and exit operations; this method can classify scenes into normal scenes, low interference scenes and excessive interference scenes, so as to implement targeted control measures according to different environmental conditions. For normal scenes, the operating process can maintain the standard, and for low interference scenes, the system will make appropriate adjustments, such as optimizing light and signals; for scenes with excessive interference, more complex adjustments may be required, such as adding signal amplifiers or replanning paths.
[0093] In addition, by comparing with the preset hidden danger threshold and failure threshold, the system can more accurately determine the risk level in the warehousing process and adjust the management and control strategy accordingly. Real-time tracking and effect evaluation of management and control measures enable the system to dynamically respond and optimize management and control measures to ensure the safety and effectiveness of operations. In general, this method improves the adaptability and flexibility of the system, makes warehousing management more refined, effectively reduces potential operational risks, and improves overall management efficiency and equipment safety.
[0094] 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.
[0095] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.
[0096] Those of ordinary skill in the art will appreciate that the modules 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.
[0097] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.
[0098] 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.
[0099] Finally: The above description is only a preferred 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. A computer vision-based target detection method, characterized in that: The steps include: When the standard boxes are in and out of the warehouse, the scene image data is collected and recognized through OCR, and the standard box image data is collected during the QR code scanning stage; Establish a warehouse entry and exit scene analysis model based on scene image data and standard box image data to generate warehouse entry and exit prediction evaluation coefficients; The scene image data includes a light gathering influence coefficient and a warehouse layout influence coefficient; the standard box image data includes a QR code guidance evaluation coefficient; The method for obtaining the light gathering influence coefficient is as follows: Obtain the light structure, light source distribution, and light gathering of the warehouse, and draw a three-dimensional image through a computer; Perform filtering on the three-dimensional image; The processed three-dimensional image is feature extracted, and the light intensity evaluation coefficient is calculated based on mass conservation and momentum conservation combined with diffusion coefficient and boundary conditions; The light path inside the warehouse is simulated through images, and a simulation equation is established to obtain the light divergence effect coefficient; The light gathering influence coefficient is calculated by combining the light intensity evaluation coefficient; The specific calculation formula of the light gathering influence coefficient is as follows: In the formula, is the light intensity evaluation coefficient, is the light divergence effect coefficient, is the light gathering influence coefficient, is the error term; The calculation formula of the warehouse layout influence coefficient is: In the formula, is the warehouse layout influence coefficient, is the total number of voxels in the voxel set, is the voxel label; is the volume of the ith voxel, is the blocking length of the ith obstacle path, and e is a natural number; The method for obtaining the QR code guided evaluation coefficient is as follows: The guidance information entropy of the QR code is obtained by solving the QR code recognition model of the finite element model; the maximum main information entropy value recognized is calculated; and the guidance evaluation coefficient of the QR code is calculated by combining the maximum main information entropy value recognized and the network signal interference degree; The calculation formula of the guided information entropy of the two-dimensional code is: The calculation formula of the maximum identified primary information entropy value is: The calculation formula of the QR code guidance evaluation coefficient is as follows: In the formula, is the guided information entropy of the QR code, is the maximum primary information entropy in the QR code, is the intermediate main information entropy in the QR code, is the adjustment item; is the maximum primary information entropy value identified; The area of the QR code recognized by OCR; is the QR code guided evaluation coefficient, is the network signal interference degree; The calculation formula of the inventory in and out forecast evaluation coefficient is as follows: In the formula, is the inventory forecast evaluation coefficient, It is the preset proportional coefficient of the light gathering influence coefficient. It is the preset proportional coefficient of the warehouse layout influence coefficient. It is the preset proportional coefficient of the QR code guided evaluation coefficient. is the light gathering influence coefficient, is the warehouse layout influence coefficient, is the QR code guided evaluation coefficient; The in-and-out scenarios are classified into scenario categories, including normal scenarios, low interference scenarios, and excessive interference scenarios. The predicted scenario categories are classified according to the in-and-out forecast evaluation coefficients, and corresponding control measures are implemented; When the scenario category is insufficient signal, the control measures are tracked, the control effects are evaluated, and feedback signals are issued.
2. The computer vision-based target detection method according to claim 1, characterized in that: The method for obtaining the warehouse layout influence coefficient is as follows: voxelize the three-dimensional model of the warehouse, divide the warehouse into multiple cubes, label them and obtain a voxel set; Count the number of voxels occupied by obstacles to get the obstacle volume; Identify and analyze the distribution of obstacle paths in the voxelized warehouse model, and calculate the curvature of the obstacle path; The warehouse layout impact coefficient is calculated by combining the curvature of the obstacle path and the obstacle volume analysis.
3. The computer vision-based target detection method according to claim 2, characterized in that: The predicted scenario categories are classified according to the inventory forecast evaluation coefficients, and corresponding control measures are implemented, specifically: Compare the in-and-out forecast evaluation coefficient with the preset in-and-out hidden danger threshold and in-and-out failure threshold; The three results after comparison are divided into normal scene, low interference scene and excessive interference scene; For low-interference scenarios, adjust the light and signal; In scenarios where the interference is too strong, it is necessary to add signal amplifiers and re-plan the path.
4. The computer vision-based target detection method according to claim 3, characterized in that: The aforementioned tracking of control measures, evaluation of control effects, and issuance of feedback signals are as follows: Obtain multiple inbound and outbound forecast evaluation coefficients within G time during the control measures, and build a control set; Calculate the standard deviation and average value of the inbound and outbound forecast evaluation coefficients in the control set, and set the control standard deviation reference threshold and control average reference threshold; Compare the standard deviation of the inbound and outbound forecast evaluation coefficients in the control set with the control standard deviation reference threshold, and compare the average value of the inbound and outbound forecast evaluation coefficients in the control set with the control average value reference threshold; Based on the comparison results, the control effects are divided into three types: control success, control failure, and control fluctuation; after classification, feedback signals are sent to relevant personnel.
5. The computer vision-based target detection method according to claim 4, characterized in that: The generation of the inbound and outbound forecast evaluation coefficient is specifically: The acquired light gathering influence coefficient, warehouse layout influence coefficient and QR code guidance evaluation coefficient are used to establish a warehouse entry and exit scenario analysis model to generate a warehouse entry and exit prediction evaluation coefficient; The calculation formula of the inventory in and out forecast evaluation coefficient is as follows: In the formula, is the inventory forecast evaluation coefficient, It is the preset proportional coefficient of the light gathering influence coefficient. It is the preset proportional coefficient of the warehouse layout influence coefficient. It is the preset proportional coefficient of the QR code guided evaluation coefficient. is the light gathering influence coefficient, is the warehouse layout influence coefficient, is the QR code guided evaluation coefficient.
6. The computer vision-based target detection method according to claim 5, characterized in that: The specific calculation formula of the light intensity evaluation coefficient is as follows: The specific calculation formula of the light divergence effect coefficient is as follows: In the formula, is the volume of the warehouse, is the light source volume of the warehouse, is the air density, is the diffusion coefficient of light in air, is the Laplace operator of light intensity, is the light energy scattering rate, is the average light energy, is the nonlinear term of light energy divergence.
7. The computer vision-based target detection method according to claim 6, characterized in that: The calculation formula of the obstacle volume is: The calculation formula of the curvature of the obstacle path is: 。
