Stone layout intelligent warehouse management method based on machine vision

Abstract

The invention relates to the technical field of inventory management, in particular to a stone layout intelligent warehouse management method based on machine vision, which comprises the following steps: acquiring stone image temperature and distance data through multi-modal sensing, generating a feature set through graying quantization and time difference, performing Kalman filtering fusion after synchronous calibration, and dynamically adjusting the weight. And outputting a positioning result, analyzing a reflection coefficient, adjusting a light source, optimizing a light field parameter, collecting a texture reconstruction three-dimensional model, correcting and updating an error, extracting position information based on the model, correcting an inventory strategy, and generating an inventory management scheme. The light field is optimized through illumination monitoring and reflection adjustment to ensure that the stone texture is complete and stable, the feature synchronous storage state is updated in real time through three-dimensional model dynamic reconstruction and correction, the stacking level and goods allocation recognition accuracy is improved, and the inventory information reliability and consistency are enhanced.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of inventory management, in particular to a stone slab intelligent warehouse management method based on machine vision. BACKGROUND

[0002] The technical field of inventory management involves optimizing and managing the storage, classification, scheduling, and transportation of goods and materials. The core content of this field includes real-time monitoring of inventory data, accurate prediction of inventory quantity, automation and intelligent operation of inventory systems, efficient inventory scheduling, and improvement of warehouse management efficiency. The technology covers multiple aspects such as Internet of Things technology, intelligent identification technology, and automated logistics equipment, aiming to achieve fine management of inventory to meet the needs of modern production and consumption.

[0003] Among them, the traditional stone slab intelligent warehouse management method refers to the processing method for positioning and management problems caused by specification differences, large weight, and large storage quantity of stone materials in the warehouse link. It relies on manual block-by-block numbering registration or barcode label pasting for identification, and then uses manual counting or semi-automatic scanning equipment for data entry and management.

[0004] Traditional stone warehouse management relies on manual block-by-block recording or barcode label identification, which is prone to label omission or information confusion in scenarios with large quantities and dense stacking. Manual counting is time-consuming and difficult to ensure stable recognition in varying environmental lighting or deep stacking levels. Semi-automatic scanning often experiences data loss or positioning deviation due to reflection and obstruction, resulting in outdated inventory data updates, inaccurate storage location positioning, and prolonged search time, which adversely affects overall management efficiency under high-frequency warehouse entry and exit conditions. SUMMARY

[0005] To solve the technical problems of traditional stone warehouse management relying on manual block-by-block recording or barcode label identification, which is prone to label omission or information confusion in scenarios with large quantities and dense stacking, manual counting being time-consuming and difficult to ensure stable recognition in varying environmental lighting or deep stacking levels, semi-automatic scanning often experiencing data loss or positioning deviation due to reflection and obstruction, resulting in outdated inventory data updates, inaccurate storage location positioning, and prolonged search time, which adversely affects overall management efficiency under high-frequency warehouse entry and exit conditions, the present application provides a stone slab intelligent warehouse management method based on machine vision, comprising the following steps: To achieve the above-mentioned purpose, the present application adopts the following technical solution: a stone slab intelligent warehouse management method based on machine vision, comprising the following steps: S1: Collect stone surface image data, infrared temperature distribution data, and ultrasonic distance measurement data through a multi-modal sensor, perform grayscale processing on the image data, quantization processing on the temperature data, and time difference calculation on the distance data, and generate a multi-modal sensing feature data set; S2: Perform time synchronization calibration based on the multi-modal sensing feature data set, input the synchronized data into a Kalman filter algorithm for fusion processing, dynamically adjust the weight coefficients according to the shelf positioning accuracy and storage space identification, and generate comprehensive positioning result data; S3: Based on the comprehensive positioning result data, analyze the reflection coefficient of the stone stacking level and the storage space coding area, detect the light intensity value, adjust the light source parameters when the reflection coefficient exceeds the preset reflection threshold, and generate an optimized light field parameter set; S4: Re-collect the stone surface texture data according to the optimized light field parameter set, input the texture data into a three-dimensional reconstruction algorithm for model construction, update the surface feature spectrum through error correction calculation, and generate dynamic three-dimensional model data.

[0006] As a further scheme of the present application, the multi-modal sensing feature data set includes texture information features, temperature distribution features, and depth information features, the comprehensive positioning result data includes position coordinate information, stacking level information, and region identification information, the optimized light field parameter set includes light intensity parameters, light source direction parameters, and light source frequency parameters, and the dynamic three-dimensional model data includes geometric structure models, surface texture models, and error correction models.

[0007] As a further scheme of the present application, the specific steps of S1 are: S101: After obtaining the stone surface image data, convert the color channel values into a gray value matrix using a pixel gray scale conversion method, re-encode the brightness according to the channel proportion of each pixel point in the matrix, and calculate the gray scale difference of the pixel points according to the brightness difference, to generate a gray pixel value set; S102: According to the collected infrared temperature distribution data, call the pixel coordinates in the gray pixel value set, discretize the temperature continuous value according to the temperature interval threshold, and superimpose the discrete value at the coordinate position, calculate the temperature level by comparing the interval threshold, and obtain the temperature interval value distribution; The temperature interval threshold is used to discretize the collected continuous temperature value into different level intervals according to a preset or adaptive division rule, to realize the standardized hierarchical representation of temperature data; S103: For ultrasonic distance measurement data, call the coordinate position in the temperature interval value distribution, calculate the distance value according to the time difference between the transmitted and received signals, and integrate it with the temperature distribution value and gray pixel value of the corresponding coordinate, to generate a multi-modal sensing feature data set.

[0008] As a further scheme of the present application, the specific steps of S2 are: S201: Based on the multi-modal sensor feature data set, the sensor output signal detection time record difference value under the same time period is detected, the offset between the sensor record time is calculated, the time reference correction is performed on the data sequence, and the time synchronization difference value sequence is generated; S202: The time synchronization difference value sequence is called and input into the Kalman filtering algorithm, the residual between the predicted state and the observation value is weighted, the state transition matrix and the covariance matrix are adjusted, the filtering process is updated, and the fusion positioning estimation value is generated; S203: According to the fusion positioning estimation value, the shelf positioning accuracy value and the storage space identification value are compared, the distribution proportion of different weight coefficients in the Kalman filtering is adjusted, the corresponding weight value of position, velocity and observation signal is updated, and the comprehensive positioning result data is generated.

[0009] As a further scheme of the present application, the specific steps of S3 are: S301: Based on the comprehensive positioning result data, the stone stacking level and the storage location coding area are traversed, the space coordinates are compared with the area boundary value, the stone surface reflection coefficient is extracted, and the area reflection coefficient value set is generated; S302: The area reflection coefficient value set is called to detect the light intensity of the storage location coding area, the reflection coefficient value is compared with the set reflection threshold value, the area exceeding the threshold value is selected and the light intensity is recorded, and the threshold light intensity value is obtained; The set reflection threshold value is used to determine whether the area surface reflection coefficient reaches the minimum standard of light demand, so as to select the target area to be optimized; S303: According to the threshold light intensity value, the brightness parameter and the angle parameter of the light source are jointly adjusted, the light intensity value range is distributed and optimized, the adjusted light source parameter is redistributed to the storage location coding area, and the optimized light field parameter set is generated.

[0010] As a further scheme of the present application, the specific steps of S4 are: S401: Obtain the optimized light field parameter set, detect the light intensity value and the stone surface reflectivity value under different incident angles, establish the corresponding relationship between the two by polynomial fitting, adjust the light intensity distribution according to the difference value, collect the stone surface texture image data and compare the pixel matrix with the light intensity value, and generate the light texture matching coefficient; S402: The light texture matching coefficient is called, the texture image pixel matrix is input into the three-dimensional reconstruction algorithm, the depth difference value is calculated according to the gray value and three-dimensional coordinate point position mapping relationship, the three-dimensional point cloud distribution is analyzed by combining the depth difference value, and the surface point cloud density value is obtained by aggregation according to the coordinate adjacency. S403: According to the surface point cloud density value, compare the point cloud density with the reconstruction error threshold value in the error correction calculation, screen the over-limit point position and adjust the three-dimensional coordinate value, update the surface feature map, and generate dynamic three-dimensional model data.

[0011] As a further scheme of the present application, the illumination texture matching coefficient refers to a coefficient value calculated from the gray scale distribution of the stone surface texture image pixel matrix and the corresponding illumination intensity value under the condition of a certain incident angle and illumination intensity; The surface point cloud density value refers to the ratio of the number of three-dimensional coordinate points after effective point cloud determination to the area per unit surface area; The reconstruction error threshold value refers to the allowed numerical range of the difference between the depth value of the point obtained by three-dimensional reconstruction and the reference depth value.

[0012] As a further scheme of the present application, the method further comprises a S5 step: S5: Based on the dynamic three-dimensional model data, the position information extraction calculation of the inventory checking unit and the warehouse entry and exit record table is performed, the position deviation value is calculated, the frequency parameter of the dynamic three-dimensional model data update is adjusted according to the position accuracy, and the inventory management scheme is generated; The inventory management scheme includes position accuracy strategy, inventory scheduling strategy and warehouse entry and exit control strategy.

[0013] As a further scheme of the present application, the specific steps of S5 are: S501: Based on the dynamic three-dimensional model data, the coordinate information of the inventory checking unit and the warehouse entry and exit record table is extracted, the coordinate difference vector is calculated according to the corresponding comparison of the material number, and the unified difference set is summarized to generate the position deviation value; S502: The position deviation value is called, the coordinate difference amount of the checking unit at different time nodes is compared with the coordinate accuracy reference value, the checking compliance proportion is calculated, the overall recognition matching degree is analyzed by weighted average analysis according to the weight setting of the checking unit quantity proportion, and the position accuracy is obtained; S503: According to the position accuracy and the inventory management strategy adjustment reference value, it is judged whether the checking frequency and the warehouse entry and exit calibration mechanism need to be adjusted, the difference range is re-set, and the inventory management scheme is generated.

[0014] As a further scheme of the present application, the position deviation value is the difference vector between the dynamic three-dimensional model data coordinate of the checking unit and the corresponding coordinate of the warehouse entry and exit record table; The position accuracy is the proportion of the number of checking units whose position deviation value is within the allowed range after weight processing to the total number of checking units under the condition of a given coordinate accuracy reference value; The inventory management strategy adjustment reference value is a threshold value set in advance based on the position accuracy.

[0015] Compared with the prior art, the application has the advantages and positive effects that: In the application, through the cooperative collection and fusion of multi-source sensing data, the positioning result is consistent in different measurement dimensions, error diffusion caused by environmental interference is avoided, in real-time monitoring of light intensity and adjustment of reflection coefficient, the light field is optimized, the integrity and stability of stone texture information in the collection process are ensured, the surface features of the three-dimensional model can be updated in real time under dynamic reconstruction and error correction, the model and the actual storage state are synchronized and matched, the continuous data optimization process improves the accuracy of stacking level and location recognition, the positioning and management in complex scenes are more precise, and the reliability and consistency of inventory information are effectively enhanced. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1 The step flowchart of the present application is shown in the figure. Figure 2 The S1 refinement schematic diagram of the present application is shown in the figure. Figure 3 The S2 refinement schematic diagram of the present application is shown in the figure. Figure 4 The S3 refinement schematic diagram of the present application is shown in the figure. Figure 5 The S4 refinement schematic diagram of the present application is shown in the figure. Figure 6 The S5 refinement schematic diagram of the present application is shown in the figure. DETAILED DESCRIPTION

[0018] The technical solutions in the present application will be described below with reference to the drawings.

[0019] In the embodiments of the present application, the words such as "example", "for example" are used to represent as an example, illustration or description. Any embodiment or design scheme described as "example" in the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the word "example" is intended to present the concept in a specific way. In addition, in the embodiments of the present application, the meaning expressed by "and / or" can be both, or can be one of the two.

[0020] In the embodiments of the present application, "image" and "picture" can be used interchangeably, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized. "Of", "corresponding" and "corresponding" can be used interchangeably, and it should be pointed out that the meanings expressed are consistent when the distinction is not emphasized.

[0021] In the embodiments of the present application, sometimes the subscript such as W1 can be written in the form of non-subscript such as W1, and the meanings expressed are consistent when the distinction is not emphasized.

[0022] In order to make the technical problems, technical solutions and advantages to be solved by the present application more clear, the following will be described in detail in conjunction with the drawings and specific embodiments.

[0023] Please refer to Figure 1 The embodiments of the present application provide a stone surface intelligent warehouse management method based on machine vision, comprising the following steps: S1: Collecting stone surface image data, infrared temperature distribution data and ultrasonic distance measurement data through multi-modal sensors, performing gray scale processing on the image data, quantization processing on the temperature data and time difference calculation on the distance data, and generating a multi-modal sensing feature data set; S2: Time synchronization calibration based on the multi-modal sensing feature data set, inputting the synchronized data into Kalman filtering algorithm for fusion processing, dynamically adjusting the weight coefficient according to the shelf positioning accuracy and storage space identification, and generating comprehensive positioning result data; S3: Based on the comprehensive positioning result data, the reflectivity of the stone stacking level and the storage space coding area is analyzed, the light intensity value is detected, and when the reflectivity exceeds the preset reflection threshold, the light source parameters are adjusted, and an optimized light field parameter set is generated; S4: Re-collecting stone surface texture data according to the optimized light field parameter set, inputting the texture data into a three-dimensional reconstruction algorithm for model construction, updating the surface feature spectrum through error correction calculation, and generating dynamic three-dimensional model data; S5: Based on the dynamic three-dimensional model data, the position information of the inventory checking unit and the warehouse entry and exit record table is extracted and calculated, the position deviation value is calculated, the frequency parameter of dynamic three-dimensional model data updating is adjusted according to the position accuracy, and a warehouse management scheme is generated; The multimodal sensing feature data set includes texture information features, temperature distribution features and depth information features; the comprehensive positioning result data includes position coordinate information, stacking level information and area identification information; the optimized light field parameter set includes light intensity parameters, light source direction parameters and light source frequency parameters; the dynamic three-dimensional model data includes geometric structure model, surface texture model and error correction model; the inventory management plan includes position accuracy strategy, inventory scheduling strategy and inbound and outbound control strategy.

[0024] See also Figure 2 , the specific steps of S1 are: S101: After obtaining the stone surface image data, the color channel values ​​are sequentially converted into a grayscale value matrix using a pixel grayscale conversion method, the brightness is re-encoded according to the channel ratio of each pixel in the matrix, and the grayscale discrimination of the pixel points is calculated based on the brightness difference to generate a grayscale pixel value set; After acquiring the image data of the stone surface, the collected color image is decomposed into three independent channels: R (red channel), G (green channel), and B (blue channel), and the integer values ​​in the range of 0 to 255 are read in turn for the pixel points of each channel. The color value is converted into a grayscale value by performing weighted proportional conversion on the three channel values ​​of a single pixel point. The weight can be set according to the reflective characteristics of the stone surface under the target lighting conditions. For example, in a light-colored granite sample, the red weight is set to 0.3, the green weight is set to 0.59, and the blue weight is set to 0.11. Multiplication operations are performed on the three channels of each pixel point and the sum is obtained to obtain a preliminary grayscale value. Subsequently, brightness recoding is performed on each pixel value. The brightness recoding here is achieved by comparing the preliminary grayscale value with a preset brightness reference value. For example, the brightness reference value is set. , when the gray value When it is greater than 128, weighted amplification is performed. , when the gray value When it is less than 128, perform weighted reduction operation The weighted coefficients of 1.1 and 0.9 are obtained by referring to the statistical analysis of the brightness difference of stones under artificial light. The statistical method is to select 100 sample pixels to measure the mean and calculate the variance, and select the minimum variance point as the setting value. After the above brightness adjustment is completed, the difference operation is performed on the grayscale values ​​of adjacent pixel coordinates, and the difference is greater than the brightness contrast threshold. The pixels with grayscale difference are regarded as pixels with significant grayscale difference. The threshold is obtained by taking the mean of the extreme difference of the grayscale difference statistics of 10 different types of stone samples. For example, the adjacent pixel difference results of sample 1 are The mean value is 15, sample 2 is 14.6, and finally 15 is selected as the standard threshold. After executing the entire image traversal, the pixel coordinates and their gray values determined by the brightness difference are added to the gray pixel value set, and a two-dimensional matrix is constructed for storage. In the two-dimensional matrix, each element records the pixel coordinates and gray values in the form of . Table 1: Stone surface sampling pixel gray data table

[0025] As shown in Table 1, 4 consecutive horizontal pixels are selected in the sampling process. First, the gray values are calculated, for example, the initial R=140, G=150, and B=160 in the first row, then =140×0.3+150×0.59+160×0.11 = 42+88.5+17.6 =148.1 ≈ 148, since =148>128, the magnification is executed, and then compared with the gray values of adjacent pixels to obtain the difference , which is less than the threshold 15, and is not selected into the gray set. This process is executed according to the full image pixel traversal to ensure the integrity of the gray pixel value set, and finally a gray value matrix organized by coordinates is obtained.

[0026] S102: According to the collected infrared temperature distribution data, the pixel coordinates in the gray pixel value set are called, the temperature continuous values are discretized according to the temperature interval threshold, and the discrete values are superimposed on the coordinate position. The temperature level is calculated by comparing the interval threshold, and the temperature interval value distribution is obtained. The temperature interval threshold is used to discretize the collected continuous temperature values into different level intervals according to a preset or adaptive division rule, so as to realize the standardized hierarchical representation of temperature data. The infrared temperature distribution data is obtained, and the corresponding position of the infrared temperature continuous value is called point by point according to the pixel coordinates (x, y) stored in the gray pixel value set. First, unit conversion and normalization are performed on the temperature data stream collected by the infrared sensor. The temperature value in Celsius (℃) is mapped to the [0, 1] interval. The Min-Max normalization method is adopted, wherein is the measured temperature value of the pixel position, and are the lowest temperature and the highest temperature of the current full image sampling, respectively. For example, , When the normalized temperature of the pixel point with coordinates (101, 200) is 25.6, the normalized temperature , and then the normalized temperature is mapped back to the temperature grade interval. Here, the temperature interval threshold is set into five grade intervals based on the temperature distribution experience of the actual storage environment: interval 1 is T < 23.5, interval 2 is 23.5 ≤ T < 24.8, interval 3 is 24.8 ≤ T < 26.1, interval 4 is 26.1 ≤ T < 27.4, and interval 5 is T ≥ 27.4. The classification is based on and The difference is divided into five equal parts and rounded to the nearest 0.1°C. For example, a temperature difference of 6.4°C is divided by 5 to get about 1.28°C. The difference is added to the starting point to get the demarcation value. Then the pixel temperature of 25.6℃ is compared with the classification interval to determine that it belongs to interval 3 and assigned a discretized temperature value of 3. The above operation needs to be repeated for the temperature value corresponding to each pixel coordinate in the grayscale set, that is, first extract (x, y), then call the temperature value at the same position in the sensor storage array, and perform the process of normalization, interval judgment, and number assignment. After the number assignment is completed for the pixel position, The triplet form is superimposed on the grayscale matrix, where To obtain the final discretized value, the adjacent temperature level differences need to be analyzed and processed in the multi-point continuous judgment. If the temperature number difference of the cluster at the same position exceeds 1, it is recorded as a temperature mutation point. For example, the adjacent pixel A , pixel B The difference is 2, and a mutation is determined. Finally, the discretized matrix of the entire image is verified. The verification mechanism is to count the pixel proportions of the numbered intervals. If the proportion of a certain temperature level is less than 1% of the total number of pixels in the entire image, the level is deleted from the final temperature interval value distribution result and the temperature classification of its coordinate point is assigned to the majority class of the adjacent level. After the above normalization, hierarchical mapping, number assignment and difference verification processing, the temperature interval value distribution arranged in pixel coordinates is finally obtained.

[0027] S103: Calling the coordinate position in the temperature interval value distribution for the ultrasonic distance measurement data, calculating the distance value based on the time difference between the transmitted and received signals, and integrating it with the temperature distribution value and grayscale pixel value of the corresponding coordinate to generate a multimodal sensing feature data set; After obtaining the ultrasonic distance measurement data, call the coordinates (x, y) in the temperature interval value distribution point by point to read the discrete temperature value at that location and grayscale value , and then read the emission time in the ultrasonic ranging raw data corresponding to the position and receiving time First, the time difference is obtained by subtracting the transmission time from the reception time. ,For example , , then the difference between the two is ; then according to the actual temperature value of the point The sound speed at this temperature is calculated, which is the base sound speed 331.4 m / s plus the temperature correction amount That is, the sound speed is 331.4 + (0.6 x 25.6) = 346.76 m / s; then the sound speed is multiplied by the time difference to obtain the round-trip distance of the sound wave 346.76 x 0.00482 = 1.6694 m, since this value is the round-trip path length, it needs to be divided by 2 again to obtain the one-way distance, the result is 0.8347 m; this value is combined with the temperature number of the same coordinate position And the gray value To form a multimodal triplet ; In the process of traversing the whole image coordinate by coordinate, the above process needs to be repeated, that is, first locate the coordinate, then obtain the temperature number and gray value, at the same time read the emission and reception time and calculate the difference, use the temperature corrected sound speed, and then combine the time difference to calculate the one-way distance; In order to avoid outliers, each calculated distance value needs to be judged for interval rationality, if the distance is less than the minimum effective stacking distance 0.50 m set by the storage system and greater than zero, it is classified as an effective distance, otherwise it is marked as an invalid point; When the effective point calculation is completed, a plurality of multimodal triplets are sequentially stored according to the coordinate index, and the multimodal data in different regions are marked with the same material category label In this way, a complete multimodal sensing feature data set is obtained, which contains the visual gray value, temperature classification information and spatial distance measurement value of each effective pixel.

[0028] Please refer to Figure 3 The specific steps of S2 are: S201: Based on the multimodal sensing feature data set, the time record difference value of the sensor output signal in the same time period is detected, the offset amount between the sensor record time is calculated, and the time reference correction is performed on the data sequence to generate a time synchronization difference value sequence; Based on the multi-modal sensor feature data set, first, for the same time period, the output signal timestamp sequence of each sensor in the time period is read, each record in the sequence is composed of a timestamp value and the sensor identifier to which it belongs, for example, a group of timestamp data is read from the infrared temperature sensor, ultrasonic ranging sensor and visual acquisition terminal respectively, which is converted into a unified time format accurate to milliseconds, after reading all the data, first select a timestamp sequence of a sensor as the reference sequence, select the timestamp of the visual acquisition unit as the reference, extract the time value of the corresponding sampling point in the reference sequence, then traverse the record time of the remaining sensors under the same sampling point number, subtract the reference time value to obtain the time difference value, through this kind of two-by-two subtraction calculation, the time record difference of each corresponding sampling point is obtained, for example, for sampling point 1, the timestamp of the visual acquisition is 15:32:10.125, the timestamp of the infrared temperature sensor is 15:32:10.140, the difference is 0.015s, for sampling point 2, the difference is-0.009s, and so on to obtain the whole time difference value set, then calculate the average value of the whole time difference value set as the overall time offset of the current device, for example, the difference values of the 10 sampling points are: ; The sum of the differences is 0.051 seconds, and the average offset is 0.0051s, then the offset is used to perform addition or subtraction correction on the timestamps in the whole sequence of this kind of sensor, if the offset is positive, the timestamps of the remaining sensors are subtracted by the value to align forward to the reference time, if it is negative, it is added to align backward, a new time synchronization sequence is generated after correction, on this basis, the modified sequence is compared with the reference sequence point by point again, a new time difference value set is calculated, and a time synchronization difference value sequence is obtained, which saves the residual asynchronization amount between the sampling points of the two devices after the time synchronization reference correction, in order to ensure data quality, a time deviation threshold is set in the generated time synchronization difference value sequence, the threshold value of this embodiment is selected as 0.002s, which refers to the maximum time drift allowed by the system, the drift is obtained through the variance stable interval of the test system under multiple synchronization calibration, when the absolute value of any difference value in the time synchronization difference value sequence exceeds 0.002s, it is marked as an abnormal difference value and the abnormal point number is recorded, the rest is regarded as a normal difference value for subsequent processing, the time synchronization difference value sequence generated in this way will be provided to the subsequent step for filtering and fusion processing.

[0029] Table 2: Sensor sampling time difference value record table

[0030] As shown in Table 2, the time difference value of the sampling point is close to zero after correction, and the residual difference value is input to the next step as a time synchronization difference value sequence.

[0031] S202: Call the time synchronization difference sequence and input it into the Kalman filter algorithm. Weight it according to the residual between the predicted state and the observed value, adjust the state transfer matrix and covariance matrix, update the filtering process, and generate a fusion positioning estimate. After calling the time synchronization difference sequence, first read the sequence point by point and use it as the input initial data for filtering processing, and read each synchronization difference and the sensor state observation value at the corresponding time , and then initialize the predicted state vector, where the position value can be set to the estimated position of the previous moment, and the speed value can be calculated based on the position difference and time interval of adjacent moments. For example, the estimated position of the previous moment is 3.254m, the current and previous moments are separated by 0.10s, and the position difference is 0.023m, then the speed prediction value is 0.23m / s. The predicted state is used as the initial input of the filtering process, and then the residual between the predicted state and the current observation value is calculated. The residual is calculated by subtracting the predicted value from the current observation value. For example, if the observed position is 3.261m and the predicted position is 3.254m, the residual is 0.007m. After obtaining the residual, weighted processing is required. At this time, the weight coefficient of the current step is called The residual is scaled. In this embodiment, the initial setting is =0.68, this value is determined by referring to the best point of residual convergence in multiple offline tests of the system. The test method is to Increasing from 0.1 to 1.0, the weight value corresponding to the minimum output mean square error is taken as the set value, the residual 0.007×0.68 is used to obtain a weighted residual of 0.00476m, and then the weighted residual is superimposed and corrected with the predicted state to obtain an updated state estimate of 3.25876m. At the same time, the coefficients of the state transfer matrix are adjusted. The position and speed transfer items in the state transfer matrix are adjusted according to the time interval. For example, the speed coefficient of the position in the matrix is ​​adjusted from 0.10 in the previous cycle to 0.093 in the actual time interval of this cycle. The covariance matrix needs to be dynamically updated according to the noise level, and its diagonal position variance item is updated by comparing the observed noise variance. and the prediction noise variance To correct, if Greater than , then the covariance increases by a proportional factor at the corresponding position , here =1.15, which comes from the average ratio when the noise in the original sampling suddenly changes, such as when =0.0005, =0.0007, the position covariance value is updated to 0.0005*1.15=0.000575, and the entire weighted matrix updating process is repeated at each sampling point of the time synchronization difference value sequence, so that each step in the continuous sequence is completed by a series of prediction state, observation value, residual calculation, weight correction and matrix updating, and the finally generated fusion positioning estimation value sequence is the position information sequence updated after iteration based on the time synchronization input.

[0032] S203: According to the fusion positioning estimation value, the shelf positioning accuracy value and the storage space identification value are compared, the distribution proportion of different weight coefficients in Kalman filtering is adjusted, the corresponding weight values of position, velocity and observation signal are updated, and the comprehensive positioning result data is generated; According to the fusion positioning estimation value, first, the estimated position value at each time is read in turn and the corresponding shelf positioning accuracy reference value The reference value is a reference value collected by a high-precision laser ranging device at the same shelf position, for example, the accuracy reference value of a certain shelf position is 3.260m, and the accuracy tolerance threshold is set to ±0.005m, when the absolute value of the difference between the fusion positioning result and the reference value is less than or equal to 0.005m, it is determined that the matching is normal, otherwise it is recorded as a deviation point, then for each estimated point, the corresponding storage space identification value is read, which is the code assigned to the specific storage space by the warehouse management system, for example represents the shelf unit of row 205, column B, the offset between the fusion positioning estimation value and the reference value is rounded to millimeters, for example, the estimated value is 3.25876m, the deviation is-0.00124m, and the rounding is-1mm, this deviation value will participate in the adjustment calculation of multi-weight distribution together with the original deviation data corresponding to the storage space identification, in this adjustment process, first, read the three types of weight coefficients in the Kalman filtering process, which are position update weight , velocity update weight and observation signal weight The initial values of this embodiment are 0.68, 0.21 and 0.11 respectively, then the adjustment proportion of each type of weight is calculated, which is derived from the ratio of the deviation value to the accuracy tolerance ratio, and then multiplied by the adjustment coefficient The is obtained from the original stable running data, and this time the setting =0.35, for example, the proportion of the current deviation-1mm to the tolerance 5mm is 0.2, multiplied by 0.07, then is adjusted to 0.68+(0.07*sign(-1))=0.61, where sign(-1) represents that the deviation direction is negative, and the position weight is reduced, and the reduced weight value is evenly distributed to With Each category increases by 0.035, and the new = 0.245, = 0.145, so the dynamically adjusted weight coefficient set will replace the original weight in the Kalman filter, and the new weight will be applied in the next positioning cycle to ensure that the weight proportion matches the current positioning accuracy requirement. The point of the fusion positioning estimation value sequence is traversed, and the above weight adjustment process is repeatedly executed, and finally the comprehensive positioning result data generated by the dynamically adjusted Kalman filter output and the storage space identification value form a corresponding relationship, ensuring that each output position value and the corresponding storage space label are matched one by one.

[0033] Please refer to Figure 4 The specific steps of S3 are: S301: Based on the comprehensive positioning result data, traverse the stone stacking level and storage space coding area, compare the space coordinates with the area boundary value, extract the stone surface reflectance coefficient, and generate a set of area reflectance coefficient values; Read the comprehensive positioning result data, first traverse the data set point by point, and extract the space coordinate value and the bound storage space coding information corresponding to each stacking position in turn, call the mapping table of storage space coding and physical space boundary in the warehouse database, and obtain the boundary range of the storage space in the three-dimensional coordinate system, wherein the boundary is composed of the minimum coordinate value and the maximum coordinate value , for example, the space range of storage space coding is to , compare each coordinate in the comprehensive positioning data with the minimum value and the maximum value of the corresponding area axis by axis during the traversal process, when the X, Y and Z three axes are all within the closed interval of the area, it is determined that the point falls into the storage space coding area, otherwise continue to compare with the boundary value of the next area, until the matching is completed, after the coordinate and the area are matched, read the stone surface brightness value L collected by the vision sensor corresponding to the coordinate point, and the pre-calibrated incident light intensity , and use the two to calculate the reflectance coefficient, first normalize the brightness value, that is, divide the brightness value L by the maximum sampling brightness of the device to get the normalized brightness , for example, the brightness of a certain point is 184, and the maximum sampling brightness is 255, then , then divide the normalized brightness by the incident light intensity to get the reflectance coefficient, for example, the incident light intensity = 0.85 (unit relative amount), the reflection coefficient is 0.849, the calculation of the reflection coefficient needs to be repeated for each matched coordinate, and the results are written into the area reflection coefficient set in turn, the storage structure of the area reflection coefficient set is a list, each item is a (storage location code, reflection coefficient) key value pair, in addition, in order to exclude outliers, a reasonable interval of the reflection coefficient needs to be set, the range of the conventional reflection coefficient obtained in the stone surface characteristic calibration process is 0.40≤R≤0.95, if the calculated reflection coefficient exceeds this range, the point data is discarded, so as to avoid the influence of environmental noise or measurement failure on the overall accuracy of the set, after the above steps are performed, the finally obtained area reflection coefficient set will contain each storage location code and its corresponding reflection coefficient value.

[0034] Table 3: Stone storage location boundary and reflection coefficient sampling data table

[0035] As shown in Table 3, the sampling points of the two storage locations respectively obtain the reflection coefficient values of 0.849 and 0.799 after boundary comparison, brightness normalization and reflection coefficient calculation, the data will enter the next step for light intensity detection and threshold comparison.

[0036] S302: Call the area reflection coefficient value set to detect the light intensity of the storage location code area, compare the reflection coefficient value with the set reflection threshold value, select the area exceeding the threshold value and record the light intensity, and obtain the threshold-exceeding light intensity value; The set reflection threshold value is used to determine whether the area surface reflection coefficient reaches the minimum standard of light requirement, so as to select the target area to be optimized; Call the area reflection coefficient value set, read each (storage location code, reflection coefficient) data item in turn, and call the stored current light intensity measurement value of the storage location code as the key value , the value is obtained by the light sensor output fixed at the top of the shelf, and has been converted into the relative light intensity unit [0, 1] interval, for example, the measured value of the storage location is 0.83, and the measured value of the storage location is 0.88, then the reflection coefficient R of each storage location is compared with the set reflection coefficient threshold value , the threshold value is based on the statistical analysis result of the reflection coefficient of the same material under different light environments, after taking the mean value 0.78 of the reflection coefficient collected from 100 stone samples and adding two standard deviations 0.06 to obtain 0.90, combined with the spectral characteristics of the warehouse lamps, it is corrected to 0.85 as the final threshold setting, that is = 0.85, in the comparison, if , it is determined as a high reflection area, otherwise it is a normal area, for example, the storage location The reflection coefficient R=0.849 is -0.001 lower than the threshold, so it is judged as normal. If the reflection coefficient R=0.799, the difference is -0.051, which is also normal. If the reflection coefficient of a certain position is 0.870, the difference with the threshold is 0.020, and it is determined to be an over-threshold area. After completing the reflection coefficient determination, it is necessary to further detect the light intensity value of the over-threshold area and change its light intensity. Corresponding to the system light intensity classification standard, the standard classification is: low illumination area <0.60, medium illumination area 0.60≤ <0.80, high illumination area ≥0.80, to determine the illumination range so that the light source power can be allocated according to different intensities in the subsequent light field parameter adjustment, such as the cargo position If R=0.872 and =0.79, then the area is the medium illumination threshold area, the light intensity value is recorded as 0.79, and the cargo position If R=0.868 and =0.81, it is determined to be a high illumination over-threshold area, and the light intensity value is recorded as 0.81. After the traversal is completed, the (cargo location code, R, ) data is appended to the super-threshold light intensity list, which only saves the precise values ​​of the light intensity of the areas judged as high-reflection areas in the classification, and is used for subsequent distribution optimization processing according to the light intensity conditions of the area when adjusting the brightness and angle of the light source.

[0037] S303: Based on the above-threshold light intensity value, jointly adjust the brightness parameter and angle parameter of the light source, optimize the distribution according to the light intensity value range, and redistribute the adjusted light source parameters to the cargo location coding area to generate an optimized light field parameter set; Read the list of super-threshold light intensity and traverse each data item (cargo location code, R, ), where the light intensity value Extract as the light source adjustment benchmark, and read the light source brightness parameters originally corresponding to the cargo location code (Unit: cd / m 2 ) and angle parameters (Unit: degrees), in the initial state, the factory-calibrated brightness of the light source is =1200cd / m 2 And the angle of light direction is =45°, then according to Calculate the brightness correction amount that needs to be adjusted at the interval position in the light intensity classification and angle correction , the brightness correction comes from the light intensity and the standard target intensity The difference multiplied by the luminance proportion coefficient Wherein Set to 0.75 (unit relative value) to ensure the middle interval light balance, the proportion coefficient Take 800 cd / m 2 The value is derived from the lamp response curve test results, such as the goods location The difference =0.79, with a target difference of 0.04, then the luminance correction amount is 0.04x800=32 cd / m 2 , the new luminance parameter =1200-32=1168 cd / m 2 Indicates that the luminance needs to be reduced, the angle correction amount Then calculate according to the exceeding amplitude of the relative threshold value of the reflection coefficient R, and the exceeding part is recorded as The coefficient is set to 25°, which is derived from the measured results of the sensitivity of the light incidence direction to the reflection intensity, such as R=0.872, , then exceeding 0.022, the corresponding angle correction amount is 0.022x25≈0.55°, when the light source is too bright and the exceeding threshold amplitude exceeds 0.015, the adjustment strategy is to reduce the luminance and increase the angle value to deviate from the reflection direction, so , on the contrary, if is low and the R exceeding threshold amplitude is small (between 0 and 0.010), the luminance is increased and the angle is slightly adjusted to a negative value, such as the goods location If =0.81, the difference 0.06 corresponds to a luminance correction of 48 cd / m 2 Down, and corresponds to an angle increase of 0.45°, after traversing the threshold exceeding area, write (goods location code, , ) into the new light field parameter list, which stores all the optimized light source luminance and angle combinations, and uses the goods location code as the index, finally generating an optimized light field parameter set, which provides a data basis for subsequent system light source power output and direction control according to the set.

[0038] Please refer to Figure 5 , the specific steps of S4 are: S401: Obtain the optimized light field parameter set, detect the light intensity value and the stone surface reflectivity value under different incidence angles, establish the corresponding relationship between the two by polynomial fitting, adjust the light intensity distribution according to the difference size, collect the stone surface texture image data and compare the pixel matrix with the light intensity value, and generate a light texture matching coefficient; Obtain the optimized light field parameter set, traverse the entries (goods location code, , ), the light brightness parameter With angle parameters Applied to the illumination controller, the light source output works according to the new set value. Then, under the illumination condition, the angle parameter is used as input to control the stepping rotation structure to set multiple incident angle measurement sequences respectively. On the basis of the three groups of incident conditions, respectively, reducing 2°, keeping the original value, and increasing 2°, the actual light intensity is obtained through the surface light sensor for each group of conditions. (normalized to the range of 0 to 1) and the stone surface reflectivity obtained by the reflectivity detection module , form a one-to-one corresponding data pair between light intensity and reflectivity, and create a temporary array for recording, such as the cargo location The three sets of measurement results are: I=0.79, R=0.842 at 43.55°, I=0.83, R=0.849 at 45.55°, I=0.78, R=0.846 at 47.55°. After establishing the light intensity-reflectivity data relationship, the light intensity distribution adjustment amount is calculated for the measurement point. The method is to compare the light intensity at each angle with the preset target light intensity. =0.80 minus, then multiply by the proportional coefficient Determine the brightness adjustment value. For example, at 45.55°, a difference of 0.03 corresponds to an adjustment value of 27cd / m². If it is a positive value, the light intensity is increased, and if it is a negative value, the light intensity is reduced. At the same time, fine-tune the incident direction. If the absolute value of the reflectivity difference between two adjacent angles is greater than 0.004, take the difference sign to adjust the angle direction. For example If the value is less than the negative threshold of -0.004, the initial angle reference of the next measurement will be shifted to the left by 0.5°. After the adjustment of the light intensity distribution is completed, the high-resolution texture camera is started to collect texture images of the stone surface in the cargo area. The acquired images are quantized into grayscale matrices by pixels. , where each pixel is located at the known spatial coordinate mapping position of the cargo location, the pixel grayscale value is normalized and compared with the light intensity value of the corresponding coordinate, and the light-texture matching coefficient is calculated. This coefficient is the average of the absolute values ​​of the pixel-by-pixel difference between the grayscale normalized value and the light intensity normalized value. For example, in the comparison of 200 randomly selected pixels, the cumulative sum of the absolute values ​​of the grayscale and light differences is 12.4. The average value is taken to obtain a matching coefficient of 0.062. The same calculation is completed by traversing the sampling data of the cargo location, and the final result is written into the light-texture matching coefficient set.

[0039] Table 4: Example table of light intensity and reflectivity measurement and matching coefficient calculation

[0040] As shown in Table 4, the light intensity and reflectivity are measured for three different incident angles of the same location, and the light intensity difference and brightness adjustment value are calculated, and the light texture matching coefficient results obtained from the measurement data are used to measure the corresponding accuracy between light and texture.

[0041] S402: Call the light texture matching coefficient, input the texture image pixel matrix to the three-dimensional reconstruction algorithm, calculate the depth difference value according to the gray value and three-dimensional coordinate point mapping relationship, combine the depth difference value to analyze the three-dimensional point cloud distribution, and then aggregate according to the coordinate adjacency to obtain the surface point cloud density value; Call the light texture matching coefficient, first for each location code, call its matching coefficient value With the collected texture image pixel matrix , the matrix is a two-dimensional array, each element is a normalized gray value and corresponds to the specific spatial position index of the location in the three-dimensional scanning coordinate system , then input the pixel matrix as input into the three-dimensional reconstruction processing flow, first according to the gray value of each pixel and the corresponding space coordinate, call the gray-depth mapping table established during the light source calibration, and match the approximate depth reference with the gray value , then subtract the depth value collected under the condition of adjacent frames or different incident angles with the same coordinate position , to get the depth difference value , for example, the gray value of a certain pixel is 0.65, the corresponding reference depth is 2.812m, and the measured depth is 2.804m, then the difference =-0.008m, record the depth difference value of the pixel to the matrix , then perform combination analysis on the depth difference matrix of the entire texture image, add the depth difference value to the spatial three-dimensional coordinate of each pixel to obtain the modified coordinate point, and form the three-dimensional point cloud data set of the surface of the location by combining the points. Each point in the point cloud set has position coordinates and gray information. After the combination, the point cloud is calculated for coordinate adjacency, and the adjacency is determined by calculating the Euclidean distance between two points. When the distance between any two points is less than the aggregation threshold , it is considered to belong to the same local surface structure. In this embodiment , 0.006m is taken, which is derived from the weighted average of the scanning resolution and noise level of the stone surface microstructure. When traversing the point cloud set, first sort by X coordinate, then calculate the three-dimensional distance with the adjacent points in turn, for example, the distance between point A coordinate and point B coordinate is: ; which is slightly higher than the threshold, so it is not aggregated; while the distance between the coordinate of another point C and the coordinate of the point B is: The points are aggregated into a class of local surface units. After the adjacency judgment of all points is completed and clustering is completed, the number of points in each class is counted. The collective count value is the surface point cloud density data. For example, the surface of a certain cargo location is aggregated to obtain 135 class units with a total number of 5400 points. The average point cloud density is 5400 / 135≈40 points / unit. The complete surface point cloud density value is generated by traversing the cargo location as the input data for the next step of error correction.

[0042] S403: Based on the surface point cloud density value, the point cloud density is compared with the reconstruction error threshold in the error correction calculation, out-of-limit points are screened and the three-dimensional coordinate values ​​are adjusted, the surface feature map is updated, and dynamic three-dimensional model data is generated; Call the surface point cloud density value, first traverse the point cloud density value of each cargo location And the corresponding total number of aggregation units, the value is compared with the preset reconstruction error threshold The threshold value is based on the correlation curve between point cloud density and reconstruction accuracy in multiple batches of different surface texture scanning experiments. =35 points / unit, which means that when the average number of points per local unit is lower than this threshold, the local accuracy of the reconstructed model is insufficient. Less than When the unit is marked as an out-of-limit point, otherwise it is a normal point. Then, for each marked out-of-limit point, all the corresponding three-dimensional coordinates of the point in the point cloud dataset are read. and texture grayscale value , then calculate the coordinate error correction amount, the correction amount The method of determining is to calculate the difference between the coordinate average of the point and its nearest n=5 adjacent points, that is, first calculate the coordinate average of the adjacent points , and then subtract them from the coordinates of the point to get the correction components in the three-dimensional direction. For example, the coordinates of the current point are , the mean of its five neighboring points is , then the correction amount in the X direction is -0.003m, the Y direction is -0.001m, and the Z direction is -0.003m. Multiply these three correction components by the direction adjustment coefficient , , Then add back the original coordinates, where the direction adjustment coefficient is selected based on the scanning system precision calibration result. =0.85, =0.90, =0.80, the final corrected coordinates are , , The operation process is repeated for each out-of-limit point to generate a corrected new three-dimensional coordinate set, and the new coordinates replace the corresponding coordinates in the original point cloud data set. After the replacement is completed, the point cloud and grayscale information of all cargo locations are reorganized to construct a three-dimensional model structure, and the local geometric morphology information is updated in the surface feature data of the model to output dynamic three-dimensional model data. This data retains the mapping relationship with the original texture image and can continue to be used in conjunction with the lighting information.

[0043] See also Figure 6 , the specific steps of S5 are: S501: Extracting coordinate information of inventory counting units and inbound and outbound record tables based on dynamic 3D model data, comparing them based on material numbers, calculating coordinate difference vectors and summarizing them into a unified difference set to generate position deviation values; Call the dynamic 3D model data, first filter the inventory counting units existing in the model, each inventory counting unit contains a unique material number and its spatial coordinates in the current 3D model Then read the original coordinates of the same material number in the inventory record table from the inventory database The original coordinates are the reference position coordinates recorded by the positioning system after the most recent outbound or inbound operation. Each inventory unit is traversed. The material number is first used as the query key to compare the dynamic three-dimensional model with the number in the outbound and inbound table. If they are consistent, the current coordinates and the reference coordinates are extracted to enter the difference calculation step. The difference vector is calculated as the result of subtracting the reference coordinates in the multi-axis direction from the current coordinates. For example, the coordinates of a stone plate material number S-009 in the three-dimensional model are m, the reference coordinate in the in / out record table is m, then the X-axis difference is 12.304-12.297=0.007m, the Y-axis difference is -0.005m, and the Z-axis difference is -0.005m. The difference triplet is used as the coordinate difference vector of the material. After the calculation is completed, the difference vector is stored in the temporary difference set. , and record the material number for summary mapping. Repeat this operation until all inventory units are processed. After the traversal is completed, the modulus of each difference vector is calculated as the position deviation value. The modulus calculation step is to sum the squares of the direction differences and then take the square root. For example, for S-009, the modulus is: ; That is, the deviation value is approximately 9.95 mm. This value directly represents the difference between the actual position of the material and the last record in the warehouse. The deviation values ​​corresponding to the material numbers will be uniformly summarized to form a position deviation value set.

[0044] Table 5: Example table for calculating the difference between inventory counting units and inventory entry and exit records

[0045] As shown in Table 5, the difference vector is obtained by subtracting the current coordinates of the two inventory checking units from the reference coordinates axis by axis, and the modulus is further calculated as the position deviation value. The deviation value directly constitutes the position deviation value set for use in the next step of analysis.

[0046] S502: Call the position deviation value, compare the coordinate difference of the checking unit at different time nodes with the coordinate accuracy reference value, calculate the checking compliance ratio, and perform weighted average analysis on the overall recognition matching degree according to the weight set by the number proportion of the checking unit, and obtain the position accuracy rate; Call the position deviation value set, first read all the material numbers and their corresponding deviation values in the set , and call the coordinate accuracy reference value set in the configuration at the same time. The reference value is derived from the maximum allowable error threshold of the scanning positioning device and the warehouse operation specification. In this embodiment, the value is set to = 0.010 m, which means that when the position deviation value is less than or equal to this value, it is considered to meet the accuracy requirement. Then traverse each checking unit in the set, compare its deviation value with the accuracy reference value, and when , count the material as "compliant", otherwise count as "non-compliant". For example, in the data in Table 5, the deviation value of S-009 is 0.00995 m, and the reference difference value is -0.00005 m, which is compliant. The deviation value of S-010 is 0.00640 m, and the reference difference value is -0.00360 m, which is also compliant. Continue to traverse the entire set to count the number of compliant materials and the total number of materials , calculate the checking compliance ratio . The step is to divide the number of compliant materials by the total number of materials. For example, when the total number is 200 checking units and the number of compliant materials is 184, the ratio is 184 / 200 = 0.92. After completing the ratio calculation, introduce the weight set by the number proportion of the checking unit to adjust the compliance ratio of different types of regions or different important degree storage locations. In this embodiment, the compliance ratio of three types of partitions is weighted: the weight of high priority storage location , the weight of medium priority storage location , and the weight of low priority storage location . The weight is derived from the average turnover rate proportion of this type of storage location in warehouse operation. The rationality of the weight is calculated from the annual in-out warehouse statistics. For example, the compliance ratio of high priority storage location is 0.95, the compliance ratio of medium priority storage location is 0.90, and the compliance ratio of low priority storage location is 0.88. Then the overall recognition matching degree of weighted average is calculated as , first calculate the weighted value of high priority part 0.475, medium priority part 0.27, and low priority part 0.176, then add them to get 0.912, which is the position accuracy, accurate to ±0.001 range, finally output the position accuracy as the overall coordinate accuracy performance indicator of the current inventory period.

[0047] S503: According to the position accuracy and inventory management strategy adjustment benchmark value, judge whether the inventory frequency and warehouse calibration mechanism need to be adjusted, reset the measures in the difference range, and generate an inventory management scheme; Call the position accuracy, first read the benchmark accuracy value set in the inventory management strategy , which comes from the minimum guarantee requirement of the warehouse operation department for the accuracy of the inventory, this embodiment sets , indicating that when the actual position accuracy is lower than this value, the inventory and warehouse management mechanism needs to be adjusted, then calculate the difference between the current accuracy and the benchmark value , for example, the position accuracy obtained in the previous step is , then the difference is 0.921-0.930=-0.009, the difference is negative, indicating that the current accuracy is lower than the benchmark, enter the adjustment strategy process, first judge the inventory frequency (unit: times / week), compare it with the strategy reference frequency , in this example, the reference frequency is 1.0 times / week, if and its absolute value is greater than 0.005, then increase the inventory frequency by one frequency step , the step size comes from the statistical median of the original adjustment record, this example sets it to 0.5 times / week, so the new frequency is the sum of the two, for example, if the current frequency is 1.0 times / week, then adjust it to 1.5 times / week, at the same time, evaluate the sampling ratio of the warehouse calibration mechanism , which represents the proportion of materials to be sampled in each warehouse operation, the initial value in this example is 15%, the reference benchmark ratio , if the accuracy is lower than the benchmark value by more than 0.010, then directly use , if it is lower than the benchmark value but the amplitude is between 0.005 and 0.010, then increase the sampling ratio in a linear proportion, that is, the new sampling ratio is equal to the current sampling ratio plus the product of the coefficient k and the absolute value of the position accuracy difference ∣ΔP∣, the coefficient k comes from the correlation experiment of the pass rate and the sampling rate, in this example, k=2.5 (unit: % / 0.001 accuracy difference), so for the case of ∣ΔP∣=0.009, the sampling ratio is increased by 0.009×2.5≈0.0225, that is, increased by 2.25%, and the new sampling ratio is , after the adjustment calculation of the frequency and the sampling ratio is completed, the measures in the difference range are reset, and the steps are as follows: in the parameter table of the inventory management system, the inventory frequency field is updated to , the warehouse-in and warehouse-out sampling ratio field is updated to , and a change writing strategy adjustment record log is written, the record log including the values before and after the adjustment, the threshold condition triggering the adjustment, the execution timestamp and the operator identification code, finally, the new inventory management strategy file is exported as a data object with the version number sequentially increasing, and a new inventory management scheme is generated corresponding to the adjustment.

[0048] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. The intelligent storage management method of stone panels based on machine vision is characterized by: The following steps are involved: S1: Use multimodal sensors to collect stone surface image data, infrared temperature distribution data, and ultrasonic distance measurement data. Perform grayscale processing on the image data, quantify the temperature data, and calculate the time difference of the distance data to generate a multimodal sensing feature data set. S2: Performing time synchronization calibration based on the multimodal sensor feature data set, inputting the synchronized data into a Kalman filter algorithm for fusion processing, dynamically adjusting the weight coefficient according to the shelf positioning accuracy and storage space identification, and generating comprehensive positioning result data; S3: Based on the comprehensive positioning result data, reflectance coefficient analysis is performed on the stone stacking layer and the cargo location coding area, and the light intensity value is detected. When the reflectance coefficient exceeds a preset reflection threshold, the light source parameters are adjusted to generate an optimized light field parameter set; S4: Recollecting stone surface texture data according to the optimized light field parameter set, inputting the texture data into a three-dimensional reconstruction algorithm for model construction, updating the surface feature map through error correction calculation, and generating dynamic three-dimensional model data.

2. The method for intelligent storage management of stone panels based on machine vision according to claim 1 is characterized in that: The multimodal sensing feature data set includes texture information features, temperature distribution features and depth information features; the comprehensive positioning result data includes position coordinate information, stacking level information and area identification information; the optimized light field parameter set includes light intensity parameters, light source direction parameters and light source frequency parameters; and the dynamic three-dimensional model data includes a geometric structure model, a surface texture model and an error correction model.

3. The intelligent storage management method for stone panels based on machine vision according to claim 1 is characterized in that: The specific steps of S1 are: S101: After obtaining the stone surface image data, the color channel values ​​are sequentially converted into a grayscale value matrix using a pixel grayscale conversion method, the brightness is re-encoded according to the channel ratio of each pixel in the matrix, and the grayscale discrimination of the pixel points is calculated based on the brightness difference to generate a grayscale pixel value set; S102: Retrieving pixel coordinates in the grayscale pixel value set based on the collected infrared temperature distribution data, discretizing the continuous temperature value according to the temperature interval threshold, superimposing the discrete values ​​at the coordinate position, calculating the temperature level by comparing the interval threshold, and obtaining the temperature interval value distribution; S103: Calling the coordinate position in the temperature interval value distribution for the ultrasonic distance measurement data, calculating the distance value based on the time difference between the transmitted and received signals, and integrating it with the temperature distribution value and grayscale pixel value of the corresponding coordinate to generate a multimodal sensing feature data set.

4. The method for intelligent storage management of stone panels based on machine vision according to claim 3 is characterized in that: The specific steps of S2 are: S201: Based on the multimodal sensing feature data set, detecting time record difference values ​​of sensor output signals in the same time period, calculating the offset between sensor recording moments, and performing time reference correction on the data sequence to generate a time synchronization difference sequence; S202: Calling the time synchronization difference sequence and inputting it into the Kalman filter algorithm, weighting it according to the residual between the predicted state and the observed value, adjusting the state transfer matrix and the covariance matrix, updating the filtering process, and generating a fusion positioning estimate; S203: Based on the fused positioning estimate value, the shelf positioning accuracy value and the storage space identification value are compared, the distribution ratio of different weight coefficients in the Kalman filter is adjusted, the corresponding weight values ​​of the position, speed and observation signal are updated, and the comprehensive positioning result data is generated.

5. The method for intelligent storage management of stone panels based on machine vision according to claim 4 is characterized in that: The specific steps of S3 are: S301: traverse the stone stacking levels and cargo location coding areas based on the comprehensive positioning result data, compare the spatial coordinates with the area boundary values, extract the stone surface reflection coefficient, and generate a regional reflection coefficient value set; S302: calling the regional reflectance coefficient value set to detect the light intensity of the cargo location coding area, comparing the reflectance coefficient value with a set reflectance threshold, screening areas exceeding the threshold and recording the light intensity to obtain an over-threshold light intensity value; S303: According to the above-threshold light intensity value, the brightness parameter and the angle parameter of the light source are jointly adjusted, and the distribution optimization is performed according to the range of the light intensity value. The adjusted light source parameters are reallocated to the cargo location coding area to generate an optimized light field parameter set.

6. The intelligent storage management method for stone panels based on machine vision according to claim 5 is characterized in that: The specific steps of S4 are: S401: Obtain the optimized light field parameter set, detect the light intensity values ​​and the stone surface reflectance values ​​at different incident angles, establish a corresponding relationship between the two through polynomial fitting, adjust the light intensity distribution according to the difference, collect stone surface texture image data, and compare the pixel matrix with the light intensity value to generate a light texture matching coefficient; S402: Calling the illumination texture matching coefficient, inputting the texture image pixel matrix into the 3D reconstruction algorithm, calculating the depth difference based on the mapping relationship between the grayscale value and the 3D coordinate point position, combining the depth difference to analyze the 3D point cloud distribution, and then aggregating according to the coordinate adjacency to obtain the surface point cloud density value; S403: Based on the surface point cloud density value, compare the point cloud density with the reconstruction error threshold in the error correction calculation, filter out-of-limit points and adjust the three-dimensional coordinate values, update the surface feature map, and generate dynamic three-dimensional model data.

7. The machine vision-based intelligent storage management method for stone panels according to claim 6 is characterized in that: The illumination texture matching coefficient refers to the coefficient value calculated from the grayscale distribution of the pixel matrix of the stone surface texture image and the corresponding illumination intensity value under specific incident angle and illumination intensity conditions; The surface point cloud density value refers to the ratio of the number of three-dimensional coordinate points determined by the valid point cloud to the area within a unit surface area; The reconstruction error threshold refers to the allowable numerical range of the difference between the depth value of a point obtained by 3D reconstruction and the reference depth value.

8. The method for intelligent storage management of stone panels based on machine vision according to claim 1, characterized in that: The method further comprises step S5: S5: extracting and calculating the position information of the inventory counting unit and the in-and-out record table based on the dynamic three-dimensional model data, calculating the position deviation value, and adjusting the frequency parameter of the dynamic three-dimensional model data update according to the position accuracy to generate an inventory management plan; The inventory management solution includes location accuracy strategy, inventory scheduling strategy and in-and-out inventory control strategy.

9. The machine vision-based intelligent storage management method for stone panels according to claim 8, characterized in that: The specific steps of S5 are: S501: Extracting coordinate information of inventory counting units and inbound and outbound record tables based on the dynamic three-dimensional model data, comparing them according to the corresponding material numbers, calculating coordinate difference vectors and summarizing them into a unified difference set to generate a position deviation value; S502: The position deviation value is called, and the coordinate difference of the inventory units at different time nodes is compared with the coordinate accuracy benchmark value to calculate the inventory compliance ratio. The weight is set according to the proportion of the inventory units, and the weighted average analysis is performed on the overall recognition matching degree to obtain the position accuracy rate. S503: Based on the comparison between the location accuracy and the inventory management strategy adjustment benchmark value, determine whether the inventory frequency and the in-and-out calibration mechanism need to be adjusted, reset the measures within the difference range, and generate an inventory management plan.

10. The machine vision-based intelligent storage management method for stone panels according to claim 9, characterized in that: The position deviation value is the difference vector between the coordinates of the dynamic three-dimensional model data of the inventory unit and the corresponding coordinates of the inventory entry and exit record table; The position accuracy rate is the ratio of the number of inventory units whose position deviation values ​​are within the allowable range to the total number of inventory units after weight processing under the given coordinate accuracy reference value; The inventory management strategy adjustment benchmark value is a preset threshold value based on the location accuracy.

Images