Method and system for detecting material level of unmanned intelligent loading

Abstract

The invention relates to the field of material level detection, in particular to a material level detection method and system for unmanned intelligent loading. Traditional manual vehicle loading and unloading are very low in efficiency, so that a ranging radar is adopted to assist loading and unloading after scientific and technological development, and the loading and unloading method uses the ranging radar for accurate ranging. However, according to the method, the properties of the materials are required to be kept relatively stable, and dust is generated in the charging process or the charging effect becomes poor in severe weather. According to the detection method and system, the dust concentration is analyzed through image recognition by the ranging radar, the dynamic compensation value of data of the ranging radar is adjusted in real time through deep learning, and the change of the hopper material level of different vehicles in the loading process is used as a data set for deep learning; and the ranging radar is used for detecting the material level and correcting the charging speed regularly in the charging process, so that the power consumption of the ranging radar is reduced while the accurate detection of the material level is ensured.

Description

Technical Field

[0001] The present invention relates to the field of material level detection, and in particular to a method and system for material level detection in unmanned intelligent loading. Background Art

[0002] Vehicles that transport materials to and from mines and other locations need to frequently load and unload goods. Most of these goods are bulk solid substances, which have the characteristics of both fluids and solids during the loading and unloading process. Initially, the loading of materials was mostly done manually, which was not only inefficient but also difficult for the loaders to accurately observe the current material level with the naked eye, which made it easy for materials to accumulate unevenly in the vehicle's cargo box, resulting in the vehicle's loading capacity not being fully utilized.

[0003] With the development of science and technology, ranging radar assisted loading and unloading has been adopted. This loading and unloading method uses ranging radar for precise distance measurement, and adjusts the loading speed and loading position in real time according to the measurement results, thereby eliminating the need for manual real-time observation of the material level height in the vehicle, so that materials can be loaded into the truck bed evenly and quickly, improving loading efficiency.

[0004] However, the use of this method requires the properties of the material to remain relatively stable. Once the weather conditions are bad or the material produces a lot of dust during the loading process, the loading accuracy will be affected, resulting in poor loading results.

[0005] Therefore, there is a need for a new unmanned intelligent loading method and system that can detect the material level of the loading in real time under different external environments and adjust the parameters when the hopper is loading. Summary of the invention

[0006] In order to achieve the above object, the present invention proposes a method for material level detection of unmanned intelligent loading, including the following contents:

[0007] A method for material level detection of unmanned intelligent loading, characterized in that it comprises the following steps:

[0008] Use sensors to obtain current weather conditions, and use manual registration or ranging radar to measure the vehicle type and truck bed size;

[0009] Use ranging radar to measure the material level, take images of the truck bed during the loading process, perform image recognition on the captured images, analyze the current dust concentration based on the recognition results, and determine the dynamic compensation value of the current ranging radar data based on humidity and dust concentration;

[0010] The changes in the bucket material level of different vehicles during the loading process are used as a data set for deep learning, so as to obtain the functional relationship between the material level changes of different vehicles and the loading speed and loading position;

[0011] During the loading process, the vehicle is moved to change the loading position and reset the material level. Based on the trained functional relationship between the material level change of different buckets and the loading position, the loading speed at the next moment is determined and whether the vehicle should be moved.

[0012] When the initial height of the material level after resetting approaches the allowable loading height of the vehicle to a certain extent, loading is stopped. After loading is stopped, the parameters of the function used for loading are corrected according to the loading effect, and the relationship between the loading speed and the vehicle position over time during the loading process is added to the training set for deep learning of the functional relationship between the material level change of different truck buckets and the loading speed and loading position.

[0013] Furthermore, the sensor is used to obtain the weather conditions of the surrounding environment and adjust the predicted changing trend of the weather conditions according to the weather forecast. The truck bucket size includes: the length of the truck bucket, the width of the truck bucket and the height of the truck bucket. The ranging radar confirms the allowable loading height and the interval to be loaded of the truck bucket through laser ranging before loading begins, and thereby determines the loading position.

[0014] Furthermore, the principle of adjusting the ranging radar is:

[0015] The ranging radar detects the material level through laser, obtains the current weather conditions based on the weather conditions obtained by the sensor and the weather forecast, and predicts the weather conditions in the next period of time, and adjusts the results of the ranging radar based on the predicted weather conditions and the humidity and dust concentration during the ranging process.

[0016] Furthermore, the image recognition is realized by photographing a top view of the loading position, and the steps of the image recognition are:

[0017] S1. Obtain RGB data of the material currently loaded, and use a camera to take a top view of the bucket;

[0018] S2. Determine the position of the loading position corresponding to the top view of the bucket and mark it;

[0019] S2. Extract the RGB data of the marked loading position, calculate the difference between the RGB data of the loading area and the RGB data of the material, and obtain the RGB data of the dust concentration in the loading position;

[0020] S3. Convert the RGB data of the dust concentration into HSV data of the dust concentration and perform quantitative analysis to obtain the dust concentration at the loading position at the current moment, obtain the blocking effect of the dust concentration on the ranging radar, and adjust the dynamic compensation value of the ranging radar data in real time.

[0021] Furthermore, the specific steps of using the RGB data of the loading position and the RGB data of the material to obtain the dust concentration are:

[0022] The RGB data of the material is extracted from the color image of the material. An arbitrary point in the center of the position corresponding to the loading position in the overhead view of the truck bucket is taken as the origin, and the origin is taken as the center. A quarter of the length of the distance from the origin to the edge of the truck bucket image is taken as the basic unit. A two-dimensional rectangular coordinate system is established on the truck bucket image with the origin as the center. Points on the coordinate system including the origin are taken as relevant points at fixed intervals on the two-dimensional rectangular coordinate system. The straight-line distance between adjacent relevant points is defined as a basic unit. The brightness data of all relevant points are extracted according to the HSV data of the truck bucket image obtained after conversion, and the weighted sum of the brightness data of all relevant points is taken as the brightness data of the truck bucket image. The calculation formula of the brightness data is:

[0023]

[0024] Among them, L is the brightness data of the truck bed image, l0 is the brightness data of the truck bed image at the origin, u is the normalization parameter, l i is the brightness data at all relevant points except the origin, c i For l i The number of related points between the corresponding related point and the origin, m is the brightness data of the material extracted from the HSV data of the material;

[0025] Afterwards, the brightness data is subtracted from the initial brightness data in the bucket image before loading begins to obtain a change value of the brightness data, and the dust concentration can be obtained by subjecting the change value of the brightness data to parameter transformation.

[0026] Furthermore, when the ranging radar detects that the current material level reaches the allowable loading height, the vehicle moves forward to reset the loading position, and the material level is reset accordingly. The initial height of the material level after the material level is reset is related to the loading position at different times. The calculation formula of the initial material level is:

[0027]

[0028] Among them, H0 is the ratio of the initial material level after the next vehicle movement reset to the allowable loading height, a i is the weight coefficient. When H0>0.955, it means that the loading is completed. L is the initial material level after each vehicle movement. i is the correlation coefficient set in advance, which determines the accuracy and calculation speed of the progress function. t represents the time after the vehicle movement is completed.

[0029] After calculating the reset initial material level, the ranging radar calculates the loading speed and loading time according to the difference between the allowable loading height and the reset initial material level.

[0030] Furthermore, the training set used by the deep learning model includes the change of the material level over time when different vehicles transport different materials, the loading speed, and the distance and time of the vehicle movement, and through deep learning, a functional relationship between the loading speed and the material level within the time of two adjacent vehicle movements is established, and a relationship between the time and distance of the vehicle movement and the size of the truck bucket is established. Finally, the functional relationship between the loading speed and the material level is linked to the relationship between the time and distance of the vehicle movement and the size of the truck bucket for secondary deep learning, and the functional relationship between the material level change and the loading speed and loading position for different vehicles is obtained;

[0031] The formula that expresses the functional relationship between the material level change and the loading speed and loading position is:

[0032]

[0033] Among them, c is a fixed coefficient, t represents different moments in the loading process, v represents the loading speed of the loading device at different moments, T represents the total time spent on loading, H0 is the initial material level after reset after the next vehicle movement, h0 is the allowable loading height, and H is the current loading height.

[0034] Furthermore, the method for correcting the corresponding parameters according to the charging effect is:

[0035] According to the functional relationship between the material level change, the loading speed and the loading position actually generated during this loading process, the parameters used in the functional relationship between the material level change, the loading speed and the loading position predicted by the deep learning model are corrected. The correction formula is:

[0036]

[0037] Among them, L t represents the actual loading speed, loading position and material level at time t, a t0 The functional relationship between the loading speed and loading position before correction and L t The corresponding parameter, a t is the functional relationship between the corrected loading speed and loading position and L t The corresponding parameter, K t Represents the loading speed, loading position and material level obtained from the functional relationship of the loading speed and loading position predicted by the depth model at time t.

[0038] Furthermore, the ranging radar detects the material level at a fixed frequency during the loading process, and uses the detected material level to make real-time corrections to the prediction of the material level based on the functional relationship between the material level change and the loading position.

[0039] Furthermore, a material level detection system for unmanned intelligent loading is characterized by comprising:

[0040] Calibration module: using sensors to obtain the current weather conditions, using manual registration and ranging radar to measure the vehicle type and truck bed size, and adjusting the ranging radar according to the weather conditions;

[0041] Learning module: taking the changes of the material level in the bucket of different vehicles during the loading process as a data set for deep learning, so as to obtain the functional relationship between the changes of the material level in different buckets and the loading position;

[0042] Loading module: loading according to the functional relationship between the material level change of different vehicles and the loading position, and regularly obtaining the current material level height through the ranging radar;

[0043] Identification module: during the loading process, the image of the loading position is acquired, the dust concentration during the loading process is identified and analyzed, and the dynamic compensation value of the ranging radar data is adjusted according to the current dust concentration and ambient humidity, so that the ranging radar can accurately measure the material level;

[0044] The feedback module amends the detection function according to the progress function generated by the loading progress at each moment in the loading process, and sends the loading data generated in the loading process to the loading data set of the training module for training the training module.

[0045] Beneficial effects:

[0046] 1. The current dust concentration is calculated by identifying the image taken from the bucket of the loading vehicle, and the detection result of the ranging radar is corrected according to the dust concentration, thereby improving the ranging accuracy of the ranging radar. At the same time, the deep model is used for training to obtain the functional relationship between the material level change and the loading speed and loading position, thereby improving the versatility of the detection method;

[0047] 2. Add weather factors to the factors to be considered, and correct the ranging results according to weather conditions, ambient humidity and dust concentration during the loading process to avoid errors caused by the ranging radar in the ranging work; during the loading process, predict the material level at the next moment and adjust the loading speed according to the function trained by deep learning, and correct the material level height data in real time to correct the loading speed, thereby ensuring the accuracy of loading while improving the loading efficiency and indirectly reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is a flow chart of the material level detection method of the present invention;

[0049] Figure 2 It is the architecture diagram of the material level detection system of the present invention. DETAILED DESCRIPTION

[0050] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0051] The present invention provides a material level detection method for unmanned intelligent loading, which includes the following contents:

[0052] One of the advantages of the material level detection method of the present invention is its versatility. The loading data of the loading process is added to the subsequent calculation to ensure that the material level detection can be close to the actual situation. In order to be close to the actual situation, the data collected include: the current state of the surrounding environment, the appearance characteristics of the materials to be loaded and the maximum material level of the loading vehicle; the present invention identifies the material level through the captured image, and the maximum material level of the loading affects the subsequent recognition and calculation of the material level, so the maximum material level that the loading vehicle can load is recorded in advance; the appearance characteristics of the material affect the recognition of the material position in the captured image, so the appearance characteristics of the material need to be collected in advance; the state of the surrounding environment includes: the temperature, humidity and current time of the surrounding environment, these data will affect the loading speed, and cause the properties of the material and the color tone in the image to change, these data are collected and corrections are made to the subsequent recognition and calculation based on these data.

[0053] like Figure 1 As shown, a method for material level detection of unmanned intelligent loading is characterized by comprising the following steps:

[0054] Use sensors to obtain current weather conditions, and use manual registration or ranging radar to measure the vehicle type and truck bed size;

[0055] Use ranging radar to measure the material level, take a vertical picture of the truck bed during the loading process, perform image recognition on the captured image, analyze the current dust concentration based on the recognition result, and determine the dynamic compensation value of the current ranging radar data based on humidity and dust concentration;

[0056] The changes in the bucket material level of different vehicles during the loading process are used as a data set for deep learning, so as to obtain the functional relationship between the material level changes of different vehicles and the loading speed and loading position;

[0057] During the loading process, the vehicle is moved to change the loading position and reset the material level. Based on the trained functional relationship between the material level change of different buckets and the loading position, the loading speed at the next moment is determined and whether the vehicle should be moved.

[0058] When the initial height of the material level after resetting approaches the allowable loading height of the vehicle to a certain extent, loading is stopped. After loading is stopped, the parameters of the function used for loading are corrected according to the loading effect, and the relationship between the loading speed and the vehicle position over time during the loading process is added to the training set for deep learning of the functional relationship between the material level change of different truck buckets and the loading speed and loading position.

[0059] The sensor is used to obtain the weather conditions of the surrounding environment and adjust the forecast of the changing trend of the weather conditions according to the weather forecast. The dimensions of the truck bucket include: the length of the truck bucket, the width of the truck bucket and the height of the truck bucket. Before loading, the ranging radar confirms the allowable loading height and the interval to be loaded of the truck bucket through laser ranging, and thus determines the loading position. The loading position is located in the center area of ​​the truck bucket. Through the movement of the vehicle, the loading device can evenly dump the materials into the truck bucket in a certain order.

[0060] The principle of adjusting the ranging radar is:

[0061] The ranging radar detects the material level through laser, obtains the current weather conditions based on the weather conditions obtained by the sensor and the weather forecast, and predicts the weather conditions in the next period of time, and adjusts the results of the ranging radar based on the predicted weather conditions and the humidity and dust concentration during the ranging process.

[0062] Image recognition is achieved by taking a top view of the loading position. The steps of image recognition are:

[0063] S1. Obtain the RGB data of the currently loaded material and use a camera to take a top view of the bucket;

[0064] S2. Determine the loading position corresponding to the position in the top view of the bucket and mark it;

[0065] S3 extracts the RGB data of the marked loading position, calculates the difference between the RGB data of the loading area and the RGB data of the material, and obtains the RGB data of the dust concentration in the loading position;

[0066] S4. Convert the RGB data of dust concentration into HSV data of dust concentration and perform quantitative analysis to obtain the dust concentration at the current loading position, obtain the blocking effect of dust concentration on the ranging radar, and adjust the dynamic compensation value of the ranging radar data in real time.

[0067] The specific steps to use the RGB data of the loading position and the RGB data of the material to obtain the dust concentration are:

[0068] The RGB data of the material is extracted from the color image of the material. An arbitrary point in the center of the position corresponding to the loading position in the overhead view of the truck bed is taken as the origin. With the origin as the center, one quarter of the length of the distance from the origin to the edge of the truck bed image is taken as the basic unit. A two-dimensional rectangular coordinate system is established on the truck bed image with the origin as the center. Points on the coordinate system including the origin are taken as relevant points at fixed intervals on the two-dimensional rectangular coordinate system. The straight-line distance between adjacent relevant points is defined as a basic unit. The brightness data of all relevant points is extracted based on the HSV data of the truck bed image obtained after conversion, and the weighted sum of the brightness data of all relevant points is taken as the brightness data of the truck bed image. The calculation formula of the brightness data is:

[0069]

[0070] Among them, L is the brightness data of the truck bed image, l0 is the brightness data of the truck bed image at the origin, u is the normalization parameter, l i is the brightness data at all relevant points except the origin, c i For l i The number of related points between the corresponding related point and the origin, m is the brightness data of the material extracted from the HSV data of the material;

[0071] The formula for calculating the initial brightness data in the bucket image before loading begins is:

[0072]

[0073] Where L0 is the initial brightness of the truck bed image, n is the number of regions of equal size divided into the truck bed image, and q j is the brightness data of each segmented area, and m is the brightness data of the material extracted from the HSV data of the material;

[0074] After that, the brightness data is subtracted from the initial brightness data in the bucket image before loading to obtain the change value of the brightness data. The change value of the brightness data is linearly related to the shading effect of the dust, and the shading effect of the dust is positively correlated with the dust concentration. Therefore, the relationship between the change value of the brightness data and the dust concentration is:

[0075] D=f(α*L c )

[0076] Where D is the dust concentration, α is the linear parameter, L c is the change value of brightness data, and f is the functional relationship between the shading effect and the dust concentration.

[0077] The dust concentration can be obtained by transforming the change value of the brightness data through parameters.

[0078] When the ranging radar detects that the current material level reaches the allowable loading height, the vehicle moves forward to change the loading position, and the material level is reset accordingly. The initial height of the material level after the material level is reset is related to the loading position at different times. The calculation formula for the initial material level is:

[0079]

[0080] Among them, H0 is the ratio of the initial material level after the next vehicle movement reset to the allowable loading height, a i is the weight coefficient. When H0>0.955, it means that the loading is completed. L is the initial material level after each vehicle movement. i is the correlation coefficient set in advance, which determines the accuracy and calculation speed of the progress function. t represents the time after the vehicle movement is completed.

[0081] After calculating the initial material level after reset, the ranging radar calculates the loading speed and loading time according to the difference between the allowable loading height and the initial material level after reset.

[0082] The training set used by the deep learning model includes the changes in material level over time when different vehicles transport different materials, the loading speed, and the distance and time of vehicle movement. Through deep learning, the functional relationship between the loading speed and the material level within the time of two consecutive vehicle movements is established, and the relationship between the time and distance of vehicle movement and the size of the truck bucket is established. Finally, the functional relationship between the loading speed and the material level is linked to the relationship between the time and distance of vehicle movement and the size of the truck bucket for secondary deep learning, and the functional relationship between the material level change and the loading speed and loading position for different vehicles is obtained.

[0083] The formula that expresses the functional relationship between the material level change and the loading speed and loading position is:

[0084]

[0085] Among them, c is a fixed coefficient, t represents different moments in the loading process, v represents the loading speed of the loading device at different moments, T represents the total time spent on loading, H0 is the initial material level after reset after the next vehicle movement, h0 is the allowable loading height, and H is the current loading height.

[0086] The function of the fixed coefficient is to quickly adjust the current loading speed by changing the fixed coefficient after the current actual material level is detected by the ranging radar during the loading process.

[0087] For the same image, the conversion into HSV data occupies less memory than RGB data, has a faster processing speed, and can more intuitively reflect the smoke and blur of dust in the image, so that the concentration of dust in the image can be accurately identified at a faster speed.

[0088] The method to correct the corresponding parameters according to the charging effect is:

[0089] According to the functional relationship between the material level change, the loading speed and the loading position actually generated during this loading process, the parameters used in the functional relationship between the material level change, the loading speed and the loading position predicted by the deep learning model are corrected. The correction formula is:

[0090]

[0091] Among them, L t represents the actual loading speed, loading position and material level at time t, a t0 The functional relationship between the loading speed and loading position before correction and L t The corresponding parameter, a t is the functional relationship between the corrected loading speed and loading position and L t The corresponding parameter, K t Represents the loading speed, loading position and material level obtained from the functional relationship of the loading speed and loading position predicted by the depth model at time t.

[0092] Due to the characteristics of the loading process, the changes in the material level and loading speed during the loading process will be affected by the material level before the last vehicle movement. Therefore, when correcting the parameters, the parameters when loading at adjacent positions should be considered.

[0093] The ranging radar detects the material level at a fixed period during the loading process, and uses the detected material level to make real-time corrections to the material level prediction based on the functional relationship between the material level change and the loading position, thereby adjusting the current loading speed.

[0094] like Figure 2 As shown, an unmanned intelligent loading material level detection system is characterized by comprising:

[0095] Calibration module: Use sensors to obtain current weather conditions, use manual registration and ranging radar to measure vehicle type and truck bed size, and adjust ranging radar according to weather conditions;

[0096] Learning module: The changes in the bucket material level of different vehicles during the loading process are used as a data set for deep learning, so as to obtain the functional relationship between the changes in the bucket material level of different vehicles and the loading position;

[0097] Loading module: Loading is performed based on the functional relationship between the material level changes of different vehicles and the loading position, and the current material level height is regularly obtained through the ranging radar;

[0098] Identification module: It acquires the image of the loading position during the loading process, identifies and analyzes the dust concentration during the loading process, and adjusts the dynamic compensation value of the ranging radar data according to the current dust concentration and ambient humidity, so that the ranging radar can accurately measure the material level;

[0099] The feedback module amends the detection function according to the progress function generated by the loading progress at each moment in the loading process, and sends the loading data generated in the loading process to the loading data set of the training module for training the training module.

[0100] The above-mentioned embodiments only express the preferred implementation modes of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present invention. It should be pointed out that, for a person skilled in the art, several modifications, improvements and substitutions can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the present invention patent shall be subject to the attached claims.

Claims

1. A method for material level detection in unmanned intelligent loading, characterized in that: The following steps are involved: Use sensors to obtain current weather conditions, and use manual registration or ranging radar to measure the vehicle type and truck bed size; Use ranging radar to measure the material level, take images of the truck bed during the loading process, perform image recognition on the captured images, analyze the current dust concentration based on the recognition results, and determine the dynamic compensation value of the current ranging radar data based on humidity and dust concentration; The changes in the bucket material level of different vehicles during the loading process are used as a data set for deep learning, so as to obtain the functional relationship between the material level changes of different vehicles and the loading speed and loading position; During the loading process, the vehicle is moved to change the loading position and reset the material level. Based on the trained functional relationship between the material level change of different buckets and the loading position, the loading speed at the next moment is determined and whether the vehicle should be moved. When the initial height of the material level after resetting approaches the allowable loading height of the vehicle to a certain extent, loading is stopped. After loading is stopped, the parameters of the function used for loading are corrected according to the loading effect, and the relationship between the loading speed and the vehicle position over time during the loading process is added to the training set for deep learning of the functional relationship between the material level change of different truck buckets and the loading speed and loading position.

2. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: The sensor is used to obtain the weather conditions of the surrounding environment and adjust the prediction of the changing trend of the weather conditions according to the weather forecast. The truck bucket size includes: the length of the truck bucket, the width of the truck bucket and the height of the truck bucket. The ranging radar confirms the allowable loading height and the interval to be loaded of the truck bucket through laser ranging before loading begins, and thereby determines the loading position.

3. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: The principle of adjusting the ranging radar is as follows: The ranging radar detects the material level through laser, obtains the current weather conditions based on the weather conditions obtained by the sensor and the weather forecast, and predicts the weather conditions in the next period of time, and adjusts the results of the ranging radar based on the predicted weather conditions and the humidity and dust concentration during the ranging process.

4. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: The image recognition is realized by photographing a top view of the loading position, and the steps of the image recognition are: S1. Obtain RGB data of the material currently loaded, and use a camera to take a top view of the bucket; S2. Determine the position of the loading position corresponding to the top view of the bucket and mark it; S3. Extract the RGB data of the marked loading position, calculate the difference between the RGB data of the loading area and the RGB data of the material, and obtain the RGB data of the dust concentration in the loading position; S4. Convert the RGB data of the dust concentration into the HSV data of the dust concentration and perform quantitative analysis to obtain the dust concentration at the loading position at the current moment, obtain the blocking effect of the dust concentration on the ranging radar, and adjust the dynamic compensation value of the ranging radar data in real time.

5. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: The specific steps of using the RGB data of the loading position and the RGB data of the material to obtain the dust concentration are: The RGB data of the material is extracted from the color image of the material. An arbitrary point in the center of the position corresponding to the loading position in the overhead view of the truck bucket is taken as the origin, and the origin is taken as the center. A quarter of the length of the distance from the origin to the edge of the truck bucket image is taken as the basic unit. A two-dimensional rectangular coordinate system is established on the truck bucket image with the origin as the center. Points on the coordinate system including the origin are taken as relevant points at fixed intervals on the two-dimensional rectangular coordinate system. The straight-line distance between adjacent relevant points is defined as a basic unit. The brightness data of all relevant points are extracted according to the HSV data of the truck bucket image obtained after conversion, and the weighted sum of the brightness data of all relevant points is taken as the brightness data of the truck bucket image. The calculation formula of the brightness data is: Among them, L is the brightness data of the truck bed image, l0 is the brightness data of the truck bed image at the origin, u is the normalization parameter, l i is the brightness data at all relevant points except the origin, c i For l i The number of related points between the corresponding related point and the origin, m is the brightness data of the material extracted from the HSV data of the material; The change value of the brightness data can be obtained by subtracting the brightness data from the initial brightness data in the bucket image before loading begins, and the dust concentration can be obtained by subjecting the change value of the brightness data to parameter transformation.

6. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: When the ranging radar detects that the current material level reaches the allowable loading height, the vehicle moves forward to change the loading position, and the material level is reset accordingly. The initial height of the material level after the material level is reset is related to the loading position at different times. The calculation formula of the initial material level is: Among them, H0 is the ratio of the initial material level after the next vehicle movement reset to the allowable loading height, a i is the weight coefficient. When H0>0.955, it means that the loading is completed. L is the initial material level after each vehicle movement. i is the correlation coefficient set in advance, which determines the accuracy and calculation speed of the progress function. t represents the time after the vehicle movement is completed. After calculating the reset initial material level, the ranging radar calculates the loading speed and loading time according to the difference between the allowable loading height and the reset initial material level.

7. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: The training set used by the deep learning model includes the change of the material level over time when different vehicles transport different materials, the loading speed, and the distance and time of the vehicle movement, and establishes a functional relationship between the loading speed and the material level within the time of two adjacent vehicle movements through deep learning, and establishes a relationship between the time and distance of the vehicle movement and the size of the truck bucket, and finally links the functional relationship between the loading speed and the material level with the relationship between the time and distance of the vehicle movement and the size of the truck bucket for secondary deep learning, and obtains the functional relationship between the material level change and the loading speed and loading position for different vehicles; The formula that expresses the functional relationship between the material level change and the loading speed and loading position is: Among them, c is a fixed coefficient, t represents different moments in the loading process, v represents the loading speed of the loading device at different moments, T represents the total time spent on loading, H0 is the initial material level after reset after the next vehicle movement, h0 is the allowable loading height, and H is the current loading height.

8. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: The method for correcting the corresponding parameters according to the charging effect is: According to the functional relationship between the material level change, the loading speed and the loading position actually generated during this loading process, the parameters used in the functional relationship between the material level change, the loading speed and the loading position predicted by the deep learning model are corrected. The correction formula is: Among them, L t represents the actual loading speed, loading position and material level at time t, a t0 The functional relationship between the loading speed and loading position before correction and L t The corresponding parameter, a t is the functional relationship between the corrected loading speed and loading position and L t The corresponding parameter, K t Represents the loading speed, loading position and material level obtained from the functional relationship of the loading speed and loading position predicted by the depth model at time t.

9. The method for material level detection of unmanned intelligent loading according to claim 1 is characterized in that: The ranging radar detects the material level at a fixed frequency during the loading process, and uses the detected material level to make real-time corrections to the prediction of the material level based on the functional relationship between the material level change and the loading position.

10. The material level detection system for unmanned intelligent loading according to any one of claims 1 to 9, characterized in that: Included are: Calibration module: using sensors to obtain the current weather conditions, using manual registration and ranging radar to measure the vehicle type and truck bed size, and adjusting the ranging radar according to the weather conditions; Learning module: taking the changes of the material level in the bucket of different vehicles during the loading process as a data set for deep learning, so as to obtain the functional relationship between the changes of the material level in different buckets and the loading position; Loading module: loading according to the functional relationship between the material level change of different vehicles and the loading position, and regularly obtaining the current material level height through the ranging radar; Identification module: during the loading process, the image of the loading position is acquired, the dust concentration during the loading process is identified and analyzed, and the dynamic compensation value of the ranging radar data is adjusted according to the current dust concentration and ambient humidity, so that the ranging radar can accurately measure the material level; The feedback module amends the detection function according to the progress function generated by the loading progress at each moment in the loading process, and sends the loading data generated in the loading process to the loading data set of the training module for training the training module.

Images