Object volume calculation method, device, equipment, and storage medium
By acquiring and processing key points and connection relationships in object images, calculating the scale ratio of objects and three-dimensional point cloud information, the problem of inaccurate volume measurement in the logistics industry in the prior art is solved, and more efficient and accurate object volume calculation is achieved .
Patent Information
- Application Number
- CN201910573335.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2019-06-28
- Publication Date
- 2024-11-29
- Estimated Expiration
- 2039-06-28
AI Technical Summary
The intelligent software calculation methods used in volume measurement in the existing logistics industry are not accurate enough, resulting in low accuracy in cargo volume measurement.
By obtaining multi-frame object images, extracting object key points and connection relationships, calculating the object's scale ratio and three-dimensional point cloud information, combining this information to calculate the object's volume, using Tiny-DSOD network and CPM network for convolution operations and feature extraction, Convert data types to improve computational efficiency.
It improves the accuracy and efficiency of object volume calculation, and can restore the volume of the object more accurately.
Smart Images

Figure CN112150533B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of data processing technology, and in particular to a method, device, equipment and storage medium for calculating the volume of an object. Background Art
[0002] The current logistics industry is in the process of transforming towards intelligence and precision. How to obtain accurate information more efficiently at a lower cost has become an urgent need in the industry.
[0003] In the field of logistics, the weight and volume of goods are the most important billing data. For a long time, the logistics industry has still used the primitive tape measure method for volume measurement. There are some intelligent software solutions at present, but the existing intelligent software solutions generally calculate the volume of the real goods by directly using the reflected length signal, width signal and height signal of the goods as the length signal, width signal and height signal of the goods, which results in low accuracy in the volume measurement of the goods. Summary of the invention
[0004] The embodiments of the present application provide a method, device, equipment and storage medium for calculating the volume of an object, which can improve the accuracy of calculating the volume of an object.
[0005] In a first aspect, an embodiment of the present application provides a method for calculating the volume of an object, comprising:
[0006] Obtain multiple frames of object images, and obtain the object key points and connection relationships in each frame of the object image;
[0007] Calculating a scale ratio of the object according to key points of the multiple frames of object images;
[0008] Acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images;
[0009] The volume of the object is calculated according to the scale ratio, the connection relationship and the three-dimensional point cloud information.
[0010] In some implementations, obtaining key points and connection relationships of objects in each frame of the object image includes:
[0011] Obtaining a trained Tiny-DSOD network and a CPM network, wherein the Tiny-DSOD network and the CPM network are trained based on sample images of object key points and their connection relationships that are labeled in sequence;
[0012] Performing a convolution operation on the object image through the trained Tiny-DSOD network to obtain a feature map;
[0013] The key points and connection relationships of the object are extracted based on the feature graph through the CPM network.
[0014] In some embodiments, before performing a convolution operation on the object image through the trained Tiny-DSOD network to obtain a feature map, the method further includes:
[0015] Converting the 32-bit binary floating-point data in the Tiny-DSOD network into 8-bit binary integer data to obtain a converted Tiny-DSOD network;
[0016] The transformed Tiny-DSOD network is used as the trained Tiny-DSOD network.
[0017] In some implementations, calculating the volume of the object according to the scale ratio, the connection relationship, and the three-dimensional point cloud information includes:
[0018] Acquire the side length of the object in multiple frames of object images according to the three-dimensional point cloud information and the connection relationship;
[0019] The volume of the object in the multiple frames of object images is calculated according to the scale ratio and the side length of the object in the multiple frames of object images.
[0020] In some implementations, after acquiring multiple frames of object images, the method further includes:
[0021] Get the preset image size;
[0022] Scaling the multiple frames of object images according to the preset image size to obtain scaled multiple frames of object images;
[0023] The obtaining of object key points and connection relationships in each frame of the object image includes:
[0024] Get the object key points and connection relationships in each frame of the zoomed object image.
[0025] In some implementations, acquiring multiple frames of object images includes:
[0026] Acquire a video including an image of an object;
[0027] Converting the analog signal in the video into a compressed image signal;
[0028] The compressed image signal is decompressed to obtain multiple frames of object images in the video.
[0029] In some implementations, calculating the scale ratio of the object according to the key points of the multiple frames of object images includes:
[0030] Obtaining the first acceleration corresponding to each moment of the captured object image;
[0031] Calculating a second acceleration based on the key points of the multiple frames of object images and the time when the object images were collected;
[0032] The first acceleration and the second acceleration are matched based on a time, and a scale ratio is calculated according to the matched first acceleration and the second acceleration.
[0033] In a second aspect, an embodiment of the present application further provides an object volume calculation device, including:
[0034] A first acquisition unit is used to acquire multiple frames of object images, and acquire key points and connection relationships of the objects in each frame of the object images;
[0035] A first calculation unit, configured to calculate a scale ratio of the object according to key points of the multiple frames of object images;
[0036] A second acquisition unit, configured to acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images;
[0037] The second calculation unit is used to calculate the volume of the object according to the scale ratio, the connection relationship and the three-dimensional point cloud information.
[0038] In some implementations, the first acquisition unit includes:
[0039] A first acquisition subunit is used to acquire a trained Tiny-DSOD network and a CPM network, wherein the Tiny-DSOD network and the CPM network are trained based on sample images of object key points and their connection relationships that are labeled in sequence;
[0040] A convolution subunit, used to perform a convolution operation on the object image through the trained Tiny-DSOD network to obtain a feature map;
[0041] The extraction subunit is used to extract the key points and connection relationships of the object based on the feature map through the CPM network.
[0042] In some implementations, the first acquisition unit further includes:
[0043] The transformation subunit is used to transform the 32-bit binary floating point data in the Tiny-DSOD network into 8-bit binary integer data to obtain a transformed Tiny-DSOD network; and use the transformed Tiny-DSOD network as the trained Tiny-DSOD network.
[0044] In some implementations, the second computing unit includes:
[0045] A second acquisition subunit is used to acquire the side length of the object in the multiple frames of object images according to the three-dimensional point cloud information and the connection relationship;
[0046] The first calculation subunit is used to calculate the volume of the object in the multiple frames of object images according to the scale ratio and the side length of the object in the multiple frames of object images.
[0047] In some implementations, the first acquisition unit further includes:
[0048] A third acquisition subunit, used to acquire a preset image size;
[0049] A scaling subunit, used for scaling the multiple frames of object images according to the preset image size to obtain scaled multiple frames of object images;
[0050] The acquisition unit further includes:
[0051] The fourth acquisition subunit is used to acquire the key points and connection relationships of the objects in each frame of the zoomed object image.
[0052] In some embodiments, the acquisition unit further includes:
[0053] a fifth acquisition subunit, configured to acquire a video including an image of the object;
[0054] Converting the analog signal in the video into a compressed image signal;
[0055] The decompression subunit is used to decompress the image signal in a compressed state to obtain multiple frames of object images in the video.
[0056] In some implementations, the first computing unit includes:
[0057] A sixth acquisition subunit, used to acquire the first acceleration corresponding to each moment of the captured object image;
[0058] The second calculation subunit is used to calculate the second acceleration based on the key points of the multiple frames of object images and the time when the object images are collected; match the first acceleration and the second acceleration based on the time, and calculate the scale ratio according to the matched first acceleration and the second acceleration.
[0059] In a third aspect, an embodiment of the present application further provides a device, including a processor and a memory, wherein the memory stores program code, and when the processor calls the program code in the memory, the method for calculating the object volume as described above is executed.
[0060] In a fourth aspect, an embodiment of the present application further provides a storage medium, wherein the storage medium stores a plurality of instructions, wherein the instructions are suitable for loading by a processor to execute the object volume calculation method provided in the embodiment of the present application.
[0061] The embodiment of the present application obtains multiple frames of object images, and obtains the key points and connection relationships of the object in each frame of the object image; calculates the scale ratio of the object based on the key points of the multiple frames of object images; obtains the three-dimensional point cloud information of the object based on the key points of the multiple frames of object images; and calculates the volume of the object based on the scale ratio, connection relationship and three-dimensional point cloud information. This solution calculates the scale ratio of the object through multiple frames of object images continuously shot by the shooting module in the terminal, thereby accurately restoring the volume of the object, extracts the key points of the object from each frame of the object image, obtains the key points and connection relationships of the object in each frame of the object image, obtains the three-dimensional point cloud information of the key points of the object based on the key points of the object, and calculates the volume of the object based on the scale ratio, connection relationship and three-dimensional point cloud information, thereby accurately restoring the volume of the object; improving the accuracy of calculating the volume of the object. BRIEF DESCRIPTION OF THE DRAWINGS
[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.
[0063] Figure 1 is a schematic diagram of a flow chart of object volume calculation provided in an embodiment of the present application;
[0064] Figure 2 This is another flow chart of the object volume calculation method provided in the embodiment of the present application.
[0065] Figure 3 is a schematic diagram of the structure of an object volume calculation device provided in an embodiment of the present application;
[0066] Figure 4 It is a schematic diagram of the structure of the device provided in the embodiment of the present application. DETAILED DESCRIPTION
[0067] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.
[0068] See also Figure 1 , Figure 1 : is a flow chart of an object volume calculation method provided by an embodiment of the present application. The execution subject of the object volume calculation method may be the object volume calculation device provided by the embodiment of the present application, or a device integrating the object volume calculation device, such as a terminal or a server, etc., wherein the object volume calculation method may be implemented in hardware or software, and the device may be a smart phone, tablet computer, PDA, or laptop computer, fixed computer, server, etc. equipped with a camera head and an IMU (Inertial measurement unit). The object volume calculation method may include:
[0069] S101, obtaining multiple frames of object images, and obtaining object key points and connection relationships in each frame of the object image.
[0070] Specifically, this embodiment can be applied to the process of calculating the volume of goods, boxes, etc. When used to calculate the volume of the box, multiple frames of object images including the box can be obtained through the shooting module in the terminal. Further, in order to improve the accuracy of the subsequent calculation of the scale ratio, multiple frames of object images including the box can be obtained by shooting the box video through the shooting module in the terminal, and then parsing it. Then obtain the object key points and connection relationships in each frame of the object image. The key points of the objects contained in each frame of the acquired multiple frames of object images are extracted to obtain the key points corresponding to each frame of the object image. Specifically, the key points of the object are extracted for each frame of the object image through a trained network, and the connection relationship is obtained, such as a target detection network and a posture estimation network. The target detection network and the posture estimation network are trained based on sample images of object key points labeled in sequence; the object image is convolved through the target detection network to obtain a feature map; the object key points and connection relationships are extracted based on the feature map through the posture estimation network.
[0071] Specifically, step S101 includes:
[0072] Obtaining a trained Tiny-DSOD network and a CPM network, wherein the Tiny-DSOD network and the CPM network are trained based on sample images of object key points and their connection relationships that are labeled in sequence;
[0073] Performing a convolution operation on the object image through the trained Tiny-DSOD network to obtain a feature map;
[0074] The object key points are extracted based on the feature map through the CPM network.
[0075] Among them, the training of the Tiny-DSOD network and the CPM network can be performed in the terminal, or after training in other devices, the terminal directly obtains it. When training is performed in the terminal, the terminal can obtain multiple object sample images, and the object sample images can be annotated with the two-dimensional coordinate positions of the object vertices and the connection relationship between each vertex, and the two-dimensional coordinate positions of the object vertices are the real two-dimensional coordinate positions. Then, the two-dimensional coordinate positions of the object vertices in each object sample image are calculated using the Tiny-DSOD network and the CPM network to obtain the predicted two-dimensional coordinate positions and the connection relationship between each vertex. At this time, the connection relationship between the real two-dimensional coordinate position and each vertex (i.e., the real value) can be converged with the predicted two-dimensional coordinate position and the connection relationship between each vertex (i.e., the predicted value), and the parameters of the Tiny-DSOD network and the CPM network are adjusted to appropriate values to reduce the error between the real value and the predicted value, so that the trained Tiny-DSOD network and the trained CPM network can be obtained. Among them, the Tiny-DSOD network and the CPM network can be connected in series as a supervised learning network.
[0076] The terminal can process the selected multiple object images through the trained Tiny-DSOD network and the trained CPM network. For example, the object image is convolved through the Tiny-DSOD network to obtain a feature map. For example, the object image is convolved through the 7-layer depth-separable convolution of the Tiny-DSOD network to obtain a feature map. Then, the CPM network extracts the two-dimensional coordinate positions of the box vertices based on the feature map, and determines the connection relationship between the vertices.
[0077] Furthermore, in order to reduce the amount of computation of image data, the object image is convolved through the trained Tiny-DSOD network to obtain a feature map, and the method further includes:
[0078] Converting the 32-bit binary floating-point data in the Tiny-DSOD network into 8-bit binary integer data to obtain a converted Tiny-DSOD network;
[0079] The transformed Tiny-DSOD network is used as the trained Tiny-DSOD network.
[0080] Further, on the basis of the above, the 32-bit binary floating point data in the TINY-DSOD basic network model can be converted into 8-bit binary integer data, and then the converted TINY-DSOD basic network model is used to extract key points. Since the relevant 32-bit binary floating point data is converted into 8-bit binary integer data, the number of data bits is reduced, the amount of calculation is reduced, and it is beneficial to improve the calculation efficiency of the box volume as a whole. The present application can use the CNN (Convolutional Neural Networks, convolutional neural network) framework to convert the float32-bit calculation of the TINY-DSOD basic network model into int8 calculation.
[0081] Further, step S101 may include:
[0082] Acquire a video captured by a camera module in the terminal;
[0083] Converting the analog signal in the video into a compressed image signal;
[0084] The compressed image signal is decompressed to obtain multiple frames of object images in the video.
[0085] In an embodiment of the present invention, in order to improve the accuracy of calculating the volume of an object in a multi-frame object image, when shooting an object, shooting conditions can be set in the terminal. For example, after the terminal starts shooting an object or scene at a preset angle, the shooting angle changes less during the shooting process, and the shooting method is a vertical push-pull method. The higher the accuracy of the subsequent calculation of the object volume of the image, the higher the accuracy. Therefore, the angle at which the terminal shoots the video and the vertical push-pull method can be set when shooting the video. During the shooting process, if it is detected that the shooting angle does not match the set shooting angle, or if it is detected that the shooting method is not a vertical push-pull method, a prompt message can be sent to the user to prompt the user to make adjustments. Furthermore, the image obtained by the terminal is usually not a 1:1 image; and in order to facilitate the placement of the entire object in the image, the size of the screen when the terminal shoots the video can be set, such as setting it to 640*480 or other sizes, and the user can directly shoot when shooting without adjustment.
[0086] Since it is necessary to analyze continuous multiple-frame images to obtain the second acceleration of the object, an analog video signal is used for shooting when shooting the video. After obtaining the video shot by the shooting module in the terminal, the analog signal in the video is converted from an electrical signal to an image signal in a compressed state. In the scheme of the embodiment of the present invention, a method for converting the analog signal of the video to the image signal can be selected according to actual conditions, and there is no restriction on the specific conversion method and algorithm. Through the conversion work of this step, the image signal of the video can be obtained, but the image signal is in a compressed state, and it is necessary to further decompress the compressed image signal to obtain multiple-frame object images of the video, and there is no restriction on the specific decompression method and algorithm.
[0087] After obtaining multiple frames of object images, in order to improve the accuracy of calculating the object volume, the multiple frames of object images may be scaled. Specifically, after obtaining multiple frames of object images, the following steps are further included:
[0088] Get the preset image size;
[0089] Scaling the multiple frames of object images according to the preset image size to obtain scaled multiple frames of object images;
[0090] The obtaining of object key points and connection relationships in each frame of the object image includes:
[0091] Get the object key points and connection relationships in each frame of the zoomed object image.
[0092] Get a preset image size, compare the size of the processed multi-frame object image with the preset image size, if the size of the processed multi-frame object image is larger than the preset image size, then proportionally reduce the image of the processed multi-frame object image to the preset image size. If the size of the processed multi-frame object image is smaller than the preset image size, then proportionally enlarge the image of the processed multi-frame object image to the preset image size. For example, in actual application scenarios, the TINY-DSOD basic network model usually processes a two-dimensional image with a length-to-width ratio of 1:1. The following description takes the preset ratio of 1:1 as an example.
[0093] For example, the object image has a length-to-width ratio of 640*480 and the object image processed by the TINY-DSOD basic network model has a length-to-width size of 368*368. First, the long side size of the object image is reduced from 640 to 368, and then the short side size of the object image is reduced according to the long side reduction ratio (i.e., 640 / 368). At this time, the aspect ratio of the reduced object image is still not 1:1. Then, the object image is expanded in the short side direction, and the grayscale values of the pixels in the expanded part on the three RGB channels are all a predetermined value, such as 128, until the short side of the object image reaches 368. Through the above operation, a multi-frame object image with a length-to-width ratio of 1:1 is obtained.
[0094] Step S102, calculating the scale ratio of the object according to the key points of the multiple frames of object images;
[0095] Specifically, the second acceleration can be calculated based on the first acceleration corresponding to each moment of collecting the object image, and based on the key points of multiple frames of object images and the moment of collecting the object image, the first acceleration and the second acceleration are matched based on the moment, and the scale ratio is calculated based on the matched first acceleration and the second acceleration.
[0096] That is, step S102 may specifically include:
[0097] Obtaining the first acceleration corresponding to each moment of the captured object image;
[0098] Calculating a second acceleration based on the key points of the multiple frames of object images and the time when the object images were collected;
[0099] The first acceleration and the second acceleration are matched based on a time, and a scale ratio is calculated according to the matched first acceleration and the second acceleration.
[0100] Specifically, a multi-frame object image including a box can be obtained through a shooting module in the terminal, and the shooting module can be a camera or a camera preset in the terminal. And the inertial measurement unit (IMU) in the terminal collects acceleration to obtain the first acceleration, or the acceleration sensor preset in the terminal collects acceleration to obtain the first acceleration. Further, in order to improve the accuracy of calculating the second acceleration of the multi-frame image, the multi-frame object image including the box can be obtained by shooting the box video through the shooting module in the terminal, and then parsing it. The first acceleration collected by the inertial measurement module in the terminal is obtained, wherein the inertial measurement unit is a device for measuring the three-axis attitude angle (or angular velocity) and acceleration of the object, so the first acceleration when the terminal shooting module is shooting can be directly obtained from the IMU of the terminal. After obtaining the first acceleration collected by the IMU, since it is mainly to obtain the acceleration data in the three-dimensional orthogonal direction in the IMU coordinate system corresponding to each moment when the shooting module is shooting, the acquired acceleration data can be processed by a low-pass filter to filter out redundant data.
[0101] According to the key points corresponding to each frame of the object image, the displacement value between each two adjacent frames of the image is calculated. Specifically, the pairing points between each two frames of the object image in the multiple frames of the object image are selected according to the key points, such as the pairing points between the A object image and the B object image, and the pixel coordinates of the pairing points, wherein the pixel coordinates can be obtained by establishing a coordinate system in the object image. Then, the numerical matrix of the pairing points is calculated according to the pairing points and the pixel coordinates; and the movement value of the pairing points is calculated according to the numerical matrix of the pairing points; and the second acceleration is calculated according to the movement value and the difference in the shooting time between each two adjacent frames of the object image.
[0102] After obtaining the second acceleration, further obtain the various moments when the IMU collects the first acceleration, and then match the various moments corresponding to the second acceleration with the various moments when the IMU collects the first acceleration. When the matching moment is consistent, the first acceleration and the second acceleration with the same moment are divided to obtain the scale ratio of the current moment. That is, extract the various moments corresponding to the collection of each first acceleration in the first acceleration, and the various moments when the shooting module shoots multiple frames of object images; through the timestamps of the various moments corresponding to the collection of each first acceleration in the first acceleration, with the timestamps of the various moments when the shooting module shoots multiple frames of object images, perform time matching to obtain the acceleration ratio of each moment. Extract the various moments corresponding to the collection of each first acceleration in the first acceleration, and the various moments when the shooting module shoots multiple frames of object images, match the timestamps of the various moments corresponding to the collection of each first acceleration in the first acceleration, with the timestamps of the various moments when the shooting module shoots multiple frames of object images, and divide the first acceleration and the second acceleration at the same moment to obtain the acceleration ratio of each moment, that is, the scale ratio.
[0103] S103: Acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images.
[0104] After obtaining the key points and connection relationships of the object in each frame of the object image, the three-dimensional point cloud information of the object can be obtained according to the key points of the object in each frame of the object image. Specifically, the visual tracking can be performed by real-time matching and parallel local object images to obtain the visual pose. The ORB algorithm can be used for real-time matching and pose estimation, as well as the maintenance and mapping of parallel local object images for multiple frames of object images to perform visual tracking, and then obtain the visual pose; further obtain the acceleration and angular velocity of the inertial measurement module, and integrate the acceleration and angular velocity values to obtain the pose prediction result of the inertial measurement module; obtain the three-dimensional point cloud information of the object according to the visual pose and pose prediction results.
[0105] S104, calculating the volume of the object according to the scale ratio, the connection relationship and the three-dimensional point cloud information.
[0106] Then, the key points are connected through the connection relationship and the 3D point cloud information to obtain the side length of the 3D model, and the real side length of the object can be obtained by multiplying the side length with the 3D scale ratio, and the object volume can be calculated based on the real side length.
[0107] That is, step S104 includes:
[0108] Acquire the side length of the object in multiple frames of object images according to the three-dimensional point cloud information and the connection relationship;
[0109] The volume of the object in the multiple frames of object images is calculated according to the scale ratio and the side length of the object in the multiple frames of object images.
[0110] The key points are connected according to the three-dimensional point cloud information and the connection relationship to obtain the side length of the object in multiple frames of object images; the side length is multiplied by the three-dimensional scale ratio to obtain the real side length; the real volume of the object in the multiple frames of object images is calculated according to the real side length, and specifically, the real side lengths can be multiplied to obtain the real volume of the object.
[0111] This embodiment obtains multiple frames of object images, and obtains the key points and connection relationships of the object in each frame of the object image; calculates the scale ratio of the object based on the key points of the multiple frames of object images; obtains the three-dimensional point cloud information of the object based on the key points of the multiple frames of object images; and calculates the volume of the object based on the scale ratio, connection relationship and three-dimensional point cloud information. This solution calculates the scale ratio of the object through multiple frames of object images continuously shot by the shooting module in the terminal, thereby accurately restoring the volume of the object, extracts the key points of the object from each frame of the object image, obtains the key points and connection relationships of the object in each frame of the object image, obtains the three-dimensional point cloud information of the key points of the object based on the key points of the object, and calculates the volume of the object based on the scale ratio, connection relationship and three-dimensional point cloud information, thereby accurately restoring the volume of the object; improving the accuracy of calculating the volume of the object.
[0112] The object volume calculation method described in the above embodiment will be further described in detail below. In this embodiment, the terminal is a mobile phone and the object is a box to describe the method for calculating the box volume.
[0113] See also Figure 2 , Figure 2 Another flow chart of the object volume calculation method provided in the embodiment of the present application.
[0114] S201, obtaining a box video captured by a camera in a mobile phone.
[0115] In an embodiment of the present invention, in order to improve the efficiency of calculating the volume of the box in the multi-frame box image, when shooting the box, the shooting conditions can be set in the mobile phone. For example, after the mobile phone starts shooting the box or scene at a preset angle, the shooting angle changes less during the shooting process, and the shooting method is a vertical push-pull method. The higher the accuracy of the box volume of the subsequent calculation image, the higher the accuracy. Therefore, the angle of the mobile phone shooting video and the vertical push-pull method when shooting the video can be set. During the shooting process, if it is detected that the shooting angle does not match the set shooting angle, or it is detected that the shooting method is not a vertical push-pull method, a prompt message can be sent to the user to prompt the user to make adjustments. Further, the image obtained by the mobile phone is usually not a 1:1 image; and in order to facilitate the entirety of the box to be placed in the image, the size of the screen when the mobile phone shoots the video can be set, such as 640*480 or other sizes, and the user can directly shoot when shooting without adjustment.
[0116] S202, converting the analog signal in the box video into an image signal in a compressed state.
[0117] S203, decompressing the compressed image signal to obtain multiple frame box images of the video.
[0118] Since it is necessary to analyze continuous multi-frame images, an analog video signal is used for shooting during video shooting. After obtaining the video shot by the shooting module in the mobile phone, the analog signal in the video is converted from an electrical signal to an image signal in a compressed state. In the scheme of the embodiment of the present invention, the method for converting the analog signal of the video to the image signal can be flexibly selected according to the actual situation, and there is no restriction on the specific conversion method and algorithm. Through the conversion work of this step, the image signal of the video can be obtained, but the image signal is in a compressed state, and it is necessary to further decompress the compressed image signal to obtain a multi-frame box image of the video. The method for decompressing the compressed image signal can be flexibly selected according to the actual situation, and there is no restriction on the specific decompression method and algorithm.
[0119] S204, obtaining a preset image size.
[0120] S205 , scaling the multiple frames of box images according to the preset image size to obtain scaled multiple frames of box images.
[0121] Get the preset image size, compare the size of the processed multi-frame box image with the preset image size, if the size of the processed multi-frame box image is larger than the preset image size, then proportionally reduce the image of the processed multi-frame box image to the preset image size. If the size of the processed multi-frame box image is smaller than the preset image size, then proportionally enlarge the image of the processed multi-frame box image to the preset image size. For example, in actual application scenarios, the TINY-DSOD basic network model usually processes a two-dimensional image with a length-to-width ratio of 1:1. The following description takes the preset ratio of 1:1 as an example.
[0122] For example, the ratio of the length to width of the box image is 640*480, and the length to width of the box image processed by the TINY-DSOD basic network model is 368*368. First, the long side of the box image is reduced from 640 to 368, and then the short side of the box image is reduced according to the long side reduction ratio (i.e., 640 / 368). At this time, the aspect ratio of the reduced box image is still not 1:1. Then, the box image is expanded in the short side direction, and the grayscale values of the pixels in the expanded part on the three RGB channels are all a predetermined value, such as 128, until the short side of the box image reaches 368. Through the above operation, a multi-frame box image with a length to width ratio of 1:1 is obtained.
[0123] S206, obtaining a trained Tiny-DSOD network and a CPM network, wherein the Tiny-DSOD network and the CPM network are trained based on sample images of box key points and their connection relationships that are labeled in sequence;
[0124] S207, converting the 32-bit binary floating-point data in the Tiny-DSOD network into 8-bit binary integer data to obtain a converted Tiny-DSOD network;
[0125] S208, using the transformed Tiny-DSOD network as the trained Tiny-DSOD network.
[0126] S209, performing a convolution operation on the box image through the trained Tiny-DSOD network to obtain a feature map;
[0127] S210, extracting box key points and connection relationships based on the feature graph through a CPM network.
[0128] Then obtain the key points and connection relationships of the box in each frame of the box image. Extract the key points of the box contained in each frame of the acquired multi-frame box image to obtain the key points corresponding to each frame of the box image. Specifically, the key points can be extracted and the connection relationship can be obtained through the Tiny-DSOD network and the CPM network, wherein the training of the Tiny-DSOD network and the CPM network can be performed in the terminal, or after training in other devices, the terminal directly obtains it. When training in the terminal, the terminal can obtain multiple box sample images, and the box sample images can be marked with the two-dimensional coordinate positions of the box vertices and the connection relationship between each vertex, and the two-dimensional coordinate position of the box vertex is the real two-dimensional coordinate position. Then, the Tiny-DSOD network and the CPM network are used to calculate the two-dimensional coordinate positions of the box vertices in each box sample image, and the predicted two-dimensional coordinate positions and the connection relationship between each vertex are obtained. At this time, the real two-dimensional coordinate position and the connection relationship between each vertex (i.e., the real value) can be converged with the predicted two-dimensional coordinate position and the connection relationship between each vertex (i.e., the predicted value), and the parameters of the Tiny-DSOD network and the CPM network can be adjusted to appropriate values to reduce the error between the real value and the predicted value, so as to obtain the trained Tiny-DSOD network and the trained CPM network. Among them, the Tiny-DSOD network and the CPM network can be connected in series as a supervised learning network.
[0129] Further, on the basis of the above, the 32-bit binary floating point data in the TINY-DSOD basic network model can be converted into 8-bit binary integer data, and then the converted TINY-DSOD basic network model is used to extract key points. Since the relevant 32-bit binary floating point data is converted into 8-bit binary integer data, the number of data bits is reduced, the amount of calculation is reduced, and it is beneficial to improve the calculation efficiency of the box volume as a whole. The present application can use the CNN (Convolutional Neural Networks, convolutional neural network) framework to convert the float32-bit calculation of the TINY-DSOD basic network model into int8 calculation.
[0130] The terminal can process the selected multiple box images through the trained Tiny-DSOD network and the trained CPM network. For example, the box image is convolved through the Tiny-DSOD network to obtain a feature map. For example, the box image is convolved through the 7-layer depth-separable convolution of the Tiny-DSOD network to obtain a feature map. Then, the CPM network extracts the two-dimensional coordinate positions of the box vertices based on the feature map, and determines the connection relationship between each vertex.
[0131] S211, obtaining the first acceleration corresponding to each moment of collecting the box image.
[0132] S212: Calculate a second acceleration based on the key points of the multiple frames of box images and the time when the box images are collected.
[0133] S213: Match the first acceleration and the second acceleration based on a time, and calculate a scale ratio according to the matched first acceleration and the second acceleration.
[0134] Specifically, a multi-frame box image including a box can be obtained through a shooting module in the terminal, and the shooting module can be a camera or a camera preset in the terminal. And the inertial measurement unit (IMU) in the terminal collects acceleration to obtain the first acceleration, or the acceleration sensor preset in the terminal collects acceleration to obtain the first acceleration. Further, in order to improve the accuracy of calculating the second acceleration of the multi-frame image, the multi-frame box image including the box can be shot by the shooting module in the terminal. The box video is then analyzed. The first acceleration collected by the inertial measurement module in the terminal is obtained, wherein the inertial measurement unit is a device for measuring the three-axis attitude angle (or angular velocity) and acceleration of the box, so the first acceleration when the terminal shooting module is shooting can be directly obtained from the IMU of the terminal. After obtaining the first acceleration collected by the IMU, since it is mainly to obtain the acceleration data in the three-dimensional orthogonal direction in the IMU coordinate system corresponding to each moment of shooting by the shooting module, the acquired acceleration data can be processed by a low-pass filter to filter out redundant data.
[0135] According to the key points corresponding to each frame of the box image, the displacement value between each two adjacent frames of the image is calculated. Specifically, the pairing points between each two frames of the box image in the multiple frames of the box image are selected according to the key points, such as the pairing points between the A box image and the B box image, and the pixel coordinates of the pairing points, wherein the pixel coordinates can be obtained by establishing a coordinate system in the box image. Then, the numerical matrix of the pairing points is calculated according to the pairing points and the pixel coordinates; and the movement value of the pairing points is calculated according to the numerical matrix of the pairing points; and the second acceleration is calculated according to the movement value and the difference in the shooting time between each two adjacent frames of the box image.
[0136] After obtaining the second acceleration, further obtain the various moments when the IMU collects the first acceleration, and then match the various moments corresponding to the second acceleration with the various moments when the IMU collects the first acceleration. When the matching moment is consistent, divide the first acceleration and the second acceleration at the same moment to obtain the scale ratio of the current moment. That is, extract the various moments corresponding to the collection of each first acceleration in the first acceleration, and the various moments when the shooting module shoots multiple frames of box images; through the timestamps of the various moments corresponding to the collection of each first acceleration in the first acceleration, with the timestamps of the various moments when the shooting module shoots multiple frames of box images, match the moments to obtain the acceleration ratios at each moment. Extract the various moments corresponding to the collection of each first acceleration in the first acceleration, and the various moments when the shooting module shoots multiple frames of box images, match the timestamps of the various moments corresponding to the collection of each first acceleration in the first acceleration, with the timestamps of the various moments when the shooting module shoots multiple frames of box images, and divide the first acceleration and the second acceleration at the same moment to obtain the acceleration ratios at each moment, that is, the scale ratio.
[0137] S214, obtaining the side length of the box in the multiple frames of box images according to the three-dimensional point cloud information and the connection relationship.
[0138] S215, calculating the volume of the box in the multiple-frame box image according to the scale ratio and the side length of the box in the multiple-frame box image.
[0139] By connecting the key points according to the three-dimensional point cloud information and the connection relationship, the side length of the box in the multi-frame box image can be obtained; the side length is multiplied by the three-dimensional scale ratio to obtain the real side length; the real volume of the box in the multi-frame box image is calculated according to the real side length, and specifically, the real side lengths can be multiplied to obtain the real volume of the box.
[0140] This solution obtains the key points and connection relationships of the box in each frame of the box image through continuous shooting of multiple frames of box images by a shooting module in a mobile phone, calculates the scale ratio of the box according to the key points of the multiple frames of box images; obtains the three-dimensional point cloud information of the box according to the key points of the multiple frames of box images; calculates the volume of the box according to the scale ratio, connection relationship and three-dimensional point cloud information, thereby accurately restoring the volume of the box; and improves the accuracy of calculating the volume of the box.
[0141] In order to facilitate better implementation of the object volume calculation method provided in the embodiment of the present application, the embodiment of the present application also provides a device for calculating the volume of an object based on the above. The meanings of the terms are the same as those in the above object volume calculation method, and the specific implementation details can refer to the description in the method embodiment.
[0142] See also Figure 3 , Figure 3 This is a schematic diagram of the structure of an object volume calculation device provided in an embodiment of the present application, wherein the object volume calculation device may include a first acquisition unit 301, an extraction unit 302, a first calculation unit 303, a second acquisition unit 304, and a second calculation unit 305.
[0143] Specifically, the object volume calculation device includes:
[0144] A first acquisition unit is used to acquire multiple frames of object images, and acquire key points and connection relationships of the objects in each frame of the object images;
[0145] A first calculation unit, configured to calculate a scale ratio of the object according to key points of the multiple frames of object images;
[0146] A second acquisition unit, configured to acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images;
[0147] The second calculation unit is used to calculate the volume of the object according to the scale ratio, the connection relationship and the three-dimensional point cloud information.
[0148] In some implementations, the first acquisition unit includes:
[0149] A first acquisition subunit is used to acquire a trained Tiny-DSOD network and a CPM network, wherein the Tiny-DSOD network and the CPM network are trained based on sample images of object key points and their connection relationships that are labeled in sequence;
[0150] A convolution subunit, used to perform a convolution operation on the object image through the trained Tiny-DSOD network to obtain a feature map;
[0151] The extraction subunit is used to extract the key points and connection relationships of the object based on the feature map through the CPM network.
[0152] In some implementations, the first acquisition unit further includes:
[0153] The transformation subunit is used to transform the 32-bit binary floating point data in the Tiny-DSOD network into 8-bit binary integer data to obtain a transformed Tiny-DSOD network; and use the transformed Tiny-DSOD network as the trained Tiny-DSOD network.
[0154] In some implementations, the second computing unit includes:
[0155] A second acquisition subunit is used to acquire the side length of the object in the multiple frames of object images according to the three-dimensional point cloud information and the connection relationship;
[0156] The first calculation subunit is used to calculate the volume of the object in the multiple frames of object images according to the scale ratio and the side length of the object in the multiple frames of object images.
[0157] In some implementations, the first acquisition unit further includes:
[0158] A third acquisition subunit, used to acquire a preset image size;
[0159] A scaling subunit, used for scaling the multiple frames of object images according to the preset image size to obtain scaled multiple frames of object images;
[0160] The acquisition unit further includes:
[0161] The fourth acquisition subunit is used to acquire the key points and connection relationships of the objects in each frame of the zoomed object image.
[0162] In some embodiments, the acquisition unit further includes:
[0163] a fifth acquisition subunit, configured to acquire a video including an image of the object;
[0164] Converting the analog signal in the video into a compressed image signal;
[0165] The decompression subunit is used to decompress the image signal in a compressed state to obtain multiple frames of object images in the video.
[0166] In some implementations, the first computing unit includes:
[0167] A sixth acquisition subunit, used to acquire the first acceleration corresponding to each moment of the captured object image;
[0168] The second calculation subunit is used to calculate the second acceleration based on the key points of the multiple frames of object images and the time when the object images are collected; match the first acceleration and the second acceleration based on the time, and calculate the scale ratio according to the matched first acceleration and the second acceleration.
[0169] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.
[0170] Figure 4 The specific structural block diagram of the device provided in the embodiment of the present invention is shown, and the device can be used to implement the object volume calculation method provided in the above embodiment. The device 400 can be a smart phone or a tablet computer.
[0171] like Figure 4 As shown, the device 400 may include an RF (Radio Frequency) circuit 110, a memory 120 including one or more (only one is shown in the figure) computer-readable storage media, an input unit 130, a display unit 140, a transmission module 170, a processor 180 including one or more (only one is shown in the figure) processing cores, and a power supply 190. Those skilled in the art will understand that Figure 4 The structure of the device 400 shown in the figure does not constitute a limitation on the device 400, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.
[0172] The RF circuit 110 is used to receive and send electromagnetic waves, realize the mutual conversion between electromagnetic waves and electrical signals, and thus communicate with a communication network or other devices. The RF circuit 110 may include various existing circuit elements for performing these functions, such as antennas, radio frequency transceivers, digital signal processors, encryption / decryption chips, user identity module (SIM) cards, memories, etc. The RF circuit 110 can communicate with various networks such as the Internet, corporate intranets, wireless networks, or communicate with other devices through wireless networks. The above-mentioned wireless networks may include cellular telephone networks, wireless local area networks, or metropolitan area networks. The above-mentioned wireless networks can use various communication standards, protocols and technologies, including but not limited to Global System for Mobile Communication (GSM), Enhanced Data GSM Environment (EDGE), Wideband Code Division Multiple Access (WCDMA), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Wireless Fidelity (Wi-Fi) (such as Institute of Electrical and Electronics Engineers standards IEEE 802.11a, IEEE 802.11b, IEEE802.11g and / or IEEE 802.11n), Voice over Internet Protocol (VoIP), Worldwide Interoperability for Microwave Access (Wi-Max), other protocols for email, instant messaging and short messages, and any other suitable communication protocols, even those that have not yet been developed.
[0173] The memory 120 can be used to store software programs and modules, such as the program instructions / modules of the object volume calculation method in the above-mentioned embodiment. The processor 180 executes various functional applications and data processing by running the software programs and modules stored in the memory 120, that is, realizes the function of object volume calculation. The memory 120 may include a high-speed random access memory, and may also include a non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, the memory 120 may further include a memory remotely arranged relative to the processor 180, and these remote memories may be connected to the device 400 via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.
[0174] The input unit 130 can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control. Specifically, the input unit 130 may include a touch-sensitive surface 131 and other input devices 132. The touch-sensitive surface 131, also known as a touch display screen or a touch pad, can collect user touch operations on or near it (such as operations performed by users using fingers, styluses, or any other suitable objects or accessories on or near the touch-sensitive surface 131), and drive corresponding connection devices according to a pre-set program. Optionally, the touch-sensitive surface 131 may include a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch orientation, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into touch point coordinates, and then sends it to the processor 180, and can receive and execute commands sent by the processor 180. In addition, the touch-sensitive surface 131 can be implemented in various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch-sensitive surface 131, the input unit 130 may further include other input devices 132. Specifically, other input devices 132 may include but are not limited to one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, etc.
[0175] The display unit 140 can be used to display information input by the user or information provided to the user and various graphical user interfaces of the device 400, which can be composed of graphics, text, icons, videos and any combination thereof. The display unit 140 may include a display panel 141. Optionally, the display panel 141 may be configured in the form of LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode), etc. Further, the touch-sensitive surface 131 may cover the display panel 141. When the touch-sensitive surface 131 detects a touch operation on or near it, it is transmitted to the processor 180 to determine the type of touch event. Subsequently, the processor 180 provides a corresponding visual output on the display panel 141 according to the type of touch event. Although in Figure 4 In the embodiment, the touch-sensitive surface 131 and the display panel 141 are implemented as two independent components to implement input and output functions. However, in some embodiments, the touch-sensitive surface 131 and the display panel 141 can be integrated to implement input and output functions.
[0176] The device 400 can help users send and receive emails, browse web pages, and access streaming media through the transmission module 170 (such as a Wi-Fi module), and it provides users with wireless broadband Internet access. Figure 4 The transmission module 170 is shown, but it is understandable that it is not a necessary component of the device 400 and can be omitted as needed without changing the essence of the invention.
[0177] The processor 180 is the control center of the device 400. It uses various interfaces and lines to connect various parts of the entire mobile phone. By running or executing software programs and / or modules stored in the memory 120, and calling data stored in the memory 120, it executes various functions of the device 400 and processes data, thereby monitoring the mobile phone as a whole. Optionally, the processor 180 may include one or more processing cores; in some embodiments, the processor 180 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, etc., and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 180.
[0178] The device 400 also includes a power supply 190 (such as a battery) for supplying power to various components. In some embodiments, the power supply can be logically connected to the processor 180 through a power management system, so that the power management system can manage charging, discharging, and power consumption management. The power supply 190 can also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.
[0179] Specifically in this embodiment, the display unit 140 of the device 400 is a touch screen display, and the device 400 further includes a memory 120 and one or more programs, wherein the one or more programs are stored in the memory 120 and are configured to be executed by one or more processors 180. The one or more programs include instructions for performing the following operations:
[0180] Obtain multiple frames of object images, and obtain the object key points and connection relationships in each frame of the object image;
[0181] Calculating a scale ratio of the object according to key points of the multiple frames of object images;
[0182] Acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images;
[0183] The volume of the object is calculated according to the scale ratio, the connection relationship and the three-dimensional point cloud information.
[0184] In the above embodiments, the description of each embodiment has its own focus. For the part that is not described in detail in a certain embodiment, please refer to the detailed description of the object volume calculation method above, which will not be repeated here.
[0185] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.
[0186] To this end, an embodiment of the present application provides a storage medium in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the steps in any object volume calculation method provided in the embodiment of the present application. For example, the instructions can execute the following steps:
[0187] Obtain multiple frames of object images, and obtain the object key points and connection relationships in each frame of the object image;
[0188] Calculating a scale ratio of the object according to key points of the multiple frames of object images;
[0189] Acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images;
[0190] The volume of the object is calculated according to the scale ratio, the connection relationship and the three-dimensional point cloud information.
[0191] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.
[0192] The storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.
[0193] Since the instructions stored in the storage medium can execute the steps in any object volume calculation method provided in the embodiments of the present application, the beneficial effects that can be achieved by any object volume calculation method provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.
[0194] The above is a detailed introduction to a method, device, equipment and storage medium for calculating the volume of an object provided in an embodiment of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for technical personnel in this field, according to the idea of the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.
Claims
1. A method for calculating the volume of an object, characterized in that: include: Obtain multiple frames of object images, and obtain the object key points and connection relationships in each frame of the object image; Calculating a scale ratio of the object according to key points of the multiple frames of object images; Acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images; Calculating the volume of the object according to the scale ratio, the connection relationship and the three-dimensional point cloud information; The calculating the scale ratio of the object according to the key points of the multiple frames of object images includes: Acquire a first acceleration corresponding to each moment of capturing an image of the object; the first acceleration is acquired by an inertial measurement module or a preset acceleration sensor in the terminal; Calculating a second acceleration based on the key points of the multiple frames of object images and the time when the object images were collected; matching the first acceleration and the second acceleration based on the time, and calculating a scale ratio according to the matched first acceleration and the second acceleration; The calculating the volume of the object according to the scale ratio, the connection relationship and the three-dimensional point cloud information includes: Acquire the side length of the object in multiple frames of object images according to the three-dimensional point cloud information and the connection relationship; The real volume of the object in the multiple frames of object images is calculated according to the scale ratio and the side length of the object in the multiple frames of object images.
2. The object volume calculation method according to claim 1, characterized in that: The obtaining of object key points and connection relationships in each frame of the object image includes: Obtaining a trained Tiny-DSOD network and a CPM network, wherein the Tiny-DSOD network and the CPM network are trained based on sample images of object key points and their connection relationships that are labeled in sequence; Convolution operation is performed on the object image through the trained Tiny-DSOD network to obtain a feature map; The key points and connection relationships of the object are extracted based on the feature graph through the CPM network.
3. The object volume calculation method according to claim 2, characterized in that: Before the object image is convolved through the trained Tiny-DSOD network to obtain a feature map, the method further includes: Converting 32-bit binary floating-point data in the Tiny-DSOD network into 8-bit binary integer data to obtain a converted Tiny-DSOD network; The transformed Tiny-DSOD network is used as the trained Tiny-DSOD network.
4. The object volume calculation method according to claim 1, characterized in that: After acquiring multiple frames of object images, the method further includes: Get the preset image size; Scaling the multiple frames of object images according to the preset image size to obtain scaled multiple frames of object images; The obtaining of object key points and connection relationships in each frame of the object image includes: Get the object key points and connection relationships in each frame of the zoomed object image.
5. The object volume calculation method according to claim 1, characterized in that: The step of acquiring multiple frames of object images comprises: Acquire a video including an image of an object; Converting the analog signal in the video into a compressed image signal; The compressed image signal is decompressed to obtain multiple frames of object images in the video.
6. An object volume calculation device, characterized in that: include: A first acquisition unit is used to acquire multiple frames of object images, and acquire key points and connection relationships of the objects in each frame of the object images; A first calculation unit, configured to calculate a scale ratio of the object according to key points of the multiple frames of object images; A second acquisition unit, configured to acquire three-dimensional point cloud information of the object according to key points of the multiple frames of object images; A second calculation unit, used for calculating the volume of the object according to the scale ratio, the connection relationship and the three-dimensional point cloud information; The calculating the scale ratio of the object according to the key points of the multiple frames of object images includes: Acquire a first acceleration corresponding to each moment of capturing an image of the object; the first acceleration is acquired by an inertial measurement module or a preset acceleration sensor in the terminal; Calculating a second acceleration based on the key points of the multiple frames of object images and the time when the object images were collected; matching the first acceleration and the second acceleration based on the time, and calculating a scale ratio according to the matched first acceleration and the second acceleration; The calculating the volume of the object according to the scale ratio, the connection relationship and the three-dimensional point cloud information includes: Acquire the side length of the object in multiple frames of object images according to the three-dimensional point cloud information and the connection relationship; The real volume of the object in the multiple frames of object images is calculated according to the scale ratio and the side length of the object in the multiple frames of object images.
7. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores a program code, and the processor executes the object volume calculation method according to any one of claims 1 to 5 when calling the program code in the memory.
8. A storage medium, characterized in that: The storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the object volume calculation method according to any one of claims 1 to 5.
Citation Information
Patent Citations
Scale robust feature-based identifiers for image identification
CN102016880A
3D camera volume measuring method and system
CN108267083A
