Abstract
Aiming at the difficulty of standardizing the action of basketball shooting training, a new method of standardizing the action of basketball shooting training is proposed based on digital video technology. The digital video signal representation, video sequence coding data structure, and video sequence compression coding method are analyzed, and the pixels of basketball shooting training action position space are sampled to collect basketball shooting training images. The time difference method is used to extract the movement target of basketball shooting training from a digital video sequence. Based on digital video technology, the initial background image is estimated, and the update rate is introduced to update the background estimation image. According to the pixel value sequence of the basketball shooting training image, the pixel model of the basketball shooting training image is defined and modified. By judging whether the defined pixel value matches the background parameter model, the standardization of shooting training can be realized. The experimental results show that the proposed method has good stability, high precision, and short time in determining the standardization of shooting movement, can correct the wrong shooting movement in real time, and can effectively guide basketball shooting training.
1. Introduction
Basketball is quite different from other sports. It is a high-intensity and comprehensive sport [1, 2]. Basketball belongs to the same field antagonistic event group dominated by technical and tactical ability, and the technical and tactical level is the decisive factor for the competitive level of basketball [3]. In the actual competition process, basketball players need to have diversified basketball qualities to ensure the victory of the competition. Among them, the more important point is the coordination and stability of athletes' physical functions. In the development of basketball, shooting is its key offensive technology. The essence of a basketball game is a shooting game, which also shows that the stability of shooting has an important relationship with the outcome of the game [4]. Among them, the factors affecting the standardization of athletes' shooting action mainly include athletes' bodies, technology, and psychology. In the process of competition, athletes need to ensure that they master the nature, time, and score of the competition [5]. In the actual basketball game, the attacking team needs to use different techniques or tactics to create more shooting opportunities and ensure shooting scores [6, 7]. The defensive team should actively defend and prevent the other team from scoring. The accuracy and standard of shooting in basketball are directly related to the score. Therefore, it is of great significance to reasonably test and judge the shooting action of basketball.
At present, scholars in related fields have carried on the research on action judgment and obtained some research results. Reference [8] proposed a motion similarity judgment method based on motion primitives. Based on the computational model of kinematics, the similarity of motion and its performance to the human similarity judgment of the same motion are determined. By performing the action similarity task and comparing it with the computational model solving the same task, the action similarity judgment was realized by classifying the actions based on the learned kinematics primitive. The method has high reliability and provides necessary basis for human action classification. Reference [9] proposed a basketball motion image target detection method based on an improved Gaussian mixture model. Edge detection, gray processing, target capture, target recognition, image detection, and other technologies are integrated into basketball sports video, and Gaussian probability density mixing is used to select the appropriate number of continuously updated parameters and each pixel area to achieve basketball sports image detection. The method is effective to some extent. However, the above methods are difficult to determine the standardization of basketball shooting training movements.
In view of the above problems, a method to judge the movement standardization of shooting training based on digital video technology is proposed. The innovation of the research method is to use the time difference method to extract the movement target of basketball shooting training. Based on digital video technology, the initial background image is estimated and the update rate is introduced to update the image. According to the pixel value sequence of the basketball shooting training image, the pixel model of the basketball shooting training image is defined and modified. By discriminating the matching relation between pixel value and background parameter model, the standardization judgment of shooting training movement can be realized. Compared with the previous research results, the method designed based on digital video technology has better stability, high accuracy, and short time and can correct the wrong shooting action in real time, which can effectively guide basketball shooting training.
2. Digital Video Technology
Digital video technology is to first use video capture equipment such as cameras to convert the color and brightness information of external images into electrical signals, and then record them into storage media [10–12]. Digital video is video recorded in digital form, as opposed to analog video. Digital video has different production methods, storage methods, and broadcast methods. For example, digital video signals are generated directly through digital cameras and stored on digital tape, P2 card, blue disc, or disk, so as to obtain different formats of digital video, which is then played on a PC, a specific player, etc.
2.1. Representation of Digital Video Signal
Video is described as a group of continuous images, and each image is regarded as a two-dimensional pixel array. The color representation of each pixel includes three components: red , green , and blue , which is called the space representation of the image. The color coordinates used for the three digital TV systems are different. For digital video capture and display, all three digital TV systems use primary colors, but the definition of each primary color spectrum is slightly different. For the transmission of digital video signal, in order to reduce the required bandwidth and be compatible with monochrome digital TV system, the brightness/chroma coordinate system is adopted [13, 14]. The color coordinates used in the NTSC, PAL, and SECAM systems are all derived from the coordinates used for PAL, and is derived from the coordinates. According to the relationship between the primary color and the primary color, the value of the luminance component can be determined by the value of . The two chromaticity values and are proportional to the color differences and , respectively, and are adjusted to the desired range. The classic conversion relationship between the coordinate system and the primary color value is
The conversion of two color spaces is based on the characteristics of a human visual system: in space, if one of the three signals , , and changes, the color of the total image will change, and the human eye can easily detect this change. However, human eyes have different responses to the changes of and signals. Among them, they are sensitive to the changes of luminance signals, but not very sensitive to the changes of chrominance signals. In this way, we can consider more luminance signals and adopt some processing methods for chrominance signals to improve the compression ratio.
2.2. Data Structure of Digital Video Sequence Coding
In the coding scheme, the video sequence is divided and multiplexed by multiple layers to establish such a data structure.(1)Sequence: the video sequence starts with the sequence header, including several image groups, and ends with a sequence end code.(2)Group of pictures (GOP): GOP is a head followed by a series of images, which allows fast random access to the sequence, fast search, and editing. It is the smallest coding unit that can be decoded independently in the sequence [15]. The first image in the GOP is an intracoded image (I frame), followed by a forward prediction coded image (P frame) and a bidirectional prediction image (B frame). Each GOP has only one I frame, and this I frame is used as the first frame to start coding. The P frame is encoded by motion-compensated prediction relative to the previous I frame or P frame, and the P frame can be used as a reference frame for other P frame or B frame coding. B frame is encoded by motion compensation prediction of two frames, one is the past frame, and the other is the future frame. The frame arrangement of the GOP is shown in Figure 1. The standard does not specify the number of P and B frames in a GOP, nor their specific sequence, except for the first frame, and only one frame is I frame. Any sequence and frame number can be used to design the encoding scheme. The prediction of each P and B frame is based on the previous reference prediction frame. Too many frames in the group layer will affect the quality and compression ratio of coding. Therefore, 10–15 B frames are generally selected in each group layer, and 2–3 B frames are separated between the two P frames.(3)Image: image is the basic coding unit of video sequence [16]. The image is composed of three rectangular matrices representing the luminance and two chrominance and values. Each video standard divides the image into macroblock groups. H.261 and H.263 use a fixed macroblock structure, whereas MPEG1/2 allows a flexible structure, and MPEG4 arranges a variable number of macroblocks into a group.(4)Group of blocks (GOB): H.261 and H.263 divide the image into GOBs. Each GOB includes three macroblock lines and 11 macroblocks in each GOB line, and the GOB header defines the position of the GOB in the image.(5)Slice: a slice consists of several successive macroblocks into a unit. The size of the slice can be changed. The slice layer provides anti-interference ability against data errors.(6)Macroblock (MB): Macro block (MB): MB is a basic concept in video coding technology. It can divide the image into many blocks of different sizes and implement different compression strategies in different locations. A coded image is usually composed of several macroblocks. A macroblock is composed of a luminance pixel block and two additional chrominance pixel blocks. In general, the brightness block is 16 × 16 pixel block, chroma block is 8 × 8-size pixel block; several macroblocks in each image are arranged in the form of slices. The video coding algorithm takes macroblocks as units, encodes macroblocks one by one, and organizes them into a continuous video code stream.(7)Block: block is the smallest coding unit in the standardized video coding algorithm. It consists of 8 × 8 pixel composition [17].
2.3. Digital Video Sequence Compression Coding Method
2.3.1. Transform Coding
Transform coding does not directly encode the spatial image signal, but first maps and transforms the spatial signal to another orthogonal vector space to generate a batch of transform coefficients, and then encodes these transform coefficients [18–20]. In the digital video sequence image compression and coding technology, the compression performance and error of discrete cosine transform (DCT) are very close to those of K-L transform, and DCT has the characteristics of moderate computational complexity, separability, and fast algorithm. Therefore, there are many schemes using DCT coding in image data compression [21–23]. At present, DCT is used in almost all transform-based image encoders.
Assuming that the range of the spatial variables is and , and the range of the frequency domain variables is , , the two-dimensional discrete cosine sine transform formula is
In formula (2), when , . The two-dimensional inverse discrete cosine transform formula is
In formula (3), when , . The core of the two-dimensional discrete cosine transform is separable, so both the forward and inverse transforms can decompose the two-dimensional transform into a series of one-dimensional transforms (rows, columns) for calculation [24, 25].
In the MPEG series, since the basic unit of DCT transformation is a luminance block or a chrominance block, the size is 8 × 8, so can be used in formula (2), so that
In practical application, considering the characteristics of separable variables in formula (4), rewriting the latter part of formula (4) can obtain
Set an intermediate variable ; then,
The coefficient can be written as
It can be seen that after such processing, the two-dimensional DCT transform is decomposed into one row DCT and one column DCT transform, which is easy to be realized by computer. For the inverse transformation, the variables can still be separated to facilitate the implementation of the algorithm.
2.3.2. Predictive Coding
Predictive coding is a technique to improve compression performance through statistical redundancy. Based on the previously encoded pixel values, the encoder can estimate and predict the pixel values to be encoded and decoded [26–28]. For a large number of static or slowly varying regions in the sequence image, the conditional patching method can be used to store the first frame image in the reference frame and send it to the other party. Then, the predicted value of the pixel sampling value of the frame image at is the restoration value of the pixel value at the same position of the frame image. The interframe difference is expressed as
For sequence images with relatively moderate amount of motion, encode the frame difference of the transmitted subblock, where is the subscript of the subblock, and use the following formula to restore the subblock:
3. Standardized Judgment Method of Shooting Training Action
The standardized judgment method of shooting training action based on digital video technology is mainly to collect the basketball shooting training image by sampling and characteristic analysis of the pixels in the position space of basketball shooting training action. Using the time difference method, the basketball shooting training target is found and extracted in real time in the digital video sequence. Based on digital video technology, the initial background estimation image is introduced, and the update rate is introduced to update the background estimation image. According to the pixel value sequence of the basketball shooting training image, the pixel model of the basketball shooting training image is defined, and the defined pixel value model parameters are modified. The standardized judgment of shooting training action is realized by judging whether the defined pixel value matches the background parameter model.
3.1. Collecting Basketball Shooting Training Images
Assuming that the Gaussian mixture model labels the spatial position rotation of basketball shooting action, at multiple points in the basketball shooting space, the shape coordinate of the basketball shooting action under the initial deformation is , the width of the entire characteristic image of the basketball court is , and the height is . The three-dimensional spatial feature image of basketball shooting is divided into several subblocks by using the grid model. The matching coordinate of the central point of the matching point along the gradient direction on the grid model is calculated as , and then, the spherical grid model of basketball in the hands of players is calculated. The triangular partition pheromone of single-frame basketball shooting action is obtained at the manual calibration point :
The basketball shooting action sampling image has 8 × 8 pixels in the grid surface. The sampling point density feature is extracted, and the mean square error between the standardized feature points of the shooting action is
In the above formula, is the total number of uniformly distributed grids of the image. Considering all the pixel feature points of spatial positions, the difference error vector of basketball players in shooting and lifting the ball is obtained as
Thus, the pixel sampling and feature analysis of three main position spaces of basketball shooting training action are realized, and the image acquisition of basketball shooting training is completed.
3.2. Extracting the Goal of Basketball Shooting Training
The time difference method mainly uses the difference of two or several consecutive frames in the digital video sequence to extract the moving target of basketball shooting training [29–31]. The basic process of the time difference method is shown in Figure 2.(1)Calculate the difference image between the frame image and the frame image ; according to the following three methods, the difference image obtained is expressed as follows. Positive difference: Negative difference: Full difference:(2)Binarize the image after the difference to obtain In the above formula, contains the change of the scene between two consecutive frames of images. This change is composed of many factors. It can be considered that the change of the moving target is obvious. Given a threshold, when the difference of a pixel value in the differential image is greater than a given threshold, the pixel is considered to be a foreground pixel, possibly a point on the target; otherwise, it is considered a background pixel [32–34].(3)Postprocessing the image to obtain , where the area of the moving target should be greater than the given threshold. Morphological filtering and noise removal can be used to eliminate noise in small areas.(4)Judge the postprocessing result , mark the area larger than the given threshold as the target, and obtain its complete location information. Through the above steps, the basketball shooting training sports goal is extracted.
3.3. Judging the Standardization of Basketball Shooting Training Action
(1)Initial background estimation image: the single Gaussian distribution background model is suitable for single-modal background situations. It establishes a model represented by a single Gaussian distribution for the color of each image point, where represents time. Let the current color value of the basketball shooting training image point be ; calculate the average brightness of each pixel and the variance of pixel brightness of the basketball shooting training image in the digital video sequence, and take the image with Gaussian distribution composed of and as the initial background estimation image, which is expressed as In formula (17), .(2)Update the background estimation image: the update of the single Gaussian distribution background model refers to the update of the Gaussian distribution parameter of the basketball shooting training image. The constant update rate that represents the update speed is introduced, and the update of the Gaussian distribution parameter of this point can be expressed as In the above formula, .(3)Define the pixel model: define the distribution model for each basketball shooting training image pixel. Set the pixel value sequence of the basketball shooting training image as , and on this basis, define a set of multiple single models: In the above formula, is each single model, which consists of three parameters, where is the weight of this single model, and its size reflects the current reliability of the pixel value represented by this model; is the mean value of this single model, which reflects the center of each single peak distribution; and is the width of the unimodal distribution of this single model, and its size reflects the degree of instability of the pixel value, and its role is equivalent to that of the aforementioned single model. is the number of single models, which reflects the number of peaks in the multipeak distribution of pixel values. Its selection depends on the pixel value distribution and also on the computing power of the system. The usual value is between 3 and 5. In order to keep the model close to the current distribution of pixel values, it is necessary to update the parameters of this model for each defined pixel value [35].(4)Correct the pixel value model parameters: the parameter correction steps are as follows: Step 1: for each new pixel value, first check whether it matches the model. The detection method is Step 2: after the detection in step 1, the weight of the single model matching the defined pixel value is corrected as The parameters of the single model that matches the defined pixel values are corrected as follows: Step 3: after completing the above correction, it is necessary to normalize the weight of single model in the model as follows:(5)Establishment of background pixel model: the above model is used to define the pixel value of the basketball shooting training image; that is, it is necessary to judge whether the defined pixel value is target pixel or background pixel, so as to realize the standardized judgment of shooting training action. Calculate each single model , arrange each model in descending order of the single model, consider the previous model to be a background model, and obtain the model of background pixels [36] expressed as
Through the above steps, the pixel value of the basketball shooting training image is defined by using the background parameter model, and the standardized judgment of shooting training action is realized by judging whether the defined pixel value matches the background parameter model.
4. Experimental Analysis
4.1. Experimental Environment and Data
In order to verify the effectiveness of the method for determining the standardization of shooting training movements based on digital video technology, the experiments were conducted on a computer with Intel Core i7-6800K, 3.4 GHz, Nvidia GeForce GTX1080 (8G) graphics card and 24G memory. The operating system is Win10, and the software platform is Anaconda3 and Visual Studio 2015. The resolution of basketball shooting action visual image sampling is 320 × 240. A group of basketball shooting action visual image simulation data express a basketball shooting action. There are 100 test sample image sets in each shooting action mode and a total of 1024 × 1000 test sets in basketball shooting action visual image database. The reference background neighborhood is 5 × 5 image blocks, that is, 20 × 20 pixels, and the model update parameters are , and the threshold is determined after many experiments. In order to ensure the absolute fairness of the experimental results, the ball selection processing in the whole experimental process is completed by the artificial intelligence robot, and the relevant participants only serve as the detection and verification personnel to supervise and investigate the ball selection operation of the robot. According to the above parameters, SolidWorks is used to establish a simplified visual analysis model of basketball shooting action, import the analysis data into ADAMS software for image processing and analysis, and make standardized judgment on basketball shooting action. The standardized action mode of basketball shooting is shown in Figure 3.
Save the basketball shooting standardized action data shown in Figure 3 as . TXT text data, load it into the image data processing software, conduct computer vision analysis, guide the actual shooting action, collect the basketball shooting training image, and obtain the original basketball shooting information. The collection results are shown in Figure 4.
Figure 4 shows the original basketball shooting information collection results. In order to realize the standardized judgment of shooting training action, it is necessary to extract the training target from the collected original basketball shooting information. The proposed standardized judgment method of shooting training action based on digital video technology is used to extract the collected original basketball shooting training target and judge the standardization of shooting training action. The results are shown in Figure 5.
It can be seen from the analysis of Figure 5 that the standardized judgment method of shooting training action based on digital video technology can effectively realize the extraction and detection of moving targets in basketball shooting training, correct shooting errors in real time, and effectively guide basketball shooting training.
In order to evaluate and compare the proposed standardized judgment method of shooting training action, the accuracy rate and recognition rate are used as evaluation indexes, and the calculation formulas of accuracy rate and recognition rate are as follows:where is the number of foreground pixels that are correctly detected, is the number of pixels whose background is misjudged as foreground, and is the number of pixels whose foreground is misjudged as background.
4.2. Comparison of Standardized Judgment Accuracy of Shooting Training Action
In order to further verify the judgment accuracy of the proposed method, the accuracy is taken as the evaluation index. The higher the accuracy, the higher the judgment accuracy. By comparing the method of reference [8] and the method of reference [9], the standardized judgment accuracy of shooting training action of different methods is obtained, and the comparison results are shown in Figure 6.
According to the analysis of Figure 6, when the number of experiments is 30, the average standardized judgment accuracy of the shooting training action of the method of reference [8] is 84.6%, the average standardized judgment accuracy of the shooting training action of the method of reference [9] is 70.3%, and the average standardized judgment accuracy of shooting training action of the proposed method is as high as 95.2%. Therefore, compared with the method of reference [8] and the method of reference [9], the proposed method has a higher accuracy of the standardized judgment of shooting training action and can effectively improve the accuracy of standardized judgment of shooting training action.
4.3. Comparison of Standardized Judgment and Stability of Shooting Training Action
Further verify the judgment stability of the proposed method, and take the recognition rate as the evaluation index. The higher the recognition rate, the better the judgment stability of the method. By comparing the method of reference [8], the method of reference [9], and the proposed methods, we get the comparison results of the standardized judgment stability of shooting training actions of different methods, as shown in Figure 7.
According to the analysis of Figure 7, when the number of experiments is 30, the average standard judgment recognition rate of shooting training action of the method of reference [8] is 88.4%, the average standard judgment recognition rate of shooting training action of the method of reference [9] is 80.2%, and the average standard judgment recognition rate of the shooting training action of the proposed method is as high as 96%. It can be seen that compared with the method of reference [8] and the method of reference [9], the proposed method has better stability in judging the standardization of shooting training action.
4.4. Comparison of Standardized Judgment Time of Shooting Training Action
On this basis, the judgment time of the proposed method is verified, and the method of reference [8], the method of reference [9], and the proposed method are compared. The standardized judgment time of the shooting training action of different methods is compared, and the comparison results are shown in Table 1.
According to the data in Table 1, with the increase in the number of experiments, the standardized judgment time of shooting training actions of different methods increases. When the number of experiments reaches 30, the standardized judgment time of the shooting training action of the method of reference [8] is 23.9 s, the standardized judgment time of the shooting training action of the method of reference [9] is 26.5 s, whereas the standardized judgment time of the shooting training action of the proposed method is only 15.3 s. Therefore, compared with the method of reference [8] and the method of reference [9], the standardized judgment time of shooting training action of the proposed method is shorter.
5. Conclusion
(1)The proposed standardized judgment method of shooting training action based on digital video technology gives full play to the advantages of digital video technology(2)The standardized judgment of shooting training action is high, which can effectively shorten the judgment time and has good judgment stability(3)Correct shooting mistakes in real time, and effectively guide basketball shooting trainingHowever, in the process of standardized judgment of shooting training action, dimension reduction is not considered to deal with the characteristics of shooting training action, so as to reduce the amount of calculation. Therefore, in the next research, the dimension of shooting training action characteristics is reduced to further reduce the judgment time.
Data Availability
The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.
Conflicts of Interest
The authors declare that they have no conflicts of interest regarding this work.
Acknowledgments
This work was supported by key topics in 2020 of the 13th Five-Year Plan of Educational Science in Gansu Province: Research on the implementation path of physical education to improve college students' physical quality.