JP2021013657A - Training device for pitching and the like, computer program and method for the same - Google Patents

Abstract

To provide a training device for pitching and the like which is sturdy, hardly broken, has the small manufacturing and installing costs, and in which balls can be easily recovered and reused, and provide a computer program and method for the same.SOLUTION: A training device includes a classifier of collision of a ball 6 or object (hereinafter referred to as "ball or the like") to a surface (that may be or may not be a plane) applied with a feature indicating a target range, and displays the classification by the classifier every time a user makes the ball or the like collide with the surface. The classifier performs classification on the basis of an image (static image or moving image) imaged before and after the collision. The classifier determines the timing of the collision or the classification result on the basis of a sensor 4 which senses the vibration, contact, contact sound or the combination of them in the entire region or partial region of the surface. The classifier classifies the collision with the image captured at the timing of sensing the collision to the surface.SELECTED DRAWING: Figure 1

Description

The present invention relates to a practice device for pitching baseball and handball, shooting soccer, and shooting tennis and volleyball. It also relates to computer programs and methods for such devices.

Conventionally, such practice devices have a bag with a plurality of physical openings, a plurality of grids whose lids (each numbered) are opened by an impact, and a complicated sensor device configuration (a complex sensor device configuration near the ball landing). It was realized by the following Patent Document 1 etc.). Although they have a certain level of physical robustness, they are easily damaged physically or electronically, and it is difficult to reuse (re-throw, etc.) the ball with a simple mechanism, so installation such as pitching practice etc.・ Operational costs (practice equipment production costs, installation and operation costs, ball collection time, etc.) were high.

Application 2001-816

Therefore, an object to be solved by the present invention is to provide a pitching practice device that is robust, hard to break, low in manufacturing and installation costs, and easy to collect and reuse balls, and a computer program and method for that purpose.

The present invention
A user collides with a ball or the like with a classifier for collision with a ball or an object (hereinafter referred to as "ball or the like") on a surface (which may or may not be a flat surface) having a feature indicating a target range. Provided is a practice device for displaying the classification by a classifier each time the ball is made.
: "Surface with features indicating the range of goals" includes strikes such as baseball (and shadows of batters of various heights on the left and right), goals such as soccer and handball (for adults and children), tennis and volleyball, etc. It may include a sturdy surface that depicts the virtual target of, and a line that divides the range of strikes, goals, targets, etc. into vertical M frames, horizontal N frames, concentric circles, and vertical, horizontal, and diagonal angles within the concentric circles. , No goal, ball, off-target, etc. may be included. It also includes curved surfaces and polyhedra of metal images that imitate humans, animals, vehicles, etc. that clearly indicate equidistant lines from the target part. When this is a flat surface, in many cases, the ball or the like bounces off the throwing user, so that the ball can be easily repaired and reused. Further, the rebound coefficient can be adjusted by adjusting the material of 1 in FIG. 1 and the properties of the spring as shown in 2.
: "Display, etc." includes display after communication to on-site equipment and remote equipment, display after integration / analysis / diagnosis of classification results by computer, display by voice, light, characters, etc., display at collision location (deep part) Includes LED blinking and laser irradiation on the surface).
> In many cases, this makes it difficult for equipment to be damaged, makes it easier to reuse balls, etc., and reduces the time and effort required for judgment and evaluation when practicing training and sports, including on-site / distant spectators. It will also be possible to produce in a more enjoyable way.

The present invention also provides a practice device in which the classifier classifies based on images (still images or moving images) captured before and after a collision.
: This includes not only classification from one still image or series of still images, but also classification by moving images, and classification by pixel color distribution analysis from still images and moving images and deep learning technology (supervised learning). Including. When the target is a geometric shape, a simple judgment or course classification can be made by calculating with a computer whether the collision center coordinates of the pixel set indicating the color of the sphere are within the theoretical range of the geometric shape. You can also do it.
> This makes it possible to classify the detailed position of the collision (for example, the position that evaluates whether or not the goal has been reached), speed, angle, etc. by using a small number of physical sensors (one acceleration sensor, etc.) on the "face" side. become.
> In addition, since imaging has few moving parts, it is resistant to rain and deterioration over time, and in the present age when digital cameras have become widespread, maintenance and operation costs are low, which is advantageous.

The present invention also
Provided is a practice device in which the classifier determines the timing or classification result of the collision based on a sensor that senses vibration, contact, contact sound, or a combination thereof in a whole area or a partial area of a surface.
: Vibration and contact can be detected by acceleration sensors arranged for each whole or partial area, and contact sound can be detected and classified by several microphones.
> The timing of collision is necessary for determining the unit of classification, and the classification result is necessary for the content of the determination. In this way, it is possible to make a multifaceted and robust judgment.

The present invention also
The classifier provides a practice device for classifying collisions based on an image captured at the timing when a collision with a surface is detected.
> As a result, it is possible to save a set of images to be learned by giving a teacher signal in the case of deep learning or the like, and to reduce the learning cost.

We also provide computer programs and methods used for such practice devices.

It is a system schematic diagram of one Example of this invention. It is the schematic of one Example of the baseball pitching practice of this invention. It is the schematic of the soccer goal kick practice device by one Example of this invention. It is the schematic of the tennis shot practice apparatus by one Example of this invention. It is a schematic diagram of a part of the catch ball practice device (example of a polyhedron) according to one embodiment of the present invention.

Hereinafter, various embodiments will be described with reference to examples, but the present invention is not limited thereto.
In any case, the classifier is configured so that the classification can be instantly determined based on the collision image caused by the user's kick or pitch. At this time, imaging and image input to the classifier are performed only when there is collision vibration on the stainless steel plate, the ground, or the doll, and by sending the classification result to a personal computer or the like and displaying it, the sensor on the goal side has collision vibration. It can be a sensor only.
In another embodiment, the imaging timing can be determined only by the audio signal from the microphone near the stainless steel plate, and the deep learning result with the moving image and the teacher signal specifying the timing on the moving image based on the determination of the collision timing. It is also possible to perform automatic classification from only moving images using only.

FIG. 1 is a schematic diagram of a system according to an embodiment of the present invention.
The stainless steel plate 1 is a stainless steel plate (6.4 mm) in which a strike zone, a grid of goals, a concentric circle for practice, etc. are drawn on the painted surface. The thickness and dimensions vary depending on the application (baseball, soccer, tennis, etc.).
The spring 2 fixes the stainless plate 1 to a wall (concrete, a building, a support for fixing, etc.) (the actual required number is omitted).
The wall 3 is made of concrete, a building, a support for fixing, or the like (may be a skeleton).
The acceleration sensor 4 is an acceleration sensor for detecting the collision of the ball 6 with the stainless plate 1, and the number of acceleration sensors 4 required for reliable detection is installed.
The pitcher 5 (in the case of a baseball pitcher in the figure) may be a pitcher for machine learning (deep learning, etc.) of a practice device, or a practitioner (user) after machine learning. .. The machine learning results can be copied as electronic data when mass-producing the practice device.
The ball or the like 6 is emitted from the pitcher 5 and collides with the stainless steel plate 1.
The communication 7 is a communication from the acceleration sensor 4 to the digital camera 8, conveys the timing of the impact of the ball 6 or the like, and connects it to imaging, strike determination, course classification, and the like.
The digital camera 8 is fixed in the imaging direction 9 by the installation jig 8a, and photographs the entire stainless steel plate 1 (including the state of the collision of the spheres 6) in response to a signal indicating a collision from the acceleration sensor 4. , The image information 10 is transmitted to the computer 11 (USB wiring, WIFI connection, etc.).
The putter 11 uses the image information 10 to perform machine learning based on the teacher signal 13 from the visual judge 12 at the time of development, and performs judgment / course classification based on the machine learning result at the time of operation. The speaker informs the user (practitioner) of the judgment and course classification.

FIG. 2 is a schematic diagram of an embodiment of baseball pitching practice.
The drawn grid 1a depicts 5 × 5 types of intrusion courses in the range of the baseball strike zone, and this line is also used in machine learning. In another embodiment (not shown), the balls are also classified according to the boundary line in the displaced direction.
The drawn image 1b imitates the left batter of baseball, and gives a sense of presence to the pitching practitioner and can also be a part of image information for machine learning. The same applies to the right batter 1c.
The stainless steel plate 1 is smaller than that of soccer (FIG. 3) or tennis (FIG. 4), and can create a compact driving range.

FIG. 3 is a schematic view of a soccer goal kick practice device according to an embodiment of the present invention.
A welded green stainless steel plate 1 (9m4m x 3.2mm) with a size that allows the front surface of the front inner diameter (about 7.32m x 2.44m) of a soccer goal to be drawn with a margin, and a concrete wall 3 (15m x 8m) with a spring 2. The acceleration sensor 4 is provided on the back surface at an appropriate density and connected to the computer 11.
Draw a grid line that clearly shows the outer frame and inner diameter of the soccer goal and divides the inner diameter into 15 vertical 3 horizontal 5 frames, and actually hit 10,000 or more volley shoots etc. to impact the stainless steel plate 1 (acceleration sensor) Take a still image at each point in time (judged by 4), and manually determine whether it is a goal, if it is a goal, which area is divided into 15 areas, or whether it is a goal that hits the grid and bounces off. A teacher signal 13 such as is given so that a judgment can be made by deep learning when a new goal is reached. Here, when it is not possible to visually judge the person 12, the result of the experiment applied to the actually created grid can be used as the teacher signal 13, and the result of the physical simulation or the like by another computer is used as the teacher signal 13. You can also do it.
The result of the judgment is that the LED etc. (not shown) embedded in the center of each grid of the stainless steel plate is turned off to indicate that the goal has been reached in the grid, and the bulletin board beside the site or the WEB site that can be confirmed on a smartphone. To be displayed in. The LED of the grid that has been turned off is programmed in the computer 11 or the like so as to be reset and turned on in response to a start signal of another game.
Handball goals, baseball strike zones, etc. can be configured in the same way, although the size will be smaller. In the case of baseball, for example, when the upper right strike, the lower left strike, the lower right ball, and the upper left strike are thrown separately on a predetermined grid or course, it is regarded as one out to produce a game. You can also do it.

FIG. 4 is a schematic view of a part of a tennis shot practice device according to an embodiment of the present invention. In the case of volleyball, the system can be configured in the same way.
In the case of volleyball, tennis, etc., the whole is not shown, but there are some concentric circles in the space where there is a ground that detects impact (not shown) and a cushion-shaped doll that detects impact (see Fig. 5). You can practice and play games after expressing the expressed target etc. (1d, 1e, etc.) and doing machine learning etc. like soccer and baseball.

FIG. 5 is a schematic view of a part of a catch ball practice device (example of polyhedron 1) according to an embodiment of the present invention. In addition to baseball catch balls, it can be used when aiming at the front of receiver 1 etc. (or conversely, a course avoiding the front) in volleyball or tennis. Considering the convenience of collecting the ball, it is not always necessary to make it a polyhedron, but in the case of a polyhedron, it may be easier to collect the ball when the ball hit at an angle hits the polyhedron vertically.

1 Stainless steel plate (6.4 mm) with a strike zone, a grid of goals, and concentric circles for practice drawn on top of the paint.
1a A grid depicting a baseball strike zone and an invasion course 1b A statue imitating a baseball left batter 1c A statue imitating a baseball right batter 1d Concentric circles of targets for baseball practice
1e Targets for tennis practice are shown by concentric circles and deviation angles 1f Targets for baseball catch balls are shown by concentric circles and deviation angles 2 Stainless steel plates are used for walls (concrete, buildings, fixing columns, etc.) ) (The actual required number is omitted)
3 walls (concrete, buildings, fixing columns, etc.)
4 Accelerometer for detecting the collision of the ball with the stainless steel plate 1 5 Pitcher for opportunity learning of the training device or pitcher after machine learning (user)
6 Ball, etc. 7 Communication from the accelerometer 4 to the camera (timing when there was an impact)
8 Digital camera 8a Digital camera installation jig 9 Imaging direction of digital camera 10 Image information from digital camera 11 Computer (equipped with display, input device for teacher signal and play reset signal, judgment speaker)
12 Visual judge of strike ball course (corresponding grid) (teacher signal provider)
13 Teacher signals, etc.

Claims (6)

A user collides with a ball or the like with a classifier for collision with a ball or an object (hereinafter referred to as "ball or the like") on a surface (which may or may not be a flat surface) having a feature indicating a target range. A practice device that displays the classification by a classifier each time it is made. The practice device according to claim 1, wherein the classifier classifies based on images (still images or moving images) captured before and after the collision. The practice device according to claim 1, wherein the classifier determines the timing or classification result of the collision based on a sensor that senses vibration, contact, contact sound, or a combination thereof in an entire region or a partial region of a surface. The practice device according to claim 2 and 3, wherein the classifier classifies the collision based on the image captured at the timing when the collision with the surface is detected. A computer program used for the practice device according to any one of claims 1 to 4. The method using the practice device according to any one of claims 1 to 4.

Images