WO2019184642A1

WO2019184642A1 - Method and system for determining positions of watch face hands

Info

Publication number: WO2019184642A1
Application number: PCT/CN2019/076213
Authority: WO
Inventors: 苏江; 范凤华; 黄树志
Original assignee: 广东乐芯智能科技有限公司
Priority date: 2018-03-26
Filing date: 2019-02-27
Publication date: 2019-10-03

Abstract

A method and system for determining the positions of watch face hands. The method comprises the following steps: step one, utilizing a camera of a photographing device to capture in real-time the watch face of a watch to produce a video; step two, the photographing device performs a recognition with respect to the video captured by the camera, the recognition being the recognition of images of the video; step three, the photographing device positions the images so that the images are positioned in a watch dial area; step four, the photographing device performs positioning processing with respect to watch hands of the watch dial area to obtain the hands; nd step five, the photographing device determines the angle at which the hands are offset from a preset point. By means of the described technical solution, first, the positions of the hands of the watch on the watch face can be calculated accurately and in real-time, secondly, no high requirement is placed on a processor of a mobile phone, the recognition is asynchronously processed, thus preventing apps of the mobile phone from lagging; moreover, the employment of a scheme of selectively ignoring image frames allows low-end mobile phones, even if incapable of quickly processing images, to perform smoothly.

Description

Method and system for determining surface pointer position

Technical field

The present invention relates to a method and system for determining the position of a surface pointer.

Background technique

At present, smart entity pointer watch calibration is more troublesome, requires visual recognition and manual adjustment. Our company has developed a method for taking photos with pictures, which gives the mobile phone a very accurate premise for the recognition of the picture.

technical problem

The photographed image belongs to the instantaneous action, and the picture is fixed. During the completion of the photographing process, the time is moving. Therefore, there may be some error in this process, and a large number of correction algorithms are needed to correct the time.

Technical solution

Based on the above technical problem, the present invention provides a method for determining the position of a surface pointer, which uses a mobile phone video capture to identify the current position of the positioning pointer from the data image, and transmits the current position and the angle to be adjusted to the watch via Bluetooth. The watch is adjusted to the correct position according to the instructions. Currently, the accuracy of the test is extremely high. The user's use is greatly facilitated and the user experience is improved.

The specific technical solutions of the present invention are as follows:

A method of determining a position of a surface pointer, the method comprising the steps of:

Step one, real-time shooting of the surface of the watch to obtain a video;

Step two, identifying a video captured by the camera, the identifying is extracting an image of the video;

Step 3: positioning the image to position the image in a dial area;

Step 4: performing positioning processing on the hands of the dial area to obtain a pointer;

In step five, the angle at which the pointer deviates from the preset point is determined.

Further, in the second step, the extracting is performed according to the processing speed of the processor. If the current image is not processed, the sent image is skipped and extracted, and the image is processed and then extracted. The first frame of image sent to the processor after completion.

Further, in the third step, the positioning comprises positioning preprocessing, and the positioning preprocessing comprises grading the image to generate a channel gray image from the four-channel image.

Further, in the third step, after the pre-processing, the Hough transform image processing is used, and the dial is positioned in the gray image according to the shape of the dial to obtain a positioning matrix, and the positioning matrix is used to cut the step. The image extracted in the second image is removed from the region outside the positioning matrix in the image, and the image in the range of the positioning matrix of the image extracted in the second step is retained, which is called a positioning image.

Further, in the step 4, the image is edged on the positioning image, so that only the line is left in the positioning image, and the line is searched by the Hough transform to filter the invalid line segment.

Further, the method for filtering the invalid line segments is as follows:

When there is only one pointer, the shortest distance H from the center point of the dial to the line segment and the length L of the line segment are determined, and the processor stores the first threshold and the second threshold:

If the H is smaller than the first threshold, and the L is greater than the second threshold, determining that the line is a pointer;

When the pointer has two, the shortest distance H from the center point of the dial to the line segment and the length L of the line segment are determined, and the processor stores the first threshold, the second threshold and the second threshold:

If the H is less than the first threshold, and the L is greater than the second threshold and less than the second threshold, determining that the line is an hour hand pointer;

If the H is smaller than the first threshold, and the L is greater than the second threshold, determining that the line is a minute hand pointer;

When there are three pointers, the shortest distance H from the center point of the dial to the line segment and the length L of the line segment are determined, and the processor stores the first threshold, the second threshold, the second threshold, and the second threshold:

If the H is smaller than the first threshold, and the L is greater than the second two threshold and less than the second three threshold, determining that the line is a minute hand pointer;

And if the H is smaller than the first threshold, and the L is greater than the second three threshold, determining that the line is a second hand pointer.

Further, in the step 5, a color point is preset in the dial, and the processor performs color recognition on the positioning image to determine the color point, and calculates the connection between the color point and the center point of the dial. The angle of the line and the pointer.

Further, repeating steps 2 to 5, processing a subsequent third threshold number of frame images and calculating an angle of the pointer from the color point to obtain an angle of the third threshold number, and taking an angle in which the probability of occurrence is the largest The pointer is offset from the angle of the color point.

A system for implementing the above method, the system comprising: a smart watch and a smart phone;

The smart watch includes a physical pointer, a driving device for controlling rotation of the physical pointer, and an MCU for controlling the driving device;

The smart phone includes a processor and a camera;

The camera can perform real-time shooting on the surface of the watch to obtain a video;

The processor includes an image processing unit, the image processing unit may identify the video, and the identifying is to extract an image of the video for recognition;

The processor also includes a computing unit that can calculate the position of the pointer to derive an angle of the pointer from the reference point.

Beneficial effect

Through the above technical solution, one can accurately and real-time calculate the position of the pointer of the watch on the surface, and secondly, the processor requirements of the mobile phone are not high, and the recognition adopts asynchronous processing, so that the mobile APP does not change the card. And the use of selective ignore image frame mode, so that low-end mobile phones, even if you can not quickly process the image, can also be very smooth.

Embodiments of the invention

The technical solutions of the present invention are described in detail below.

The invention only needs to solve the algorithm problems of image recognition, noise coming over and time calibration, and the optimization problem which is different for the requirements of the mobile phone processor.

The extraction of the surface pointer already has the extraction of the image. However, due to the problem of light and color, the extraction of the image is difficult to determine the position of the pointer more accurately, and the shooting of multiple images is due to the time interval of shooting. Causes the pointer to have a very large chance of moving around.

The invention adopts the method of video extracting pictures, the video has continuity, and the picture of many frames can be extracted in one second of the video, and when the several frames are processed, the position of the pointer is substantially unchanged.

Determining the position of the surface pointer, in addition to the smart watch, requires the use of an external device such as a smart phone, and requires the smart phone to shoot a unit, the smart watch including a physical pointer, a driving device for controlling the rotation of the physical pointer, and an MCU for controlling the driving device;

The smart phone includes a processor and a camera;

The specific implementation is as follows:

First, the surface of the watch is photographed in real time using a photographing device such as a mobile phone, and of course, a mobile phone having an IOS system to obtain a video, and then the video is processed.

When processing video, because the number of frames captured by the video is large, it is limited to the processing power of the processor. The most ideal state is that the processor processes and calculates each frame of image, but in fact, when processing the image, It is very likely that the processor has not processed the current image when the next frame image is sent. At this time, the processor can ignore the next frame image sent until the processor processes the current image.

Of course, it is also possible to perform forced processing. For example, according to the law, the image currently being processed is numbered as the No. 0 image, and the subsequent images are numbered according to the natural number, and the frame image of the multiple of 2, the multiple of 3, or the multiple of 4 is processed. . If the current image is not processed, the current processing can be forcibly stopped, and the image sent by the rule can be processed.

After extracting the image, in order to better locate the dial, reduce the interference of other factors (such as light, color), grayscale the image, and generate a 1-channel grayscale image from the original 4-channel image, which can reduce the Mobile phone performance requirements. Gaussian filtering is performed on the grayscale image to smooth the grayscale image.

Due to the particularity of the dial and the specific geometry, the Hough transform image processing can locate the dial. For example, when the disc is square, a positioning matrix can be located, and the range in the positioning matrix is the dial. range. Or when the dial is circular, a matrix can also be located, and the range within the matrix is the range of the dial. Even when positioning a square matrix when the dial is circular, you can use all black or all white to fill the gap between the circles.

The extracted image is clipped using the positioning matrix or the edge of the positioning matrix (the image is not the grayscale image described above, and the second is the color image extracted from the video) such that the colored image has only the region of the dial. The cropped image can be defined as a positioning image.

Then, the positioning image is edged, so that only the lines are left in the image, and then the Hough line transformation is used to find the straight line. At this time, there are many straight lines to be searched, and the invalid line segments need to be filtered out. The filtering method for invalid line segments is as follows:

Invalid filtering is to determine that the shortest distance from the center point of the dial to the line segment is less than a specific threshold (adjusted according to the actual situation), and that the length of the line segment is greater than or equal to a certain threshold (determined according to the distance of the dial pointer) is considered valid. of. Otherwise, the task line is invalid.

Of course, in the actual operation process, there may be two pointers coincident or substantially coincident (the angle between the two pointers does not exceed 2 degrees or less), then the mobile phone sends a signal to the watch, so that the two pointers of the watch are Any of the pointers are rotated a certain angle first.

At the time of shooting, most of the time, the mobile phone can not be photographed correctly. In order to fit the watch, it is not necessarily a correct scan. You need to find a reference point for the pointer, such as the reference point at 12 o'clock. We can select Set a color point in the direction of 12 o'clock, such as red.

The processor uses color recognition to judge the position of the color point (red dot), and confirms that the red dot algorithm is based on the RGB value of the image pixel (0-255), R minus G is greater than or equal to 80 and lacks R minus B. Greater than or equal to 80 is considered a red dot.

According to the length and angle of the line segment, the minute hand and the hour hand are distinguished. The second hand can also distinguish the second hand. Finally, the angle of each hand deviates from 12 o'clock by the center point.

The extracted images are all calculated once. After one calculation, the subsequent transmitted frame images are further calculated. For example, 10 times are calculated, and an angle with the highest probability of occurrence in 10 times is taken as the angle at which the pointer deviates from the color point. For example, in 10 times, there are 3 times 14 degrees, 4 times 15 degrees, 3 times 16 degrees, then 15 degrees is the angle at which the pointer deviates from the color point.

It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not limited thereto; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that The technical solutions described in the foregoing embodiments are modified, or the equivalents of the technical features are replaced. The modifications and substitutions do not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

A method of determining a position of a surface pointer, the method comprising the steps of:

Step one, real-time shooting of the surface of the watch to obtain a video;

Step two, identifying a video captured by the camera, the identifying is extracting an image of the video;

Step 3: positioning the image to position the image in a dial area;

Step 4: performing positioning processing on the hands of the dial area to obtain a pointer;

In step five, the angle at which the pointer deviates from the preset point is determined.
The method for determining the position of a surface pointer according to claim 1, wherein in the second step, the extracting is performed according to a processing speed of the processor, and if the current image is not processed, the sending is performed. The image is skipped without being extracted for recognition until the first frame image is sent to the processor after the current image is processed and the extraction process is completed.
A method for determining a position of a surface pointer according to claim 1, wherein in said step three, said positioning comprises positioning preprocessing, said positioning preprocessing comprising grading an image to make it A channel of grayscale image is generated from a four-channel image.
A method for determining a position of a surface pointer according to claim 3, wherein in said step three, after pre-processing, Hough transform image processing is employed, and the grayscale image is paired according to the shape of the dial. The dial is positioned to obtain a positioning matrix, and the positioning matrix is used to crop the image extracted in the second step, and the area outside the positioning matrix in the image is removed, and the positioning matrix of the image extracted in the second step is retained. The image is called a positioning image.
The method for determining the position of a surface pointer according to claim 4, wherein in the step four, image edge processing is performed on the positioning image, so that only the line is left in the positioning image, The transformation searches for a line and filters the invalid line segments.
A method for determining a position of a surface pointer according to claim 5, wherein the method of filtering the invalid line segments is as follows:

When there is only one pointer, the shortest distance H from the center point of the dial to the line segment and the length L of the line segment are determined, and the processor stores the first threshold and the second threshold:

If the H is smaller than the first threshold, and the L is greater than the second threshold, determining that the line is a pointer;

When the pointer has two, the shortest distance H from the center point of the dial to the line segment and the length L of the line segment are determined, and the processor stores the first threshold, the second threshold and the second threshold:

If the H is less than the first threshold, and the L is greater than the second threshold and less than the second threshold, determining that the line is an hour hand pointer;

If the H is smaller than the first threshold, and the L is greater than the second threshold, determining that the line is a minute hand pointer;

When there are three pointers, the shortest distance H from the center point of the dial to the line segment and the length L of the line segment are determined, and the processor stores the first threshold, the second threshold, the second threshold, and the second threshold:

If the H is less than the first threshold, and the L is greater than the second threshold and less than the second threshold, determining that the line is an hour hand pointer;

If the H is smaller than the first threshold, and the L is greater than the second two threshold and less than the second three threshold, determining that the line is a minute hand pointer;

And if the H is smaller than the first threshold, and the L is greater than the second three threshold, determining that the line is a second hand pointer.
A method for determining a position of a surface pointer according to claim 6, wherein in the step five, a color point is preset in the dial, and the processor performs color recognition on the positioning image. The color point is determined, and the angle between the color point and the center point of the dial is calculated.
A method for determining a position of a surface pointer according to claim 7, wherein steps 2 to 5 are repeated to process a subsequent third threshold number of frame images and calculate an angle at which the pointer deviates from the color point. An angle of the third threshold number is obtained, and an angle in which the probability of occurrence is the largest is taken as the angle at which the pointer deviates from the color point.
A system for implementing the method of any of the preceding claims 1-8, the system comprising: a smart watch and a smart phone;

The smart watch includes a physical pointer, a driving device for controlling rotation of the physical pointer, and an MCU for controlling the driving device;

The smart phone includes a processor and a camera;

The camera can perform real-time shooting on the surface of the watch to obtain a video;

The processor includes an image processing unit, the image processing unit may identify the video, and the identifying is to extract an image of the video for recognition;

The processor also includes a computing unit that can calculate the position of the pointer to derive an angle of the pointer from the reference point.