TW201721523A - Image-based circular and triangular traffic sign detection and recognition method utilizing convolution calculation in an image space to implement the circular and triangular traffic sign detection and recognition method - Google Patents

Abstract

An image-based circular and triangular traffic sign detection and recognition method is disclosed, which mainly utilizes convolution calculation in an image space to implement the circular and triangular traffic sign detection and recognition method. The implementation of the present invention is to utilize a method of calculating feature points based on convolution in an image space of an image processing technique, and to arbitrarily pick feature point pixels to calculate a circular formula and a trigonometric function formula; next, the present invention detects image regions of all possible circular and triangular traffic signs in an image; then, the present invention conducts sample feature comparison between the image regions of all possible circular and triangular traffic signs and existing road traffic sign image data, so as to detect and recognize all possible circular and triangular traffic signs in the image.

Description

Image type circular and triangular traffic sign detection and identification method

A method for detecting and recognizing circular and triangular traffic signs by using a convolution calculation of image space, especially an image area that utilizes all possible circular and triangular traffic signs that will detect the image, and the original road Traffic sign image data is used to compare the image type circular and triangular traffic signs detection and identification methods.

In the patented technology, in order to make the traffic signs on the road easier to identify for motorists, Mr. Lin Yongxing proposed a new patent “Road Traffic Warning Sign with Dynamic Effects” in the Republic of China in 1987 (announcement) No.: 00240849), its creation is to improve the warning signs in the traditional roads, with the light source of the light-emitting diodes, to play a clear and eye-catching effect on the fault warning or limiting the revelation. However, for car drivers, it is still necessary to pay attention to the road traffic signs on the roadside, and it is easy to ignore the road traffic warning signs on the roadside. In the 82nd year of the Republic of China, Mr. Chen Fudi and Mr. Chen Jintang proposed an invention patent “Voice Sign Voice Reminder” (Announcement No.: 00230812). Before the road was equipped with traffic signs, set up “flag code transmitter” in steam locomotives, etc. A "signal voice receiver" is installed on the vehicle to voice the content of the traffic sign corresponding to the code to achieve the purpose of alerting the driver. Although this invention can actively remind drivers of road traffic signs, it is necessary to set a "flag code transmitter" in front of each traffic sign, which will cost the traffic signs and maintain them regularly, and not every car will be installed. "Mark Voice Receiver".

It can be seen that there are still many shortcomings in the above-mentioned methods of use, which is not a good design, but needs to be improved.

In view of the shortcomings derived from the above-mentioned conventional methods, the inventors of the present invention have improved and innovated, and after years of painstaking research, they have successfully developed and completed the image detection method for circular and triangular traffic signs.

The main purpose of the present invention is to use a computer vision and image processing technology to apply the technology to a smart driving image monitoring system. Once a traffic sign is displayed on a road scene monitored by the camera, a notification message can be issued in time to remind the road driver. In addition, it can reduce and reduce traffic accidents or traffic penalties caused by the inability to identify road traffic signs in time.

The method for detecting and recognizing the image type circular and triangular traffic signs of the above invention is firstly carried out on a steam locomotive equipped with a driving recorder and an image analysis platform, while driving the recording and implementing the method on the working platform. There are triangular or circular traffic signs in the image when driving. On the system platform, digital image processing technology can be used to detect and identify the traffic signs through innovative triangle or circular traffic sign detection and identification methods. Finally, on the driving recorder, according to the driver's behavior needs, timely remind the driver of the road ahead.

The patented technology uses a driving recorder image device to capture a digital image of a computer, and then uses a digital image processing method and computer vision technology to detect whether a road scene scene has a triangular or circular road traffic sign. The invention is mainly divided into three parts: image convolution calculation, feature point group matching and traffic sign image comparison.

1. Image convolution calculation: Apply the image space convolution calculation formula in image processing, calculate the convolution value of each pixel, and find out that the first 20% of the convolution value is the feature point.

2. Feature point grouping point: use the concept of three sides of equal triangle to combine the paired triangle traffic signs Chi, then use the circular center point to the concept of the radius of the circle to the same length to combine round traffic signs.

3. Traffic Sign Image Comparison: The candidate triangle or circular image area is compared with the traffic sign image published by the Ministry of Communications using the image subtraction sample comparison method, and the error value needs to be less than 5%.

The method of the invention comprises calculating the image space of the computer digital image in the gray scale color space, and designing different convolution masks for the triangle and the circular traffic sign, such as a triangle or a circle with a successful pairing in the picture. The shape feature points can be used to compare the traffic sign images. If the comparison is successful, the identification results can be output and the motor vehicle driver of the road can be reminded in time.

101‧‧‧Computer digital image data

102‧‧‧Converted to grayscale color space

103‧‧‧ Calculating the formula of the convolution of a triangle

104‧‧‧Retrieve the triangular convolution feature points for group matching

105‧‧‧Comparative triangular traffic signs

106‧‧‧Output triangle traffic sign identification result

107‧‧‧ Calculate the circular formula of the circle

108‧‧‧Receiving circular convolution feature points for group matching

109‧‧‧Compared round traffic signs

110‧‧‧Output circular traffic sign identification result

201‧‧‧Triangular Convolution Formula Mask One

202‧‧‧Triangular Convolution Formula Mask II

203‧‧‧Triangular Convolution Formula Mask III

301‧‧‧Circular formula

302‧‧‧Circular Convergence Formula Mask II

303‧‧‧Circular formula formula mask three

304‧‧‧Circular formula formula mask four

1 is a flow chart of an implementation step of the technical method of the present invention; FIG. 2 is a set of three masks for calculating a triangular convolution formula.

Figure 3 is a set of four masks for calculating the circular convolution formula.

Referring to FIG. 1 , the image circular and triangular traffic sign detection and identification method of the present invention is to input the original image image into the format of the computer digital image data 101, and then perform conversion into the gray scale color space 102. , that is, divided into triangular traffic sign detection and circular traffic sign detection The large flow, triangle traffic sign detection is divided into the calculation triangle of the calculation triangle 103, the triangle rotation feature points are taken out for grouping 104, the triangular traffic sign data 105 is compared, the triangular traffic sign identification result 106 is output; the circular traffic sign detection is also Divided into a circular formula v for calculating a circle, the circular convolution feature points are taken out for grouping pairing 108, and the circular traffic sign data 109 is compared to output a circular traffic sign recognition result 110.

Next, there will be two main pairing and comparison processes, as shown in Figure 2. In the calculation of the triangle's convolution formula 103, the triangular convolution formula mask shown in Figure 2 is used. Mask two 202, triangular convolution formula mask three 203, the rotation formula of the image processing is executed, in the first pairing process, the calculated convolution value of each pixel point, taking the general quantized convolution value The first 20% of the pixels are feature points. If there are 1000 pixels in the picture, there will be 200 feature points in each group, and the three convolution masks will have three sets of candidate feature points. Then, the triangle feature points are grouped and paired, and the three sets of feature points are randomly paired. If each group has 200 feature points, the computer will perform 8 million comparisons, and each selected three feature points will be Need to meet the characteristics of the equilateral triangle, that is, the distance between the three sides is the same length, because the image processing needs to have the error tolerance value, so the distance error of the three side pixels of the three feature points is not more than five percent, and then the picture Three feature points in the successful pairing For the area that may be a triangular traffic sign, the second important comparison process is to compare the traffic signs of the triangle, and use the triangular traffic sign image data published by the Ministry of Communications to match the successfully paired triangular image data. Subtracting the sample comparison, subtracting the sum of the pixel values of the triangular image frame and subtracting the sum of the pixel values of the sample image to obtain an error value, and dividing the error value by the sum of the pixel values of the sample image is less than 5 percent. The triangle traffic sign is then executed to output the triangle traffic sign recognition result. In the calculation of the circular convolution formula, the circular convolution formula mask shown in Figure 3 is used. A mask 301, a circular convolution formula mask 302, a circular convolution formula mask 303, a circle The shape of the formula is masked by four 304, and the calculated convolution value is obtained for each pixel. The first 20% of the pixels with the highest convolution value after normal quantization are the feature points, and the four convolutions are covered. The cover will have four sets of candidate feature points, and then the round feature point grouping will be paired, and the four sets of feature points will be randomly paired. If each group has 100 feature points, the computer will perform 100 million comparisons. Each of the four feature points selected must conform to the circular feature. The coordinates of the four feature points are divided by four to obtain a center point, and the center point must be in the middle of the geometric position of the four feature points. And the position of the center point to the four feature points must be the same length. Since the image processing needs to have an error tolerance value, the pixel distance error ratio of the four feature points changed to the center point is not more than 5%. The second important The comparison process compares the four feature points of the successful picture matching with the circular traffic sign data, and uses the circular traffic sign image data published by the Ministry of Communications and the paired successful circular image frame data to perform image subtraction. kind This comparison is the same as the triangular alignment method of the triangle. The error value is divided by the sum of the pixel values of the sample image. If the error value is less than five percent, the circular traffic sign is obtained, and then the circular traffic sign recognition result is output. .

The image type circular and triangular traffic sign detection and identification method provided by the invention has the following advantages when compared with other conventional technologies:

1. The invention can actively remind the driver of the steam turbine driving on the road, so that the driver can improve the road use state in front of the police.

2. The invention uses the intelligent image type detection and identification method, and does not need to install any wireless message transmitting device on the traffic sign, so as to achieve the purpose of actively reminding the driver.

3. The invention can be combined with the use of a driving recorder to perform a value-added intelligent function, which can simultaneously record the condition of the road ahead, and can also detect and identify the road traffic sign in front.

The detailed description of the present invention is intended to be illustrative of a preferred embodiment of the invention, and is not intended to limit the scope of the invention. The patent scope of this case.

To sum up, this case is not only innovative in terms of technical thinking, but also has many of the above-mentioned functions that are not in the traditional methods of the past. It has fully complied with the statutory invention patent requirements of novelty and progressiveness, and applied for it according to law. Approved this invention patent application, in order to invent invention, to the sense of virtue.

101‧‧‧Computer digital image data

102‧‧‧Converted to grayscale color space

103‧‧‧ Calculating the formula of the convolution of a triangle

104‧‧‧Retrieve the triangular convolution feature points for group matching

105‧‧‧Comparative triangular traffic signs

106‧‧‧Output triangle traffic sign identification result

107‧‧‧ Calculate the circular formula of the circle

108‧‧‧Receiving circular convolution feature points for group matching

109‧‧‧Compared round traffic signs

110‧‧‧Output circular traffic sign identification result

Claims (8)

An image type circular traffic sign detection and identification method, the main steps thereof include: a. using the driving recorder image device to capture the original image image into a computer digital image; b. converting the captured computer digital image into gray scale Color space; c. perform a convolution formula, calculate the convolution value of each image point; d. judge the calculated convolution value, and calculate the first 20th pixel point of the convolution value The number of circular convolution masks is grouped as a feature point; e. The feature points are added and divided by the number of feature points to obtain the center point of the circle; f. The circular traffic sign and traffic that match the picture successfully The video data of the circular traffic signs published by the Ministry are compared with the image subtraction data to obtain the correct circular traffic signs. The method for detecting and identifying an image circular traffic sign according to claim 1, wherein the circle has four convolutional masks, and when the circular convolution formula is executed, each pixel is detected when the circular traffic sign is detected. Four convolution values can be calculated. The method for detecting and identifying an image circular traffic sign as described in claim 1, wherein the circle has four convolution masks, and the first twenty percent pixels of the calculated convolution value are allocated as four Group feature points, and the center point of the circle is in the middle of the four feature points, and the position distance from the center point to the four feature points is equal or the error is within 5 percent. The method for detecting and identifying an image circular traffic sign as described in claim 1, wherein the information of the circular traffic sign paired successfully with the image of the circular traffic sign published by the Ministry of Communications is subtracted In the right time, the circular traffic sign with the smallest error and the error value less than 5 percent is the correct circular traffic sign. An image type triangular traffic sign detection and identification method, the main steps of which include: Using the driving recorder image device to capture the original image into a computer digital image; a. Converting the captured digital image into a grayscale color space; b. Performing a convolution formula to calculate the convolution of each image point Value; c. judging the calculated convolution value, and grouping the first 20th pixel of the calculated convolution value into the feature points according to the number of convolution masks of the triangle; d. successfully pairing the pictures The triangular traffic sign is compared with the image data of the circular traffic sign published by the Ministry of Communications for image subtraction to obtain the correct triangular traffic sign. The method for detecting and identifying an image type triangular traffic sign according to claim 5, wherein the triangle has four convolution masks, and when the triangle convolution formula is executed, each pixel point can be calculated when detecting the circular traffic sign. Three convolution values. The method for detecting and identifying an image type triangular traffic sign according to claim 5, wherein the triangle has three convolution masks, and the first 20 pixels of the calculated convolution value are allocated as three sets of feature points. And the three side pixels of the three sets of feature points are equal in distance or within five percent of the error. The method for detecting and identifying an image-type triangular traffic sign as described in claim 5, wherein the image of the triangular traffic sign successfully paired with the image of the circular traffic sign published by the Ministry of Communications is compared with the image subtraction data. The triangular traffic sign with the smallest error and less than 5 percent error is the correct triangular traffic sign.