[0019] The embodiments of the present invention will be described in detail below with reference to the drawings: this embodiment is implemented on the premise of the technical solution of the present invention, and detailed implementation modes and specific operation procedures are given, but the scope of protection of the present invention is not limited to the following Mentioned examples.
[0020] Such as figure 1 As shown, the blind guide device of this embodiment includes: a black-and-white camera (ie, an external information acquisition module), a street sign detection module, a sound conversion module, and a peripheral headset. The black and white camera collects the information in front of the field of view and converts it into grayscale information and outputs it to the street sign detection module; the street sign detection module stores and processes the image information collected in the black and white camera, and obtains the area of the image related to the street sign information through the image processing method , Use the pattern recognition method to process the area where the sign is located and get the characters on the sign, find the corresponding digital audio signal of the character and output it to the sound conversion module; the sound conversion module converts the digital audio signal into an analog audio signal; the headset will The analog audio signal is converted into a sound signal and output to the blind; in this embodiment, the street sign detection module is implemented by TMS320DM642, and the sound conversion module is implemented by the AIC23 stereo digital signal codec.
[0021] The street sign detection module includes: a street sign information preprocessing unit, a street sign searching unit, a character segmentation unit, a feature code and a template recognition unit. Among them: the road sign information preprocessing unit converts the gray information obtained in the external information acquisition module into a gray matrix, and performs the preprocessing in sequence with the method of histogram equalization, binarization, and removal of dot noise. And output the result to the street sign finding unit. The sign finding unit is used to find the position and size of the sign in the preprocessed gray matrix, and output this information to the character segmentation unit. The character segmentation unit uses the gray value of most points in the street signs and the gray level difference of the dot matrix where the characters are located to obtain each character area, and perform denoising, smoothing, gradient sharpening, texture detection, and texture detection for each character area. Image rotation and normalization processing. Convert it into a standard size character dot matrix, and output the result to the feature code and template recognition unit. The feature code and template recognition unit is to obtain a serial number from the character dot matrix through feature recognition or template matching, obtain the corresponding character or number through the number, find the corresponding digital audio signal in the database, and output it to Sound conversion module.
[0022] The blind guide method of this embodiment includes the following steps:
[0023] Step 1. Classify and summarize the different types of street signs, and determine the fonts and fonts in the street signs. Classify and label different types of road signs, and estimate the position and size of the same type of road signs in Chinese and English;
[0024] Common road signs are divided into intersection signs and road signs. Intersection signs are generally used at intersections to indicate the road name corresponding to each direction after the intersection. The main feature is that there is an arrow in the middle, so during the recognition process You need to recognize the arrow first and then re-divide it in the four directions of up, down, left, and right. The road sign is used to indicate the road name and direction of the road (specifically refers to the direction of east, west, south, north). It can be divided into two lines. The upper line indicates Chinese characters and the lower line indicates English instructions. It is recognized during the recognition process. After going up and down the corresponding dividing line, it can be divided into multiple areas for identification.
[0025] Step 2: Reading in external information: The information in a certain scene in front of the outside world is converted into digital information through analog-to-digital conversion through a camera. Each frame of image corresponds to a matrix that expresses the gray information value of all pixels in the front. Specifically, the image is collected according to the YCbCr video format (where Y represents the brightness component, Cb represents the blue chroma component, and Cr represents the red chroma component), and the pure gray value is obtained after the Cb and Cr values are discarded. Store the row and column number of the corresponding pixel in a two-dimensional array. The images of different frames are correspondingly stored in different arrays;
[0026] Step 3. Preprocessing of street sign information: use histogram equalization, binarization, point noise removal, gradient sharpening, texture detection, and image rotation to analyze and process the external scene information in order to obtain the contour area of the street sign. Then analyze and classify the area where the street signs are located. If there are multiple street signs, the street sign with the largest area will be processed first. The specific process is: first, the image is equalized by histogram to obtain the average gray value of the entire image; for the first frame, the gray value is directly used as the threshold , Other frames use the target area obtained after the image processing of the previous frame to calculate the average gray value and set it as a threshold; the image is binarized through the intermediate threshold to obtain a black and white image and remove the dot noise, in order to achieve the edge of the character It is easy to distinguish, and gradient sharpening is required; in order to achieve the character itself without large information loss, texture detection is required; finally, the image is rotated as a whole to minimize the possible interference of the camera in the next step of recognition;
[0027]Step 4. Reading and recognizing characters: Segment the existing road signs according to their characteristics. After obtaining the range of each individual Chinese character, letter and number, use bilinear interpolation to split the image. Perform standard normalization. Finally, through the feature recognition method and the template matching method, the Chinese characters, letters and numbers are recognized to obtain the corresponding feature sequence, and the content is recorded. The specific process is:
[0028] (1) Perform character segmentation on the street sign information obtained in step 3 combined with the characteristics of the stored street signs, and find the position and edge of each character in the original image;
[0029] (2) Use bilinear interpolation to normalize the character matrix and convert it into a standard size character dot matrix to prepare for the next step of feature recognition or font matching;
[0030] (3) Perform feature recognition and template matching for the character dot matrix that has been standardized.
[0031] The feature recognition method is as follows: For a fixed-size number or character, it usually has the same feature. The features used in this embodiment include:
[0032] ①. Whether there is a long horizontal line at the top of the character, such as figure 2 shown in a;
[0033] ②. Whether there is a long horizontal line at the bottom of the character;
[0034] ③. Whether there is a long horizontal line in the middle of the character;
[0035] ④. Whether there is a long vertical crossing on the left or right side of the character, such as figure 2 shown in b;
[0036] ⑤. The number of different lines crossing from left to right in the middle of the character, such as figure 2 c shows;
[0037] ⑥. A series of characteristics such as the number of different line segments crossing from top to bottom in the middle of the character.
[0038] For different characters corresponding to different features, the corresponding characters can be identified by obtaining the character features.
[0039] The template matching method is as follows: because the characters on the road signs are standard characters, a standard character template is set, and the original standard characters are sampled with 16 points in the horizontal direction and 16 points in the vertical direction and stored in a standard template. The real-time image is sampled at 16 points horizontally and vertically, and a standardized size character lattice of 16*16 special points is obtained as the lattice to be measured. By comparing the dot matrix to be tested with the standard template dot matrix, the character template with the highest matching rate is regarded as the matching result.
[0040] (4) Compare the feature sequences obtained by the above two recognition methods. If the feature sequences obtained by the two methods are the same, it is considered that the correct recognition result has been obtained.
[0041] Step 5: Search for corresponding Chinese characters and letters in the database for the identified characteristic sequence, and at the same time search for the digital audio signal in the system database, and finally convert the audio signal into a sound signal for output.
[0042] In this embodiment, the ICETEK-DM642-C evaluation board produced by Beijing Ruitai Innovation Company is used as the development platform. This evaluation board is a low-power independent development platform. The on-board equipment mainly includes the TMS320DM642DSP chip of Texas Instruments (TI), independent HPI interface, 2 video ports, 4M synchronous dynamic memory, 32MFlash, AIC23 stereo digital signal codec, JTAG emulation interface, and 4 user indicators .
[0043] This embodiment is applied to assist the blind to obtain street sign information and convert it into sound, which is used for guiding and guiding functions of the blind when walking on the road. The method of this embodiment is simple and easy to implement, the hardware is convenient to carry, and the running speed is fast, which provides convenience for the daily life of the blind.
