Image processing recognition algorithm based on visual acquisition

Abstract

The invention relates to the technical field of visual image processing and identification, and discloses an image processing and identification algorithm based on visual acquisition, an image acquisition module performs image acquisition on a target area to obtain original image data, the original image data is sent to an identification processing module, and then target standard image data is obtained. And the image acquisition module identifies the image of the field area by utilizing algorithm analysis, and judges whether the field image is consistent with the target image or not. And after the image of the field area is consistent with the target standard image data, the recognition processing module performs image filtering and image segmentation operation by using the field image information, and then performs comparison to complete the recognition process. According to the invention, the original image data is obtained during image acquisition and is sent to the identification processing module, and then the target standard image data is obtained after the original image data passes the standard detection. Through establishment of target standard image data, the reliability of subsequent detection can be improved, and the range of image acquisition can also be enlarged.

Description

Technical Field

[0001] The present invention relates to the technical field of visual image processing and recognition, and in particular to an image processing and recognition algorithm based on visual acquisition. Background Art

[0002] In recent years, with the continuous development of artificial intelligence technology, image processing technology has gradually moved towards intelligence. The principle of image recognition is to use computer technology and mathematical methods to pre-process the obtained target image information and extract features. Then the classifier classifies the image into corresponding categories, and finally obtains the recognition result by matching with the stored information.

[0003] After searching, the Chinese patent number CN106033537A discloses a visual recognition device for robot positioning; it includes an image acquisition module, an image processing module with a pre-stored standard image, and a servo control system; its visual recognition method includes the following steps: the first step, image acquisition; the second step, image enhancement; the third step, image filtering, by filtering the image completed in the second step through a median filtering algorithm; the fourth step, edge detection, the fifth step, image segmentation, and the sixth step, comparative recognition; the visual recognition device for robot positioning of the present invention and its recognition method improve the anti-interference and accuracy of the visual recognition system by improving the image processing method; and the algorithms involved in its image processing are all relatively mature algorithms in the prior art, and the present invention can directly quote them to ensure the accuracy of the system. In addition, through a series of previous preprocessing, the interference in the acquisition process is reduced, and finally the comparative recognition can accurately identify the image and improve the accuracy.

[0004] After searching, the Chinese patent number CN110619336A discloses a product recognition algorithm based on image processing, which mainly includes four parts: image preprocessing, feature extraction, classification recognition, layer number determination and quantity statistics. The present invention can well judge the type of goods and the change of the quantity of goods in the video data. It adopts the feature extraction method based on KL transformation and Laplace mapping, and the decision algorithm based on error correction coding SVM to improve the accuracy of product recognition; at the same time, it adopts the visual ranging method to judge the change of the quantity of goods, obtains a cheap and simple visual system, and reduces the time cost and economic cost of the system.

[0005] However, when the working environment of the existing device changes or the target object changes, the traditional image processing and recognition algorithm imitates the human visual ability to perceive the external environment, but its own intelligence, self-adaptation and self-learning capabilities are general. Therefore, when facing external changes, the image processing method is relatively complicated and cannot form a good recognition system. Based on this, the present invention designs an image processing and recognition algorithm based on visual acquisition to solve the above problems. Summary of the invention

[0006] The purpose of the present invention is to provide an image processing and recognition algorithm based on visual acquisition, which solves the problem of poor systematicity in the background technology.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0008] An image processing recognition algorithm based on visual acquisition includes the following steps:

[0009] Step S1: The creator collects basic features of the target image, stores them in a data storage module, and establishes a target image feature library, in which the feature values ​​of the target image are stored;

[0010] Step S2: The image acquisition module acquires images of the target area, obtains original image data, and sends it to the recognition processing module;

[0011] Step S3: After receiving the original image data, the recognition processing module detects the original image data. If it is qualified, it jumps to step 4; if it is not qualified, it jumps to step 2;

[0012] Step S4: After the original image data passes the standard detection, image enhancement is performed, including histogram enhancement, histogram averaging and contrast enhancement, to obtain target standard image data;

[0013] Step S5: performing algorithm processing on the three-dimensional image contour data of the target standard image data to obtain center and edge contour information;

[0014] Step S6: The image acquisition module uses an algorithm to analyze and identify the image of the scene area, and determines whether the scene image is consistent with the target image. If they are consistent, it jumps to step 7; if not, it jumps to step 2;

[0015] Step S7: After the image of the on-site area is consistent with the target standard image data, the recognition processing module uses the on-site image information to perform image filtering and image segmentation operations, wherein the image filtering algorithm is a median filtering algorithm, and the image segmentation method is a binary segmentation method;

[0016] Step S8: Match the feature values ​​of the scene image with the feature values ​​in the image feature library, and compare them with the standard images pre-stored in the image processing module through the module matching method to complete the recognition process. If the match is successful, it is judged as a target, if the match fails, it is judged as a non-target.

[0017] Preferably, the algorithm includes a storage module, a power module, an image acquisition module, an identification processing module, a fault processing module and a power protection module.

[0018] Preferably, step S4 comprises the following steps:

[0019] Step S4.1: enhancing the image using a grayscale density function;

[0020] Step S4.2: Expand the grayscale range of the image and make the grayscale with a smaller grayscale frequency larger, that is, make the grayscale histogram tend to be consistent within a larger dynamic range;

[0021] Step S4.3: Modify the grayscale of each pixel of the input image according to certain rules, thereby changing the dynamic range of the image grayscale.

[0022] Preferably, step S8 includes the following steps: storing the characteristic value of the extracted image in data variable A, and using the data of data variable A to compare the data feature library, normalizing the data variable B, and judging whether it is a matching value of the target image based on the variance of data variable A relative to data variable B by the size of the variance.

[0023] Preferably, in step S2, the original image data is filtered, grayed, and histogram equalized in sequence to obtain the standard image data for subsequent comparison.

[0024] Preferably, in step S2, the target object is photographed by a digital camera and converted into an image signal, and then the image signal is converted into digital information by an image acquisition card and sent to the recognition processing module;

[0025] The power module provides power supply for the entire device;

[0026] The recognition processing module performs compliance detection on the original image data.

[0027] Preferably, in step 5, the standard image data is binarized to obtain center and edge contour information to facilitate subsequent recognition processing.

[0028] Compared with the prior art, the beneficial effects achieved by the present invention are:

[0029] 1. The present invention obtains raw image data through the cooperation of the recognition processing module of the image acquisition module, sends it to the recognition processing module, and then performs image enhancement after the raw image data passes the standard detection, including histogram enhancement, histogram averaging and contrast enhancement, to obtain the target standard image data. The establishment of the target standard image data can improve the reliability of subsequent detection and also expand the scope of image acquisition.

[0030] 2. In the present invention, the image acquisition module uses algorithm analysis to identify the image of the scene area, and the recognition processing module uses the scene image information to perform image filtering and image segmentation operations to further identify the similarities and differences between the scene image and the target image, facilitate matching, and improve matching accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is the algorithm flow chart of the present invention;

[0032] Figure 2 The figure is a flow chart of the image enhancement of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] Embodiment 1;

[0035] See also Figure 1-Figure 2 In an embodiment of the present invention, an image processing and recognition algorithm based on visual acquisition includes the following steps:

[0036] Step S1: The creator collects basic features of the target image, stores them in a data storage module, and establishes a target image feature library, in which the feature values ​​of the target image are stored;

[0037] Step S2: The image acquisition module acquires images of the target area, obtains original image data, and sends it to the recognition processing module;

[0038] Step S3: After receiving the original image data, the recognition processing module detects the original image data. If it is qualified, it jumps to step 4; if it is not qualified, it jumps to step 2;

[0039] Step S4: After the original image data passes the standard detection, image enhancement is performed, including histogram enhancement, histogram averaging and contrast enhancement, to obtain target standard image data;

[0040] Step S5: performing algorithm processing on the three-dimensional image contour data of the target standard image data to obtain center and edge contour information;

[0041] Step S6: The image acquisition module uses an algorithm to analyze and identify the image of the scene area, and determines whether the scene image is consistent with the target image. If they are consistent, it jumps to step 7; if not, it jumps to step 2;

[0042] Step S7: After the image of the on-site area is consistent with the target standard image data, the recognition processing module uses the on-site image information to perform image filtering and image segmentation operations, wherein the image filtering algorithm is a median filtering algorithm, and the image segmentation method is a binary segmentation method;

[0043] Step S8: Match the feature values ​​of the scene image with the feature values ​​in the image feature library, and compare them with the standard images pre-stored in the image processing module through the module matching method to complete the recognition process. If the match is successful, it is judged as a target, if the match fails, it is judged as a non-target.

[0044] The working principle of the embodiment of the present invention is as follows: the creator collects the basic features of the target image, stores them in the data storage module, and establishes the target image feature library, in which the feature values ​​of the target image are stored. The image acquisition module performs image acquisition on the target area, obtains the original image data, and sends it to the recognition processing module. After receiving the original image data, the recognition processing module detects the original image data. After the original image data passes the standard detection, image enhancement is performed, including histogram enhancement, histogram averaging and contrast enhancement, to obtain the target standard image data. The three-dimensional image contour data of the target standard image data is processed by algorithm to obtain the center and edge contour information. The image acquisition module uses algorithm analysis to identify the image of the scene area and judges whether the scene image is consistent with the target image. After the image of the scene area is consistent with the target standard image data, the recognition processing module uses the scene image information to perform image filtering and image segmentation operations, wherein the image filtering algorithm is the median filtering algorithm, and the image segmentation method is the binary segmentation method. The feature value of the scene image is matched with the feature value in the image feature library, and compared with the standard image pre-stored in the image processing module through the module matching method to complete the recognition process.

[0045] Embodiment 2;

[0046] See also Figure 1-Figure 2 , step S4 comprises the following steps:

[0047] Step S4.1: enhancing the image using a grayscale density function;

[0048] Step S4.2: Expand the grayscale range of the image and make the grayscale with a smaller grayscale frequency larger, that is, make the grayscale histogram tend to be consistent within a larger dynamic range;

[0049] Step S4.3: Modify the grayscale of each pixel of the input image according to certain rules, thereby changing the dynamic range of the image grayscale.

[0050] Step S8 includes the following steps: storing the feature value of the extracted image in data variable A, and using the data of data variable A to compare the data feature library, normalizing the data variable B, and judging whether it is a match value of the target image based on the variance of data variable A relative to data variable B.

[0051] The working principle of the embodiment of the present invention is: image enhancement includes histogram enhancement, histogram equalization and contrast enhancement. The histogram of an image is an important statistical feature of the image. It can be considered as an approximation of the grayscale density function of the image, and therefore has corresponding statistical features, among which the most important feature is the grayscale density function. The grayscale histogram of a natural image with histogram equalization usually has a large frequency in the low-value grayscale interval. In order to make the image clear, the grayscale range of the image can be widened, and the grayscale with a small grayscale frequency can be increased, that is, the grayscale histogram can be made consistent within a larger dynamic range. Contrast enhancement is to modify the grayscale of each pixel of the input image according to certain rules, thereby changing the dynamic range of the image grayscale, which can expand the grayscale dynamic range or compress it, or segment the grayscale, compress it in a certain interval and expand it in another interval according to the characteristics and requirements of the image.

[0052] Embodiment 3;

[0053] See also Figure 1-Figure 2 ,The algorithm includes a storage module, a power module, an image acquisition module, an ,identification processing module, a fault processing module and a power protection module.

[0054] In step S2, the original image data is filtered, grayed, and histogram equalized in sequence to obtain standard image data for subsequent comparison.

[0055] In step S2, the target object is photographed by a digital camera and converted into an image signal, and then the image signal is converted into digital information by an image acquisition card and sent to the recognition processing module;

[0056] The power module provides power supply for the entire device;

[0057] The recognition processing module performs compliance detection on the original image data.

[0058] In step 5, the standard image data is binarized to obtain center and edge contour information to facilitate subsequent recognition processing.

[0059] The working principle of the embodiment of the present invention is as follows: the image processing part includes grayscale processing of the image, smoothing filtering and Canny edge detection of the obtained grayscale image. After obtaining the image data, the contour outside the threshold is removed to eliminate the influence of external noise and the environment on the image contour. After finding the center point of the image, the maximum range and minimum range of the image are solved according to the coordinates of the edge points of the image.

[0060] Image feature selection, first use KL transform and Laplace map to extract the features of the target object, and establish a feature library of different objects for screening. Generally speaking, the color of the target object is an important feature to distinguish different types of objects. Therefore, in terms of feature extraction of objects, compared with grayscale space, extracting the features of objects based on HSV and RGB color space can improve the accuracy of target object recognition.

[0061] Working principle: The creator collects the basic features of the target image, stores them in the data storage module, and establishes the target image feature library, in which the feature values ​​of the target image are stored. The image acquisition module acquires the image of the target area, obtains the original image data, and sends it to the recognition processing module. After receiving the original image data, the recognition processing module detects the original image data. After the original image data passes the standard detection, image enhancement is performed, including histogram enhancement, histogram averaging and contrast enhancement, to obtain the target standard image data. The three-dimensional image contour data of the target standard image data is processed by algorithm to obtain the center and edge contour information. The image acquisition module uses algorithm analysis to identify the image of the scene area and judge whether the scene image is consistent with the target image. After the image of the scene area is consistent with the target standard image data, the recognition processing module uses the scene image information to perform image filtering and image segmentation operations, wherein the image filtering algorithm is the median filtering algorithm, and the image segmentation method is the binary segmentation method. The feature value of the scene image is matched with the feature value in the image feature library, and compared with the standard image pre-stored in the image processing module through the module matching method to complete the recognition process.

[0062] Image enhancement includes histogram enhancement, histogram averaging and contrast enhancement. The histogram of an image is an important statistical feature of the image. It can be considered as an approximation of the grayscale density function of the image, so it has corresponding statistical features, among which the most important feature is the grayscale density function. The grayscale histogram of a natural image with histogram averaging usually has a higher frequency in the low-value grayscale interval. In order to make the image clear, the grayscale range of the image can be widened, and the grayscale with a smaller grayscale frequency can be increased, that is, the grayscale histogram can be made consistent within a larger dynamic range. Contrast enhancement is to modify the grayscale of each pixel of the input image according to certain rules, thereby changing the dynamic range of the image grayscale. It can expand the grayscale dynamic range, compress it, or segment the grayscale, compress it in a certain interval and expand it in another interval according to the characteristics and requirements of the image.

[0063] The image processing part includes grayscale processing of the image, smoothing filtering of the obtained grayscale image and Canny edge detection. After obtaining the image data, the contour outside the threshold is removed to eliminate the influence of external noise and the environment on the image contour. After finding the center point of the image, the maximum and minimum ranges of the image are solved according to the coordinates of the edge points of the image.

[0064] Image feature selection, first use KL transform and Laplace map to extract the features of the target object, and establish a feature library of different objects for screening. Generally speaking, the color of the target object is an important feature to distinguish different types of objects. Therefore, in terms of feature extraction of objects, compared with grayscale space, extracting the features of objects based on HSV and RGB color space can improve the accuracy of target object recognition.

[0065] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit thereof, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An image processing and recognition algorithm based on visual acquisition, characterized in that: The following steps are involved: Step S1: The creator collects basic features of the target image, stores them in a data storage module, and establishes a target image feature library, in which the feature values ​​of the target image are stored; Step S2: The image acquisition module acquires images of the target area, obtains original image data, and sends it to the recognition processing module; Step S3: After receiving the original image data, the recognition processing module detects the original image data. If it is qualified, it jumps to step 4; if it is not qualified, it jumps to step 2; Step S4: After the original image data passes the standard detection, image enhancement is performed, including histogram enhancement, histogram averaging and contrast enhancement, to obtain target standard image data; Step S5: performing algorithm processing on the three-dimensional image contour data of the target standard image data to obtain center and edge contour information; Step S6: The image acquisition module uses an algorithm to analyze and identify the image of the scene area, and determines whether the scene image is consistent with the target image. If they are consistent, it jumps to step 7; if not, it jumps to step 2; Step S7: After the image of the on-site area is consistent with the target standard image data, the recognition processing module uses the on-site image information to perform image filtering and image segmentation operations, wherein the image filtering algorithm is a median filtering algorithm, and the image segmentation method is a binary segmentation method; Step S8: Match the feature values ​​of the scene image with the feature values ​​in the image feature library, and compare them with the standard images pre-stored in the image processing module through the module matching method to complete the recognition process. If the match is successful, it is judged as a target, if the match fails, it is judged as a non-target.

2. The image processing and recognition algorithm based on visual acquisition according to claim 1, characterized in that: The algorithm includes a storage module, a power module, an image acquisition module, an identification processing module, a fault processing module and a power protection module.

3. The image processing and recognition algorithm based on visual acquisition according to claim 1, characterized in that: The step S4 comprises the following steps: Step S4.1: enhancing the image using a grayscale density function; Step S4.2: Expand the grayscale range of the image and make the grayscale with a smaller grayscale frequency larger, that is, make the grayscale histogram tend to be consistent within a larger dynamic range; Step S4.3: Modify the grayscale of each pixel of the input image according to certain rules, thereby changing the dynamic range of the image grayscale.

4. The image processing and recognition algorithm based on visual acquisition according to claim 1, characterized in that: The step S8 includes the following steps: storing the feature value of the extracted image in data variable A, and using the data of data variable A to compare with the data feature library, normalizing the data variable B, and judging whether it is a matching value of the target image based on the variance of data variable A relative to data variable B.

5. The image processing and recognition algorithm based on visual acquisition according to claim 1, characterized in that: In the step S2, the original image data is filtered, grayed, and histogram equalized in sequence to obtain the standard image data for subsequent comparison.

6. The image processing and recognition algorithm based on visual acquisition according to claim 2 is characterized in that: In step S2, the target object is photographed by a digital camera and converted into an image signal, and then the image signal is converted into digital information by an image acquisition card and sent to the recognition processing module; The power module provides power supply for the entire device; The recognition processing module performs compliance detection on the original image data.

7. The image processing and recognition algorithm based on visual acquisition according to claim 1, characterized in that: In step 5, the standard image data is binarized to obtain center and edge contour information to facilitate subsequent recognition processing.

Images