CN114170158A - Image target detection method and device, computer equipment and readable storage medium - Google Patents
Image target detection method and device, computer equipment and readable storage medium Download PDFInfo
- Publication number
- CN114170158A CN114170158A CN202111400297.7A CN202111400297A CN114170158A CN 114170158 A CN114170158 A CN 114170158A CN 202111400297 A CN202111400297 A CN 202111400297A CN 114170158 A CN114170158 A CN 114170158A
- Authority
- CN
- China
- Prior art keywords
- image
- target object
- target
- characteristic
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20021—Dividing image into blocks, subimages or windows
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
Abstract
The invention belongs to the technical field of image processing, and particularly relates to an image target detection method, an image target detection device, computer equipment and a readable storage medium. The method comprises the following steps: acquiring at least one initial image sample containing a multi-target object, and obtaining a regional image block set of all characteristics of the initial image sample through image characteristic separation processing; selecting regional image blocks of the multi-target object characteristics, and configuring a determined neural network model for the selected regional image blocks of each target object; acquiring an image text to be detected, and processing by using the respective neural network model of each target object to obtain an edge coordinate point set of the detected target object; the obtained edge coordinate point sets are mapped into the same image text to mark a plurality of features associated with the multi-target object. The method and the device realize the detection of the multiple target objects in the image without acquiring a large number of training samples in advance, accelerate the efficiency of image target detection and avoid the problems of detection omission or inaccurate detection.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to an image target detection method, an image target detection device, computer equipment and a readable storage medium.
Background
With the rapid development of computer vision related theories and application research, the superiority of computer vision technology in daily life is increasingly highlighted, and a computer is adopted to identify and extract features in related videos or images, so that the image processing technology is gradually popularized and continuously applied.
In the existing application, in order to detect and recognize a target object in an image, a detection network is generally trained by using artificial intelligence machine learning, and then the detection of the object in the image can be realized through the detection network. When the detection network training is performed, a group of sample data sets with the same category are usually adopted for training, so that the detection of a target object is realized, an image carrying the target object needs to be collected in advance for training the detection network, only when the magnitude of the target object to be detected is large, the target object is added into the detection network to generate target identification, and for the target object which is not trained in advance in the image, the problems of missing detection or inaccurate detection usually exist, so that the image target detection becomes a huge project. Time and labor are wasted, and the detection requirement of the multi-target object in the image cannot be met.
Disclosure of Invention
The invention provides an image target detection method, an image target detection device, a computer device and a readable storage medium, aiming at solving the problems that sample data with large magnitude order needs to be acquired in advance for target object detection in an image and detection is not accurate for a target object which is not trained in advance.
In order to achieve the above purpose, the embodiment of the present invention provides the following technical solutions:
in a first aspect, in an embodiment provided by the present invention, an image target detection method is provided, including:
acquiring at least one initial image sample containing a multi-target object, and obtaining a regional block set of all characteristics of the initial image sample through image characteristic separation processing;
selecting regional image blocks of the multi-target object characteristics based on the regional image block set, and configuring a determined neural network model for the selected regional image blocks of each target object;
acquiring an image text to be detected, and processing by using the respective neural network model of each target object to obtain an edge coordinate point set of the detected target object;
the obtained edge coordinate point sets are mapped into the same image text to mark a plurality of features associated with the multi-target object.
In some embodiments provided herein, the method of obtaining a region tile set of all features of the initial image sample comprises:
acquiring at least one initial image sample containing a multi-target object;
acquiring pixel characteristic points of all objects by adopting an image characteristic identification method;
carrying out binarization processing on the initial image sample, setting the pixel characteristic point part as a foreground, and setting the rest part as a background to obtain a characteristic image block containing all objects;
and (4) utilizing an image segmentation technology to segment the characteristic image blocks to obtain a regional image block set of all the characteristics.
In some embodiments provided by the present invention, the step of acquiring the pixel feature points by the image feature identification method includes:
reading image pixel points of an initial image sample, and inputting the image pixel points into an image RGB three channel to obtain RGB three channel data of the current initial image sample;
and comparing the calibrated background image with RGB three-channel data of the current initial image sample, and detecting foreground pixel characteristic points and background pixel points in the initial image sample.
In some embodiments provided by the present invention, a method for obtaining a feature image block including all objects includes:
setting separation threshold values of the foreground pixel characteristic points and the background pixel points;
carrying out binarization processing on the initial image sample to obtain the contour line characteristics of the foreground pixels;
and carrying out parameterization and fitting processing on the contour line of the foreground pixel, and segmenting along the processed contour line to obtain a characteristic image block containing all objects.
In some embodiments provided by the present invention, the method for performing parameterization and fitting processing on the contour line includes:
obtaining the contour line of the foreground pixel;
acquiring a discrete edge feature point set along the contour line according to a set step length;
fitting a contour curve of the characteristic region based on the discrete edge characteristic point set to obtain a fitted contour line;
and (5) segmenting along the fitted contour line to obtain a characteristic image block containing all the objects.
In some embodiments provided by the present invention, the set of edge feature points is an edge position attribute of the object feature region, and the edge position attribute includes edge point position coordinates of the subject object.
In some embodiments provided herein, a method of configuring a determined neural network, comprising:
acquiring a matching file in advance, wherein the configuration file comprises a neural network model and parameter information;
configuring the neural network model according to the number of target objects;
naming a name of the respective neural network model based on the name of each target object;
configuring a region image block of the target object as a parameter in a neural network model;
wherein the neural network model of each target object is the same neural network model.
In some embodiments provided by the present invention, before obtaining the set of edge coordinate points of the detected target object, further includes:
copying the obtained detection image text, and inputting a neural network model configured for each target object in parallel;
detecting each input detection image text by using the respective neural network model of each target object to acquire whether a region picture block associated with each target object exists or not;
synchronously detecting an image text to be detected containing a target object, and acquiring an edge coordinate point set of the detected target object in the image text to be detected.
In some embodiments provided herein, the method of marking a plurality of features associated with a multi-target object includes:
determining a current detection image text from the detection image texts containing the edge coordinate point set;
determining each edge coordinate point set except the current detection image text in a plurality of detection image texts containing edge coordinate point sets as an edge coordinate point set to be mapped;
and sequentially mapping the edge coordinate point sets to be mapped of a plurality of target objects to the same current detection image text based on the edge coordinate point set to be mapped, and completing the detection of the multi-target object in the detection image text.
In a second aspect, in another embodiment provided by the present invention, an image target detection system is provided, which detects multiple target objects in an image by using the image target detection method; the image target detection system comprises a characteristic image block determining module, a model configuration module, a target identification module and a multi-target fusion module.
The characteristic image block determining module is used for separating the regional image block sets of all the characteristics from the obtained initial image sample and selecting the regional image blocks of the multi-target object characteristics from the regional image block sets;
the model configuration module is used for configuring the corresponding determined neural network models one by one according to the number of the regional image blocks of the selected multi-target object characteristics;
the target identification module is used for copying the acquired image text to be detected, then respectively processing the image text by using the respective neural network of each target object, identifying the target object through the respective neural network, and acquiring an edge coordinate point set of the detected target object;
and the multi-target fusion module is used for mapping the obtained edge coordinate point set to the same image text and fusing the detected multi-target object to the same image text to be detected.
In a third aspect, in a further embodiment provided by the present invention, there is provided a computer device comprising a memory storing a computer program and a processor implementing the steps of the image object detection method when the computer program is loaded and executed.
In a fourth aspect, in a further embodiment provided by the present invention, a readable storage medium is provided, storing a computer program which, when loaded and executed by a processor, performs the steps of the image object detection method.
The technical scheme provided by the invention has the following beneficial effects:
according to the image target detection method, the image target detection device, the computer equipment and the readable storage medium, the determined neural network model is configured for each target object aiming at the specified multi-target object; the method comprises the steps of copying an image text to be detected, inputting the copied image text into a neural network model in parallel, synchronously detecting each target object, fusing the coordinates of the detected and identified target objects in the image text to the same image text, realizing the detection of the multi-target objects in the image, not needing to acquire a large number of training samples in advance and solving the problem of multi-target detection.
These and other aspects of the invention are apparent from and will be elucidated with reference to the embodiments described hereinafter. It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention. In the drawings:
fig. 1 is a flowchart of an image target detection method according to an embodiment of the present invention.
Fig. 2 is a schematic structural diagram of multi-target detection in the image target detection method according to the embodiment of the present invention.
Fig. 3 is a flowchart of acquiring all region tile sets in an image target detection method according to an embodiment of the present invention.
Fig. 4 is a flowchart of acquiring pixel feature points in an image target detection method according to an embodiment of the present invention.
Fig. 5 is a flowchart illustrating obtaining a feature image block including all objects in an image target detection method according to an embodiment of the present invention.
Fig. 6 is a flowchart of parameterization and fitting processing in an image target detection method according to an embodiment of the present invention.
Fig. 7 is a flowchart of configuring a neural network model in an image target detection method according to an embodiment of the present invention.
Fig. 8 is a flowchart before an edge coordinate point set is obtained in an image target detection method according to an embodiment of the present invention.
Fig. 9 is a flowchart illustrating a method for detecting an image target according to an embodiment of the present invention.
Fig. 10 is a flowchart of configuring a neural network model in an image target detection method according to an embodiment of the present invention.
Fig. 11 is a system block diagram of an image target detection system according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
Because the target object in the image is detected and identified in the existing application, the detection network is usually trained by utilizing artificial intelligence machine learning, and then the detection of the object in the image can be realized through the detection network. Usually, a set of sample data sets with the same category is used for training to realize the detection of a target object. The image carrying the target object with large magnitude needs to be collected in advance to train the detection network, so that the efficiency of image target detection is low, and the problems of detection omission or inaccurate detection exist.
In view of the above problems, the present invention provides an image target detection method, an image target detection apparatus, a computer device, and a readable storage medium, which do not require a large number of training samples to be obtained in advance. Aiming at a specified multi-target object, configuring a determined neural network model for each target object; and copying the image text to be detected, inputting the copied image text into a neural network model in parallel, synchronously detecting each target object, and fusing the coordinates of the detected and identified target objects in the image text to the same image text to realize the detection of the multi-target object in the image.
Specifically, the embodiments of the present application will be further explained below with reference to the drawings.
As shown in fig. 1 and 2, an embodiment of the present invention provides an image target detection method, including the following steps:
s1: at least one initial image sample containing the multi-target object is obtained, and a regional block set of all characteristics of the initial image sample is obtained through image characteristic separation processing.
In this embodiment, the obtained initial image sample may be one image sample containing a multi-target object, or may be a set of multiple image samples that are combined together to contain a desired target object. It is sufficient to ensure that all desired target objects are contained in the initial image sample.
In this embodiment, as shown in fig. 3, the method for obtaining a region tile set of all features of the initial image sample includes:
and S11, acquiring at least one initial image sample containing the multi-target object.
And S12, acquiring pixel characteristic points of all objects by adopting an image characteristic identification method.
In this embodiment, as shown in fig. 4, the step of acquiring the pixel feature point by the image feature identification method includes:
s121, reading image pixel points of an initial image sample, and inputting the image pixel points into an image RGB three channel to obtain RGB three channel data of the current initial image sample;
and S122, comparing the calibrated background image with RGB three-channel data of the current initial image sample, and detecting foreground pixel characteristic points and background pixel points in the initial image sample.
In the specific implementation of step S122 provided in the embodiment of the present invention, a difference operation is performed on the RGB three-channel data of the current initial image sample and the calibrated RGB three-channel data of the background image to determine whether there is a foreground pixel feature point containing an image feature in the initial image sample, and when there is a foreground pixel feature point in the video sequence image, there is an image feature in the initial image sample. And when the foreground pixel characteristic points do not exist in the video sequence image, the image characteristic does not exist in the video sequence image.
And S13, performing binarization processing on the initial image sample, setting the pixel characteristic point part as a foreground, and setting the rest part as a background to obtain a characteristic image block containing all objects.
In this embodiment, when setting the foreground and background regions, the feature image block including all the objects is obtained according to the pixel value of the feature point of the pixel being 255 and the pixel values of the other regions being 0.
In this embodiment, as shown in fig. 5, the method for obtaining a feature image block including all objects includes:
s131, setting a separation threshold value of the foreground pixel feature point and the background pixel point;
s132, carrying out binarization processing on the initial image sample to obtain the contour line characteristics of the foreground pixels;
and S133, carrying out parameterization and fitting processing on the contour line of the foreground pixel, and segmenting along the processed contour line to obtain a characteristic image block containing all objects.
Further, as shown in fig. 6, the method for performing parameterization and fitting processing on the contour line includes:
s1331, obtaining a contour line of the foreground pixel;
s1332, obtaining a discrete edge feature point set along the contour line according to a set step length;
s1333, fitting a contour curve of the characteristic region based on the discrete edge characteristic point set to obtain a fitted contour line;
and S1334, segmenting along the fitted contour line to obtain a characteristic image block containing all the objects.
In this embodiment, when obtaining the discrete edge feature point set, assuming that the set step size is 0.1, the discrete edge feature point set is obtained along the contour line. And obtaining a contour curve of each target object after refitting, and separating along the fitted contour curve.
And S14, dividing the characteristic image blocks by using an image division technology to obtain an area image block set of all the characteristics.
In this embodiment, the edge feature point set is an edge position attribute of the object feature region, where the edge position attribute includes an edge point position coordinate of the main object.
S2: region blocks of the multi-target object features are selected based on the region block set, and the determined neural network model is configured for the selected region blocks of each target object.
In this embodiment, referring to fig. 7, a method for configuring a determined neural network includes:
s21, acquiring a matching file in advance, wherein the configuration file comprises a neural network model and parameter information;
s22, configuring the neural network model according to the number of the target objects;
s23, naming the name of each neural network model based on the name of each target object;
and S24, configuring the region image block of the target object in the neural network model as a parameter.
In the present embodiment, the neural network model of each target object is the same neural network model.
S3: and acquiring an image text to be detected, and processing by using the respective neural network model of each target object to obtain an edge coordinate point set of the detected target object.
In this embodiment, referring to fig. 8, before obtaining the set of edge coordinate points where the target object is detected, the method further includes:
s31, copying the obtained detection image text, and inputting the neural network model configured by each target object in parallel;
s32, detecting each input detection image text by using the respective neural network model of each target object to acquire whether a region picture block associated with each target object exists or not;
s33, synchronously detecting the image text to be detected containing the target object, and acquiring the edge coordinate point set of the detected target object in the image text to be detected.
In this embodiment, a detection image text to be detected is copied or copied and then input to the neural network model corresponding to each target object, a single target object is identified, and an edge coordinate point set is obtained for the detection image text containing the corresponding target object, so that edge coordinate point sets contained in image texts output by a plurality of neural network models are mapped to the same image text. And the efficiency of multi-target detection is accelerated by adopting a parallel processing mode.
S4: the obtained edge coordinate point sets are mapped into the same image text to mark a plurality of features associated with the multi-target object.
In this embodiment, referring to fig. 9, the method for marking a plurality of features associated with a multi-target object includes:
s41, determining a current detection image text from the detection image texts containing the edge coordinate point set;
s42, determining each edge coordinate point set except the current detection image text in the detection image texts containing the edge coordinate point sets as edge coordinate point sets to be mapped;
and S43, sequentially mapping the edge coordinate point sets to be mapped of a plurality of target objects to the same current detection image text based on the edge coordinate point sets to be mapped, and completing the detection of the plurality of target objects in the detection image text.
In an embodiment of the present invention, as shown in fig. 10, the method for configuring the determined neural network for the selected region tile of each target object further includes:
s201, connecting the selected region image blocks of the multi-target object characteristics to n detection layers in sequence;
s202, each detection layer comprises a neural network model which inputs a target object to be used as a characteristic for recognition;
and S203, embedding each target object as a feature vector into a corresponding neural network model so as to detect the target object in the input image.
The method of the invention is that aiming at the appointed multi-target object, a determined neural network model is configured for each target object; and copying the image text to be detected, inputting the copied image text into a neural network model in parallel, synchronously detecting each target object, and fusing the coordinates of the detected and identified target object in the image text to the same image text, so that the detection of the multi-target object in the image is realized, a large number of training samples are not required to be obtained in advance, and the problem of multi-target detection can be solved.
In an embodiment of the present invention, referring to fig. 11, the present invention further discloses an image target detection system, which detects multiple target objects in an image by using the image target detection method; the image target detection system includes a feature patch determination module 100, a model configuration module 200, a target identification module 300, and a multi-target fusion module 400.
The characteristic block determining module 100 is configured to separate a region block set of all characteristics from the acquired initial image sample, and select a region block set of the multi-target object characteristics from the region block set. When the initial image sample is obtained, the initial image sample may be one image sample containing the multi-target object, or a set of multiple image samples that are combined together to contain the desired target object. The initial image sample contains all of the desired target objects. When separating out the region tile sets, the manner of obtaining the region tile sets of all the features of the initial image sample is as follows: acquiring at least one initial image sample containing a multi-target object; acquiring pixel characteristic points of all objects by adopting an image characteristic identification method; carrying out binarization processing on the initial image sample, setting the pixel characteristic point part as a foreground, and setting the rest part as a background to obtain a characteristic image block containing all objects; and (4) utilizing an image segmentation technology to segment the characteristic image blocks to obtain a regional image block set of all the characteristics.
When the image characteristic identification method obtains the pixel characteristic points, reading image pixel points of an initial image sample, and inputting the image pixel points into an image RGB three channel to obtain RGB three channel data of the current initial image sample. Comparing RGB three-channel data of a calibrated background image and a current initial image sample, and detecting foreground pixel characteristic points and background pixel points in the initial image sample.
When the initial image sample is subjected to binarization processing and foreground and background areas are set, a characteristic image block containing all objects is obtained according to the fact that the pixel value of a pixel characteristic point part is 255 and the pixel values of other areas are 0.
In this embodiment, the manner of obtaining the feature image blocks including all the objects is as follows: setting separation threshold values of the foreground pixel characteristic points and the background pixel points; carrying out binarization processing on the initial image sample to obtain the contour line characteristics of the foreground pixels; and carrying out parameterization and fitting processing on the contour line of the foreground pixel, and segmenting along the processed contour line to obtain a characteristic image block containing all objects.
Obtaining the contour line of the foreground pixel when carrying out parameterization and fitting processing on the contour line; acquiring a discrete edge feature point set along the contour line according to a set step length; fitting a contour curve of the characteristic region based on the discrete edge characteristic point set to obtain a fitted contour line; and (5) segmenting along the fitted contour line to obtain a characteristic image block containing all the objects. And when the discrete edge feature point set is obtained, assuming that the set step is 0.1, obtaining the discrete edge feature point set along the contour line. And obtaining a contour curve of each target object after refitting, and separating along the fitted contour curve.
The model configuration module 200 is configured to configure the corresponding determined neural network models one by one according to the number of the region blocks of the selected multi-target object features. The determined neural network model includes, but is not limited to, a Convolutional Neural Network (CNN), a deconvolution neural network (DN), and a Recurrent Neural Network (RNN). When the neural network is configured, a matching file is obtained in advance, and the configuration file comprises a neural network model and parameter information; configuring the neural network model according to the number of target objects; naming a name of the respective neural network model based on the name of each target object; and configuring the region image blocks of the target object as parameters in the neural network model.
The target identification module 300 is configured to copy the acquired image text to be detected, perform respective processing by using the respective neural network of each target object, identify the target object through the respective neural network, and obtain an edge coordinate point set of the detected target object. Before the edge coordinate point set of the target object is acquired, the acquired detection image text is copied, and a neural network model configured for each target object is input in parallel; detecting each input detection image text by using the respective neural network model of each target object to acquire whether a region picture block associated with each target object exists or not; synchronously detecting an image text to be detected containing a target object, and acquiring an edge coordinate point set of the detected target object in the image text to be detected.
The multi-target fusion module 400 is configured to map the obtained edge coordinate point sets to the same image text, and fuse the detected multi-target objects to the same image text to be detected. When a plurality of features associated with the multi-target object are marked, determining a current detection image text from the detection image texts containing the edge coordinate point set; determining each edge coordinate point set except the current detection image text in a plurality of detection image texts containing edge coordinate point sets as an edge coordinate point set to be mapped; and sequentially mapping the edge coordinate point sets to be mapped of a plurality of target objects to the same current detection image text based on the edge coordinate point set to be mapped, and completing the detection of the multi-target object in the detection image text.
In this embodiment, the image target detection system adopts the steps of the image target detection method as described above when executing, and therefore, the operation process of the image target detection system in this embodiment will not be described in detail.
In one embodiment, there is further provided, in an embodiment of the present invention, a computer device, including at least one processor, and a memory communicatively connected to the at least one processor, where the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to cause the at least one processor to execute the image object detection method, and the processor executes the instructions to implement the steps in the method embodiments:
acquiring at least one initial image sample containing a multi-target object, and obtaining a regional block set of all characteristics of the initial image sample through image characteristic separation processing;
selecting regional image blocks of the multi-target object characteristics based on the regional image block set, and configuring a determined neural network model for the selected regional image blocks of each target object;
acquiring an image text to be detected, and processing by using the respective neural network model of each target object to obtain an edge coordinate point set of the detected target object;
the obtained edge coordinate point sets are mapped into the same image text to mark a plurality of features associated with the multi-target object.
In an embodiment of the present invention, there is further provided a computer device, including a memory and a processor, where the memory stores a computer program, and the processor implements the steps in the foregoing method embodiments when executing the computer program:
acquiring at least one initial image sample containing a multi-target object, and obtaining a regional block set of all characteristics of the initial image sample through image characteristic separation processing;
selecting regional image blocks of the multi-target object characteristics based on the regional image block set, and configuring a determined neural network model for the selected regional image blocks of each target object;
acquiring an image text to be detected, and processing by using the respective neural network model of each target object to obtain an edge coordinate point set of the detected target object;
the obtained edge coordinate point sets are mapped into the same image text to mark a plurality of features associated with the multi-target object.
In an embodiment of the invention, a readable storage medium is also provided, on which a computer program is stored which, when being executed by a processor, carries out the steps of the above-mentioned method embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database or other medium used in the embodiments provided herein can include at least one of non-volatile and volatile memory.
In summary, according to the image target detection method, the image target detection device, the computer device and the readable storage medium provided by the invention, for a specified multi-target object, a determined neural network model is configured for each target object; the method comprises the steps of copying an image text to be detected, inputting the copied image text into a neural network model in parallel, synchronously detecting each target object, fusing the coordinates of the detected and identified target objects in the image text to the same image text, realizing the detection of the multi-target objects in the image, not needing to acquire a large number of training samples in advance and solving the problem of multi-target detection.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention are intended to be included within the scope of the present invention.
Claims (9)
1. An image object detection method, comprising:
acquiring at least one initial image sample containing a multi-target object, and obtaining a regional block set of all characteristics of the initial image sample through image characteristic separation processing;
selecting regional image blocks of the multi-target object characteristics based on the regional image block set, and configuring a determined neural network model for the selected regional image blocks of each target object;
acquiring an image text to be detected, and processing by using the respective neural network model of each target object to obtain an edge coordinate point set of the detected target object;
the obtained edge coordinate point sets are mapped into the same image text to mark a plurality of features associated with the multi-target object.
2. The image object detection method of claim 1, characterized by: the method of obtaining a region tile set of all features of the initial image sample, comprising:
acquiring at least one initial image sample containing a multi-target object;
acquiring pixel characteristic points of all objects by adopting an image characteristic identification method;
carrying out binarization processing on the initial image sample, setting the pixel characteristic point part as a foreground, and setting the rest part as a background to obtain a characteristic image block containing all objects;
and (4) utilizing an image segmentation technology to segment the characteristic image blocks to obtain a regional image block set of all the characteristics.
3. The image object detection method of claim 2, characterized by: the image feature identification method for obtaining the pixel feature points comprises the following steps:
reading image pixel points of an initial image sample, and inputting the image pixel points into an image RGB three channel to obtain RGB three channel data of the current initial image sample;
and comparing the calibrated background image with RGB three-channel data of the current initial image sample, and detecting foreground pixel characteristic points and background pixel points in the initial image sample.
4. The image object detection method of claim 3, characterized by: the method for obtaining the characteristic image block containing all the objects comprises the following steps:
setting separation threshold values of the foreground pixel characteristic points and the background pixel points;
carrying out binarization processing on the initial image sample to obtain the contour line characteristics of the foreground pixels;
and carrying out parameterization and fitting processing on the contour line of the foreground pixel, and segmenting along the processed contour line to obtain a characteristic image block containing all objects.
5. The image object detection method of claim 4, characterized in that: the method for carrying out parameterization and fitting processing on the contour line comprises the following steps:
obtaining the contour line of the foreground pixel;
acquiring a discrete edge feature point set along the contour line according to a set step length;
fitting a contour curve of the characteristic region based on the discrete edge characteristic point set to obtain a fitted contour line;
and (5) segmenting along the fitted contour line to obtain a characteristic image block containing all the objects.
6. The image object detection method of claim 5, characterized by: the scattered edge feature point set is an edge position attribute of the object feature area, and the edge position attribute comprises edge point position coordinates of the main object.
7. An image object detection system characterized by: the image target detection system adopts the image target detection method of any one of claims 1 to 6 to detect multiple target objects in an image; the image target detection system includes:
the characteristic image block determining module is used for separating a regional image block set of all characteristics from the obtained initial image sample and selecting regional image blocks of the characteristics of the multi-target object from the regional image block set;
the model configuration module is used for configuring the corresponding determined neural network models one by one according to the number of the regional image blocks of the selected multi-target object characteristics;
the target identification module is used for copying the acquired image text to be detected, then respectively processing the image text by using the respective neural network of each target object, identifying the target object through the respective neural network, and acquiring an edge coordinate point set of the detected target object;
and the multi-target fusion module is used for mapping the obtained edge coordinate point set to the same image text and fusing the detected multi-target object to the same image text to be detected.
8. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the steps of the method of any one of claims 1 to 6 are implemented when the computer program is loaded and executed by the processor.
9. A readable storage medium storing a computer program, wherein the computer program is loaded by a processor and executed to implement the steps of the method according to any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111400297.7A CN114170158A (en) | 2021-11-24 | 2021-11-24 | Image target detection method and device, computer equipment and readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111400297.7A CN114170158A (en) | 2021-11-24 | 2021-11-24 | Image target detection method and device, computer equipment and readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114170158A true CN114170158A (en) | 2022-03-11 |
Family
ID=80480429
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111400297.7A Pending CN114170158A (en) | 2021-11-24 | 2021-11-24 | Image target detection method and device, computer equipment and readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114170158A (en) |
-
2021
- 2021-11-24 CN CN202111400297.7A patent/CN114170158A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3620964B1 (en) | Lane line processing method and device | |
| KR101640998B1 (en) | Image processing apparatus and image processing method | |
| US9704017B2 (en) | Image processing device, program, image processing method, computer-readable medium, and image processing system | |
| CN110210483B (en) | Medical image lesion region segmentation method, model training method and device | |
| CN107239794B (en) | Point cloud data segmentation method and terminal | |
| CN110675407B (en) | Image instance segmentation method and device, electronic equipment and storage medium | |
| CN110264444B (en) | Damage detection method and device based on weak segmentation | |
| EP3843036B1 (en) | Sample labeling method and device, and damage category identification method and device | |
| CN111028261B (en) | High-precision semi-automatic image data annotation method, electronic device and storage medium | |
| CN112884782B (en) | Biological object segmentation method, apparatus, computer device, and storage medium | |
| CN111126393A (en) | Vehicle appearance refitting judgment method and device, computer equipment and storage medium | |
| CN110992384B (en) | Semi-automatic image data labeling method, electronic device and storage medium | |
| CN112381837A (en) | Image processing method and electronic equipment | |
| CN111738036A (en) | Image processing method, device, equipment and storage medium | |
| CN111415364A (en) | Method, system and storage medium for converting image segmentation samples in computer vision | |
| CN111008576A (en) | Pedestrian detection and model training and updating method, device and readable storage medium thereof | |
| CN109241893B (en) | Road selection method and device based on artificial intelligence technology and readable storage medium | |
| CN110874170A (en) | Image area correction method, image segmentation method and device | |
| CN110880003B (en) | Image matching method and device, storage medium and automobile | |
| CN114627397A (en) | Behavior recognition model construction method and behavior recognition method | |
| CN110555344B (en) | Lane line recognition method, lane line recognition device, electronic device, and storage medium | |
| CN114170158A (en) | Image target detection method and device, computer equipment and readable storage medium | |
| CN110751163A (en) | Target positioning method and device, computer readable storage medium and electronic equipment | |
| CN114359332A (en) | Target tracking method, device, equipment and medium based on depth image | |
| CN114494693A (en) | Method and device for performing semantic segmentation on image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |