CN110232660B - Novel infrared image recognition preprocessing gray stretching method - Google Patents
Novel infrared image recognition preprocessing gray stretching method Download PDFInfo
- Publication number
- CN110232660B CN110232660B CN201910347677.5A CN201910347677A CN110232660B CN 110232660 B CN110232660 B CN 110232660B CN 201910347677 A CN201910347677 A CN 201910347677A CN 110232660 B CN110232660 B CN 110232660B
- Authority
- CN
- China
- Prior art keywords
- image
- temperature
- stretching
- gray
- recognition
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 13
- 238000007781 pre-processing Methods 0.000 title claims abstract description 13
- 238000012545 processing Methods 0.000 claims abstract description 50
- 230000007613 environmental effect Effects 0.000 abstract description 6
- 238000012549 training Methods 0.000 description 9
- 238000013135 deep learning Methods 0.000 description 5
- 238000003672 processing method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000000605 extraction Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000005855 radiation Effects 0.000 description 2
- WJCNZQLZVWNLKY-UHFFFAOYSA-N thiabendazole Chemical compound S1C=NC(C=2NC3=CC=CC=C3N=2)=C1 WJCNZQLZVWNLKY-UHFFFAOYSA-N 0.000 description 2
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G06T5/90—
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
-
- 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
- G06T2207/10—Image acquisition modality
- G06T2207/10048—Infrared image
Abstract
The invention discloses a novel infrared image recognition preprocessing gray level stretching method, and belongs to the technical field of image processing. The invention can be used for detecting the constant temperature object in the infrared image, such as pedestrian recognition, animal recognition, vehicle recognition and the like. According to the invention, the temperature and the gray level are in one-to-one correspondence, the gray level detail of the object is improved by stretching the gray level in the temperature range of the identified object, the influence of the environmental temperature on the gray level characteristic of the identified object is reduced, the universality of gray level stretching processing on the image is improved, the processing link of image gray level stretching preprocessing in an object identification system in an infrared image is improved, and the identification accuracy of the target object is improved.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a gray level stretching processing technology for an infrared image.
Background
Pedestrian recognition in infrared images is always a popular field of image recognition, and has a wide application field, and how to improve the recognition precision of infrared images, and the recognition accuracy is always an important research problem in the industry.
In the pedestrian recognition processing of the infrared image, firstly, acquiring the infrared image and performing image preprocessing (including gray stretching processing), then setting a corresponding training data set (training samples are subjected to image preprocessing including stretching processing) based on a preset recognition model (usually, a deep learning network), and training the recognition model based on the training data set; and finally, inputting the image to be detected after the image preprocessing including the gray stretching processing into the trained recognition model by utilizing the trained recognition model, and performing pedestrian recognition detection processing.
Therein, there areThe specific processing method of the image gray scale stretching (also called image gray scale display) scheme is as follows:
wherein G is the gray value of the corresponding coordinate point in the image, and T isfmaxFor the highest temperature, T, in the image identified by the sensor in the imagefminThe lowest temperature within the image is identified for the sensor in the image, and t is the sensor acceptance temperature value at the coordinate point.
The main idea of the existing image gray scale stretching processing is to divide the highest temperature and the lowest temperature into 256 gray scale values, and then correspond the gray scale values with the temperature values identified by the sensor and the gray scale values of the pictures in an isocratic manner, so as to achieve the purpose of imaging. However, this processing method has a large defect in object recognition, and when the temperature difference is too large, it is difficult to recognize a low-temperature object and separate it, and when the ambient temperature and the recognized temperature are too close, it may also affect the detection of the recognized object. I.e. it lacks self-adjustment according to different conditions of the image, so that the best presentation of the detected object is achieved, and the flexibility is poor.
The image gray scale stretching processing scheme in the current processing scheme for pedestrian identification lacks universality for various conditions all the year round, so that the technical problem that the identified object or error identification cannot be easily identified when the environmental difference is large or the environmental temperature is close is solved.
Disclosure of Invention
The invention aims to: aiming at the existing problems, a novel infrared image recognition preprocessing gray level stretching method is provided.
The technical scheme adopted by the invention is as follows:
temperature acceptance range [ T ] of sensor based on image to be processed (image to be subjected to gradation stretching processing)min,Tmax]Setting gray values R displayed on the image by all pixel points of the image to be processed as follows:
wherein t is the temperature value of each pixel point received by the sensor;
stretching the grey value R to obtain a grey value g, wherein
The values of the parameters m and E are respectively as follows:
setting a parameter m based on a median value of the thermostatic object to be identified, which indicates a temperature range;
according to a preset gray scale stretching range [ K ]1,K2]And setting the value of the parameter E by the parameter m:
calculating parameters
According to the formula E ═ min { E ═1，E2Calculating the value of the parameter E;
and obtaining the final gray value G of each pixel point of the image to be processed based on the G being 255 multiplied by G, and obtaining the gray stretching processing image of the image to be processed.
In summary, due to the adoption of the technical scheme, the invention has the beneficial effects that:
on the basis of the existing gray scale stretching processing scheme, the gray scale stretching processing scheme is combined with the pixel temperature of the image to stretch the temperature range of the object to be recognized, so that the detail and the brightness degree of the recognized object are improved, and the gray scale stretching processing scheme has good self-adjusting capability when the environment difference is large. Therefore, the object identification accuracy rate when the ambient temperature difference is too large or the identified object and the ambient temperature are close in the target identification processing scheme in the infrared image is effectively improved.
Drawings
Fig. 1 is a schematic diagram of an image effect after processing by a conventional gray stretching processing method.
Fig. 2 is a schematic diagram of an image effect after processing by the gray scale stretching processing method of the present invention.
Fig. 3 is a schematic diagram showing the recognition result of the gradation stretch-processed image shown in fig. 1.
Fig. 4 is a schematic diagram of the recognition result of the gray scale stretch processed image shown in fig. 2, wherein the recognition processing is performed in the same manner as in fig. 3.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail with reference to the following embodiments and accompanying drawings.
The method can be used for detecting the constant-temperature object in the infrared image, such as pedestrian recognition, animal recognition, vehicle recognition and the like, and based on the constant-temperature object recognition processing after the gray scale stretching preprocessing of the invention (the recognition processing scheme can adopt the target recognition processing scheme in the existing infrared image, but the gray scale stretching processing in the image preprocessing adopts the gray scale stretching processing mode of the invention), the corresponding target object can be accurately recognized under the conditions of larger environmental difference or similar environmental temperature.
In order to realize the gray scale stretching processing of the invention, the temperature and the gray scale are firstly in one-to-one correspondence, the gray scale detail of the object is improved by stretching the gray scale in the temperature range of the recognized object, the influence of the environmental temperature on the gray scale characteristic of the recognized object is reduced, the universality of the gray scale stretching processing of the image is improved, the processing link of the image gray scale stretching preprocessing in the object recognition system in the infrared image is improved, and the recognition accuracy of the target object is improved.
In an identification system combining digital image processing and artificial intelligence, in order to detect a constant-temperature object (such as a pedestrian) in an infrared gray-scale image, it is considered that the constant-temperature object is a relatively stable radiation source, the surface temperature of the constant-temperature object does not greatly float throughout the year, and the constant-temperature object is not greatly influenced by the ambient temperature. Therefore, the invention establishes the one-to-one correspondence relationship between the receiving temperature range of the sensor and the gray value range represented on the image, so that the imaging effect of the constant temperature object on the image imaged by the sensor is more stable.
Sensor-based temperature acceptance range [ T ]min,Tmax]Setting the gray value R displayed by each pixel point on the image as:
and t is the temperature value of the pixel point received by the sensor in the region. So that the temperature received by each sensor corresponds to a particular gray value.
However, the temperature range measured by the sensor is far from being in one-to-one correspondence with the gray value of the image, and the defect that all gray values displayed by the image cannot be fully utilized is shown, because the extreme temperatures of-30 ℃ and 350 ℃ in the natural environment are not common, the gray value corresponding to the surface radiation temperature range of the constant-temperature object is stretched in a specific shape, so that the gray value is used in the image display range as much as possible, more specific details of the constant-temperature object are shown, and the accuracy of subsequent identification is improved.
The gray value after stretching is expressed by g, which is specifically:
that is, the range of the gradation value R is stretched, and the stretched image is an s-shaped curve. Wherein m and E determine the shape of the stretching curve, and the values are derived according to the temperature range of the identified object.
The invention is based on the temperature range [ T ] indicated by the constant-temperature object to be identifiedo_minTo_max]Setting the value of the parameter m as follows: intermediate value, i.e. T, of the temperature range of the object to be identifiedo_minAnd To_maxMedian value of (d); and based on a preset gray scale stretching range [ K ]1,K2]And the parameter m sets the value of the parameter E as follows:
E＝min{E1，E2}；
finally, the gray value G after the gray stretching process of the present invention is obtained based on that G being 255 × G, i.e., the gray value G in the range of [0,1] is expanded to [0,255] for 256 gray values.
In other words, in order to realize the gray scale stretching processing in the infrared image recognition preprocessing, a corresponding relation is established for the temperature received by the sensor and the presented gray scale, then the gray scale range of the temperature range of the detected object is found through the corresponding relation, and the gray scale range is stretched to the whole display range of the image so as to best display the recognized object.
Examples
The constant temperature object to be detected is a pedestrian, the temperature acceptance range of the adopted sensor is-30 ℃ to 350 ℃, so T is setmax＝350,Tmin-30; for winter, the surface temperature of the pedestrian is usually between 10 ℃ and 40 ℃, so m is taken as the median 26 ℃ of 10 ℃ and 40 ℃, and the gray scale stretch range is set as K1＝0.01，K20.99; based on the setting of the parameter E, E — 7 can be obtained.
In order to realize the pedestrian identification processing in the infrared image, the embodiment mainly comprises three processing links; establishing a data set, building a deep learning environment (a recognition model based on a deep learning network) and training and recognizing a target.
In the data set establishment, firstly, the acquisition of a sample picture is carried out, a Testo875-i thermal infrared imager is used in the embodiment, the imaging temperature range of the thermal imager is-30 ℃ to 350 ℃, the picture resolution is 160 x 120, the acquisition place is near a campus building, and 1260 pictures are collected in total.
And then, carrying out secondary processing on the picture, wherein the picture acquired by the thermal imager is in a picture format specific to Testo corporation, the acquired picture is exported by using picture export software IRsoft of Testo, and the export result is an Excel table, so that the acquired picture can be processed secondarily in the later period.
Reading the exported form file by using Matlab software, and processing the form data by respectively adopting the existing gray scale stretching processing and the improved gray scale stretching processing of the invention, wherein the processing results are respectively shown in FIGS. 1 and 2; and after the image is converted into a JPEG file format, the image is stored into two groups, one group is the existing gray level stretching processing result, and the other group is the gray level stretching processing result of the invention.
The data set needs to be divided into a training set and a testing set when being established, the training set is about 1000 pictures, visual calibration is carried out on the two groups of pictures by using a label tool labellimg, calibration objects are all corresponding training sets in the two groups, and the remaining 260 pictures are used as testing sets for testing and inspection. The data set was in the Pascal VOC format.
Then, a deep learning environment is built, the deep learning environment is many, such as TensorFlow, Keras, Caffe and the like, and the TensorFlow framework is adopted in the embodiment because the framework is developed rapidly and has high maturity and concentration.
Then, a specific recognition algorithm is selected, the selected algorithms include Fast-RCNN, Yolo, Fast R-CNN and the like, in the embodiment, the adopted recognition algorithm is Fast R-CNN, because the recognition accuracy of the algorithm is high, the recognition speed is high, and the practical significance is high, the running computer system is ubuntu 18.04, and the display card model is Yingwei 1050 Ti.
And finally, target training and recognition, wherein the Faster R-CNN Network comprises 3 parts including a feature extraction Network, an RPN (resilient protocol Network) Network and a classification regression Network.
The selection of the feature extraction network can be flexible, and the VGG16 network is adopted in the embodiment.
The RPN network has a main function of processing an output of the feature extraction network to obtain an output category, and then generating a RoI (region of interest) and corresponding tag information required by the next network.
And (4) classifying the regression network, classifying the detected pictures after determining the loss function, obtaining the confidence of classification, and outputting the identified pictures.
Compared with the prior art, the gray scale stretching treatment provided by the invention has certain advantages compared with the prior gray scale stretching treatment through simulation experiments, and as can be found by comparison between fig. 1 and fig. 2, the gray scale stretching treatment provided by the invention has the advantages of clearer and clearer display of the identified image, easiness in identification and no influence of the highest-temperature object (the brightest place is a lamp) in the environment. As seen from comparison between the result graphs of fig. 3 and fig. 4, the recognition processing result based on the image after the gray-scale stretching processing of the present invention enables a pedestrian not recognized originally to be recognized together with another pedestrian, and a better recognition result is obtained.
In order to compare the identification performance of the target to be identified more intuitively, mAP (mean Average Precision) is adopted as an index for measuring the accuracy of target identification, and the accuracy and the Precision of the target identification are shown.
In the present embodiment, the target recognition performance of the recognition processing based on the present invention and the recognition processing based on the conventional gradation stretching processing is shown in table 1:
TABLE 1
| Method | Existing methods | The invention |
| mAP | 87.71％ | 88.54％ |
As shown in table 1, the recognition processing based on the present invention is improved by nearly one percent compared with the recognition processing method based on the conventional gray stretching processing, and the performance improvement of the target recognition is obvious.
While the invention has been described with reference to specific embodiments, any feature disclosed in this specification may be replaced by alternative features serving the same, equivalent or similar purpose, unless expressly stated otherwise; all of the disclosed features, or all of the method or process steps, may be combined in any combination, except mutually exclusive features and/or steps.
Claims (2)
1. A novel infrared image recognition preprocessing gray level stretching method is characterized by comprising the following steps:
temperature acceptance range [ T ] of sensor based on image to be processedmin，Tmax]Setting gray values R displayed on the image by all pixel points of the image to be processed as follows:
wherein t is the temperature value of each pixel point received by the sensor;
stretching the grey value R to obtain a grey value g, wherein
The values of the parameters m and E are respectively as follows:
setting a parameter m based on a median value of the thermostatic object to be identified, which indicates a temperature range;
according to a preset gray scale stretching range [ K ]1，K2]And setting the value of the parameter E by the parameter m:
calculating parameters
According to the formula E ═ min { E ═1，E2Calculating the value of the parameter E;
and obtaining the final gray value G of each pixel point of the image to be processed based on the G being 255 multiplied by G, and obtaining the gray stretching processing image of the image to be processed.
2. The method of claim 1, wherein the gray scale stretch range [ K [ ]1，K2]The method comprises the following steps: k1＝0.01，K2＝0.99。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910347677.5A CN110232660B (en) | 2019-04-28 | 2019-04-28 | Novel infrared image recognition preprocessing gray stretching method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910347677.5A CN110232660B (en) | 2019-04-28 | 2019-04-28 | Novel infrared image recognition preprocessing gray stretching method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110232660A CN110232660A (en) | 2019-09-13 |
| CN110232660B true CN110232660B (en) | 2022-05-03 |
Family
ID=67860302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910347677.5A Active CN110232660B (en) | 2019-04-28 | 2019-04-28 | Novel infrared image recognition preprocessing gray stretching method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110232660B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110824311A (en) * | 2019-11-22 | 2020-02-21 | 南京信息工程大学 | Zinc oxide valve plate breakdown point detection device and method based on image recognition |
| CN111724301B (en) * | 2020-06-19 | 2023-05-12 | 电子科技大学 | Self-adaptive stretching method and system based on histogram statistics |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102222323A (en) * | 2011-06-13 | 2011-10-19 | 北京理工大学 | Histogram statistic extension and gradient filtering-based method for enhancing infrared image details |
| CN102798475A (en) * | 2012-08-02 | 2012-11-28 | 电子科技大学 | Method for determining temperature of object according to standard infrared video image |
| CN105354814A (en) * | 2015-08-17 | 2016-02-24 | 电子科技大学 | Infrared image segmentation method |
| CN106570831A (en) * | 2016-10-09 | 2017-04-19 | 中国航空工业集团公司洛阳电光设备研究所 | Gray image contrast equalization enhancement method |
| CN107403410A (en) * | 2017-07-25 | 2017-11-28 | 北京华新创科信息技术有限公司 | A kind of joining method of thermal infrared images |
| CN107507200A (en) * | 2017-08-31 | 2017-12-22 | 电子科技大学 | A kind of a wide range of water area extraction method of SAR image high-precisions based on communication with detection and noise suppressed |
| CN109377469A (en) * | 2018-11-07 | 2019-02-22 | 永州市诺方舟电子科技有限公司 | A kind of processing method, system and the storage medium of thermal imaging fusion visible images |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2006098162A1 (en) * | 2005-02-25 | 2006-09-21 | Fujifilm Corporation | Optical compensation sheet, polarizing plate and liquid crystal display |
| JP2009545355A (en) * | 2006-07-31 | 2009-12-24 | コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ | Method, apparatus, and computer-readable medium for visualizing scale-based image data sets |
-
2019
- 2019-04-28 CN CN201910347677.5A patent/CN110232660B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102222323A (en) * | 2011-06-13 | 2011-10-19 | 北京理工大学 | Histogram statistic extension and gradient filtering-based method for enhancing infrared image details |
| CN102798475A (en) * | 2012-08-02 | 2012-11-28 | 电子科技大学 | Method for determining temperature of object according to standard infrared video image |
| CN105354814A (en) * | 2015-08-17 | 2016-02-24 | 电子科技大学 | Infrared image segmentation method |
| CN106570831A (en) * | 2016-10-09 | 2017-04-19 | 中国航空工业集团公司洛阳电光设备研究所 | Gray image contrast equalization enhancement method |
| CN107403410A (en) * | 2017-07-25 | 2017-11-28 | 北京华新创科信息技术有限公司 | A kind of joining method of thermal infrared images |
| CN107507200A (en) * | 2017-08-31 | 2017-12-22 | 电子科技大学 | A kind of a wide range of water area extraction method of SAR image high-precisions based on communication with detection and noise suppressed |
| CN109377469A (en) * | 2018-11-07 | 2019-02-22 | 永州市诺方舟电子科技有限公司 | A kind of processing method, system and the storage medium of thermal imaging fusion visible images |
Non-Patent Citations (1)
| Title |
|---|
| 一种基于引导滤波图像分层的红外图像细节增强算法;葛朋 等;《红外技术》;20181231;第1161-1169页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110232660A (en) | 2019-09-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112380952B (en) | Power equipment infrared image real-time detection and identification method based on artificial intelligence | |
| CN110648310B (en) | Weak supervision casting defect identification method based on attention mechanism | |
| CN111709909A (en) | General printing defect detection method based on deep learning and model thereof | |
| CN111242123A (en) | Power equipment fault diagnosis method based on infrared image | |
| CN110232660B (en) | Novel infrared image recognition preprocessing gray stretching method | |
| CN111860106B (en) | Unsupervised bridge crack identification method | |
| CN116416250B (en) | Finished product detecting system of fast food canned product production line | |
| CN111882664A (en) | Multi-window accumulated difference crack extraction method | |
| CN111461036A (en) | Real-time pedestrian detection method using background modeling enhanced data | |
| CN110659637A (en) | Electric energy meter number and label automatic identification method combining deep neural network and SIFT features | |
| CN108921872B (en) | Robust visual target tracking method suitable for long-range tracking | |
| CN104268536B (en) | A kind of image method for detecting human face | |
| CN108985216B (en) | Pedestrian head detection method based on multivariate logistic regression feature fusion | |
| CN107563327B (en) | Pedestrian re-identification method and system based on self-walking feedback | |
| CN114662594B (en) | Target feature recognition analysis system | |
| CN112001894B (en) | Thyroid gland boundary smoothness detection device | |
| CN113688819B (en) | Target object expected point tracking and matching method based on marked points | |
| CN113591548B (en) | Target ring identification method and system | |
| CN111046861B (en) | Method for identifying infrared image, method for constructing identification model and application | |
| CN112001896B (en) | Thyroid gland border irregularity detection device | |
| CN114792300A (en) | Multi-scale attention X-ray broken needle detection method | |
| CN114120013A (en) | Infrared and RGB cross-modal feature point matching method | |
| CN112115824A (en) | Fruit and vegetable detection method and device, electronic equipment and computer readable medium | |
| CN103971118A (en) | Detection method of wine bottles in static pictures | |
| CN117197139B (en) | Tongue diagnosis image multi-label classification method based on AI |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |