Image processing method for indoor environment and collision-free system

Abstract

The application relates to the technical field of image data processing, in particular to an image processing method for an indoor environment and a collision-free system, which comprises the following steps: acquiring two initial indoor environment images corresponding to a target vehicle; performing ambient light extraction processing on each initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image; performing adaptive reflection interference elimination processing on each pixel point in the initial indoor environment image based on the reference image corresponding to each initial indoor environment image, and a preset sliding window and a preset convolution window corresponding to each pixel point in the initial indoor environment image, to obtain a target gray value corresponding to the pixel point; and updating the gray value corresponding to each pixel point in each initial indoor environment image into the target gray value corresponding to the pixel point. The application improves the image quality and the traffic safety by performing image processing on the two initial indoor environment images.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image data processing, in particular to an image processing method for indoor environment and a collision-free system. BACKGROUND

[0002] The indoor environment of an indoor parking place (for example, an underground parking lot) is often dim, and the GPS (Global Positioning System) often cannot be positioned correctly in the underground parking lot, which leads to frequent collision accidents in the underground parking lot. In order to reduce the occurrence of indoor collision accidents, a collision-free system is often used to provide a collision-free prompt for a vehicle. A commonly used collision-free system is a camera collision-free system, which mainly includes: analyzing a collision-free object in a collected image by using the collision-free system, and providing a collision-free prompt for the vehicle according to an analysis result, wherein the collision-free object can be an object that needs to be avoided by the vehicle. For example, the collision-free object can be, but is not limited to, a vehicle, a person, and a column that needs to be avoided by the vehicle. The collision-free prompt can be a prompt for the driver to avoid the collision-free object.

[0003] However, when the above method is used, the following technical problems often exist:

[0004] Due to the influence of indoor light and vehicle lights, the quality of the collected image is often poor, and thus directly analyzing the collision-free object in the collected image by using the collision-free system often leads to low accuracy of the analysis result, thereby leading to low traffic safety. SUMMARY

[0005] The summary part of the present application is used to introduce the concepts in a brief form, which will be described in detail in the specific embodiment part. The summary part of the present application is not intended to identify the key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0006] In order to solve the technical problem of low traffic safety, the present application provides an image processing method for indoor environment and a collision-free system.

[0007] In a first aspect, the present application provides an image processing method for indoor environment, which comprises:

[0008] obtaining two initial indoor environment images corresponding to a target vehicle;

[0009] performing ambient light extraction processing on each initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image;

[0010] performing adaptive reflection interference elimination processing on each pixel point in the initial indoor environment image based on a reference image corresponding to each frame of the initial indoor environment image and a preset sliding window and a preset convolution window corresponding to each pixel point in the initial indoor environment image, to obtain a target gray value corresponding to the pixel point;

[0011] updating the gray value corresponding to each pixel point in each frame of the initial indoor environment image to the target gray value corresponding to the pixel point, to obtain a target environment image corresponding to the initial indoor environment image.

[0012] Further, the environment light extraction processing on each frame of the initial indoor environment image to obtain the reference image corresponding to the initial indoor environment image comprises:

[0013] performing high-frequency removal processing on the initial indoor environment image to obtain a low-frequency component image;

[0014] performing smoothing on the low-frequency component image to obtain the reference image corresponding to the initial indoor environment image.

[0015] Further, the adaptive reflection interference elimination processing on each pixel point in the initial indoor environment image to obtain the target gray value corresponding to the pixel point comprises:

[0016] performing negative correlation mapping on the accumulated sum of the gray values corresponding to each pixel point in the preset convolution window corresponding to the pixel point to obtain a first gray value;

[0017] determining the product of the gray value corresponding to each pixel point in the preset convolution window corresponding to the pixel point and the first gray value as the reference weight corresponding to each pixel point in the preset convolution window, to obtain a reference weight set;

[0018] determining the product of each reference weight in the reference weight set and the gray value corresponding to each pixel point in the preset sliding window corresponding to the pixel point as a first index, to obtain a first index set;

[0019] determining the accumulated sum of each first index in the first index set as the target gray value corresponding to the pixel point.

[0020] Further, the method further comprises:

[0021] performing feature point detection processing on each obtained target environment image to obtain a target feature point set included in the target environment image;

[0022] performing optical flow matching processing on the two obtained target environment images according to the two obtained target feature point sets to obtain a matching optical flow field;

[0023] According to the matching optical flow field, the pixel points in the obtained two target environment images are grouped to obtain a pixel point group set corresponding to each target environment image, and each pixel point group in the pixel point group set is subjected to importance analysis processing to obtain an importance level corresponding to the pixel point group;

[0024] According to the importance level corresponding to each pixel point group in the pixel point group set corresponding to each target environment image, the pixel point groups in the target environment image are adaptively encoded to obtain a target coding set corresponding to the target environment image;

[0025] According to the target coding set corresponding to the target environment image, the obtained two target environment images are compressed.

[0026] Further, the optical flow matching processing of the obtained two target environment images according to the obtained two target feature point sets to obtain a matching optical flow field comprises:

[0027] According to the acquisition time corresponding to the obtained two target feature point sets, the two target feature point sets are determined as a first target feature point set and a second target feature point set respectively;

[0028] For each first target feature point in the first target feature point set and each second target feature point in the second target feature point set, a target difference index between the first target feature point and the second target feature point is determined according to a preset window corresponding to the first target feature point and a preset window corresponding to the second target feature point;

[0029] For each first target feature point in the first target feature point set, a second target feature point with the smallest target difference index between the first target feature point and the second target feature point is selected from the second target feature point set as a third target feature point corresponding to the first target feature point;

[0030] When the target difference index between the first target feature point in the first target feature point set and the third target feature point corresponding to the first target feature point is less than or equal to a pre-set difference threshold, the first target feature point and the third target feature point corresponding to the first target feature point are determined as two matching feature points matched with each other;

[0031] The matching feature points in the first target feature point set are combined into a matching feature point set, and the matching feature points in the second target feature point set are combined into another matching feature point set, to obtain two matching feature point sets;

[0032] The optical flow field between the obtained two matching feature point sets is determined as a matching optical flow field.

[0033] Further, the target difference index between the first target feature point and the second target feature point is determined according to the preset window corresponding to the first target feature point and the preset window corresponding to the second target feature point, including:

[0034] The reference image corresponding to the target environment image in which the first target feature point is located is determined as a first reference image, and the reference image corresponding to the target environment image in which the second target feature point is located is determined as a second reference image.

[0035] The pixel point at the same position as the first target feature point in the first reference image is determined as a first reference feature point, and the pixel point at the same position as the second target feature point in the second reference image is determined as a second reference feature point.

[0036] The gray values of the pixel points at the same position in the preset window corresponding to the first target feature point and the first reference feature point are combined as a first coordinate to obtain a first coordinate sequence, and the gray values of the pixel points at the same position in the preset window corresponding to the second target feature point and the second reference feature point are combined as a second coordinate to obtain a second coordinate sequence.

[0037] The Euclidean distance between the coordinates at the same position in the first coordinate sequence and the second coordinate sequence is determined as a first distance to obtain a first distance sequence.

[0038] The mean value of all first distances in the first distance sequence is determined as a first difference index.

[0039] The first difference index is normalized to obtain the target difference index between the first target feature point and the second target feature point.

[0040] Further, the pixel points in the two target environment images obtained are grouped according to the matching optical flow field to obtain a pixel point group set corresponding to each target environment image, including:

[0041] Each velocity vector in the matching optical flow field is clustered to obtain a clustering cluster set.

[0042] The matching feature points corresponding to each velocity vector in each clustering cluster in the clustering cluster set are determined as motion feature points to obtain a motion feature point group corresponding to the clustering cluster.

[0043] Group the motion feature points in the motion feature point group corresponding to each cluster, to obtain a first motion feature point group and a second motion feature point group corresponding to the cluster, wherein the motion feature points in the first motion feature point group are all located in the first target environment image, the motion feature points in the second motion feature point group are all located in the second target environment image, and the first target environment image and the second target environment image are two target environment images obtained;

[0044] Combine the pixel points in the minimum circumscribed rectangle corresponding to each first motion feature point group in the first target environment image into a pixel point group, and combine the pixel points other than the pixel points in the minimum circumscribed rectangle corresponding to each first motion feature point group in the first target environment image into a pixel point group, to obtain a pixel point group set corresponding to the first target environment image;

[0045] Combine the pixel points in the minimum circumscribed rectangle corresponding to each second motion feature point group in the second target environment image into a pixel point group, and combine the pixel points other than the pixel points in the minimum circumscribed rectangle corresponding to each second motion feature point group in the second target environment image into a pixel point group, to obtain a pixel point group set corresponding to the second target environment image.

[0046] Further, the important degree analysis processing on each pixel point group in the pixel point group set is performed to obtain an importance level corresponding to the pixel point group, including:

[0047] The region where the pixel point group is located is input into the importance degree discrimination network which is pre-trained to obtain an importance level corresponding to the pixel point group;

[0048] The training process of the importance degree discrimination network includes:

[0049] The importance degree discrimination network is constructed;

[0050] The target region set and the importance level corresponding to each target region in the target region set are obtained;

[0051] According to the target region set and the importance level corresponding to each target region in the target region set, the constructed importance degree discrimination network is trained to obtain the trained importance degree discrimination network.

[0052] Further, the importance level corresponding to each pixel point group in the pixel point group set corresponding to each target environment image is used to adaptively encode the pixel point group in the target environment image, including:

[0053] According to the importance level corresponding to each pixel point group in the pixel point group set corresponding to the target environment image, each pixel point group in the pixel point group set is subjected to Huffman coding to obtain a target code corresponding to the pixel point group.

[0054] In a second aspect, the present application provides a collision avoidance system, comprising a processor and a memory, wherein the processor is configured to execute instructions stored in the memory to implement the image processing method for indoor environment.

[0055] The present application has the following advantages:

[0056] The image processing method for indoor environment provided by the present application improves the image quality and the safety of traffic by processing the initial indoor environment image. First, since the initial indoor environment image often contains the indoor environment information around the target vehicle, obtaining the initial indoor environment image can facilitate subsequent collision avoidance object analysis of the driving environment around the target vehicle. Furthermore, obtaining two initial indoor environment images corresponding to the target vehicle can facilitate subsequent analysis of the running track of the collision avoidance object and subsequent collision avoidance prompt for the target vehicle. The collision avoidance object can be an object that needs to be avoided by the vehicle. For example, the collision avoidance object can be, but is not limited to, a vehicle, a person, and a column that needs to be avoided by the vehicle. The collision avoidance prompt can be a prompt to the driver to avoid the collision avoidance object. Then, the ambient light extraction processing is performed on each initial indoor environment image to obtain a reference image representing the intensity of ambient light. Then, based on the reference image and the preset sliding window and the preset convolution window corresponding to each pixel point in the initial indoor environment image, the adaptive reflection interference elimination processing is performed on the pixel points in the initial indoor environment image, which can eliminate the reflection interference of the pixel points in the initial indoor environment image caused by the intensity of ambient light to some extent, thereby improving the quality of the initial indoor environment image and the accuracy of subsequent optical flow matching processing of the two target environment images, and facilitating subsequent collision avoidance object analysis. BRIEF DESCRIPTION OF DRAWINGS

[0057] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, the drawings needed in the embodiments or the prior art description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can be obtained without creative labor based on these drawings.

[0058] Figure 1 The flowchart of the image processing method for indoor environment of the present application;

[0059] Figure 2 The flowchart of some steps that can be included in the present application;

[0060] Figure 3 The position diagram of the first reference feature point and the first target feature point according to the present application;

[0061] Figure 4 A light flow matching schematic diagram according to the present application;

[0062] Figure 5 An encoding process schematic diagram according to the present application.

[0063] Wherein, the reference signs include: an initial indoor environment image 301, a target environment image 302, a reference image 303, a first target feature point 304 and a first reference feature point 305. DETAILED DESCRIPTION

[0064] In order to further clarify the technical means and effects taken by the present application to achieve the predetermined object of the application, the specific embodiments, structures, features and effects of the technical solutions proposed by the present application are described in detail below in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0065] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0066] The present application provides an image processing method for indoor environment, which comprises the following steps:

[0067] Obtaining two frames of initial indoor environment images corresponding to a target vehicle;

[0068] Performing ambient light extraction processing on each frame of initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image;

[0069] Based on the reference image corresponding to each frame of initial indoor environment image, and the preset sliding window and the preset convolution window corresponding to each pixel point in the initial indoor environment image, performing adaptive reflection interference elimination processing on each pixel point in the initial indoor environment image to obtain a target gray value corresponding to the pixel point;

[0070] Updating the gray value corresponding to each pixel point in each frame of initial indoor environment image to the target gray value corresponding to the pixel point to obtain a target environment image corresponding to the initial indoor environment image.

[0071] The above steps are described in detail as follows:

[0072] Reference Figure 1 Fig. 1 shows the flow of some embodiments of an image processing method for indoor environment according to the present application. The image processing method for indoor environment comprises the following steps:

[0073] Step S1, obtaining two frames of initial indoor environment images corresponding to the target vehicle.

[0074] In some embodiments, two frames of initial indoor environment images corresponding to the target vehicle can be obtained.

[0075] The target vehicle can be a vehicle in an indoor parking place to be detected for an environmental obstacle. The indoor parking place can be a place where a vehicle can be parked indoors. For example, the indoor parking place can be an underground parking lot. The environmental obstacle detection can be used to detect whether there is a collision-free object in the vicinity of the target vehicle that has a collision risk with the target vehicle. The collision-free object can be an object that needs to be avoided by the vehicle. For example, the collision-free object can be, but is not limited to, a vehicle, a person, and a column that needs to be avoided by the vehicle. The reference image can be an initial indoor environment image with foreground information removed. The two frames of initial indoor environment images can be two frames of images obtained by preprocessing two frames of vehicle environment images. The two frames of vehicle environment images can be two frames of images of the environment near the target vehicle captured by a camera installed on the target vehicle. For example, the two frames of vehicle environment images can be two adjacent frames of images in a video captured by a camera installed on the target vehicle. The preprocessing can include, but is not limited to, grayscale processing, light compensation processing, and contrast enhancement processing.

[0076] It should be noted that since the initial indoor environment image often contains indoor environment information near the target vehicle, obtaining the initial indoor environment image can facilitate subsequent analysis of the collision-free object in the driving environment near the target vehicle, and obtaining two frames of initial indoor environment images corresponding to the target vehicle can facilitate subsequent analysis of the running track of the collision-free object, and can facilitate subsequent collision-free prompting of the target vehicle. The collision-free prompt can be a prompt to the driver to avoid the collision-free object.

[0077] As an example, the present step can include the following steps:

[0078] First, two frames of vehicle environment images are collected by a camera installed on the target vehicle.

[0079] The time interval between the collection of the two frames of vehicle environment images can be a pre-set time interval. For example, the time interval between the collection of the two frames of vehicle environment images can be seconds.

[0080] Second, each vehicle environment image is preprocessed to obtain an initial indoor environment image.

[0081] For example, preprocessing each vehicle environment image to obtain an initial indoor environment image can include the following sub-steps:

[0082] The first sub-step is to perform a grayscale processing on the vehicle environment image to obtain a grayscale image.

[0083] For example, a grayscale algorithm can be used to perform the grayscale processing on the vehicle environment image to obtain the grayscale image. The grayscale algorithm can be, but is not limited to, a maximum value method, an average value method, and a weighted average value method.

[0084] The second sub-step is to perform a light compensation processing on the grayscale image to obtain a compensated light image.

[0085] For example, a PS (Adobe Photoshop, image processing software) technique can be used to perform the light compensation processing on the grayscale image to obtain the compensated light image.

[0086] The third sub-step is to perform a contrast enhancement processing on the compensated light image to obtain an initial indoor environment image.

[0087] For example, a histogram equalization algorithm can be used to perform the contrast enhancement processing on the compensated light image to obtain the initial indoor environment image.

[0088] The step S2 is to perform an ambient light extraction processing on each initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image.

[0089] In some embodiments, the ambient light extraction processing can be performed on each initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image.

[0090] It should be noted that the ambient light extraction processing on each initial indoor environment image can obtain a reference image representing the intensity of the ambient light, which can facilitate a subsequent adaptive reflection interference elimination processing on the pixel points in the initial indoor environment image to remove the reflection interference in the initial indoor environment image caused by the intensity of the ambient light.

[0091] As an example, the step can include the following steps:

[0092] The first step is to perform a high frequency elimination processing on the initial indoor environment image to obtain a low frequency component image.

[0093] For example, a low-pass filter can be used to perform the high frequency elimination processing on the initial indoor environment image to obtain the low frequency component image.

[0094] The filter cutoff frequency for the high frequency elimination processing can be pre-set. For example, the filter cutoff frequency can be set as small as possible, such as 5 Hz.

[0095] The second step is to smooth the low frequency component image to obtain the reference image corresponding to the initial indoor environment image.

[0096] For example, a minimum filter can be used to smooth the low-frequency component image, thereby removing almost all foreground information and suppressing part of the highlight area of the reflection light, leaving only a blurred dark background image as a reference image of the ambient light of the initial indoor environment image. There is almost no foreground feature information on the reference image, and therefore the gray value corresponding to the pixel point in the reference image can represent the strength of the background ambient light. The filter window when using the minimum filter can be a 11x11 window. Since the light change has a great influence on the optical flow field when the object moves in the room, the reference image representing the strength of the ambient light can facilitate subsequent adaptive reflection interference elimination processing of the pixel points in the initial indoor environment image.

[0097] In step S3, based on the reference image corresponding to each frame of the initial indoor environment image and the preset sliding window and the preset convolution window corresponding to each pixel point in the initial indoor environment image, an adaptive reflection interference elimination processing is performed on each pixel point in the initial indoor environment image to obtain a target gray value corresponding to the pixel point.

[0098] In some embodiments, based on the reference image corresponding to each frame of the initial indoor environment image and the preset sliding window and the preset convolution window corresponding to each pixel point in the initial indoor environment image, an adaptive reflection interference elimination processing is performed on each pixel point in the initial indoor environment image to obtain a target gray value corresponding to the pixel point.

[0099] The preset sliding window can be a preset sliding window. The preset convolution window can be a preset convolution window. The preset sliding window and the preset convolution window can have the same size. For example, the preset sliding window can be a 5x5 sliding window. The preset convolution window can be a 5x5 convolution window. The pixel point can be located at the center of the preset sliding window, i.e., the pixel point is located at the center of the preset sliding window corresponding to the pixel point. The pixel point at the same position as the pixel point in the reference image can be located at the center of the preset convolution window, and the reference image can be the reference image corresponding to the initial indoor environment image in which the pixel point is located.

[0100] It should be noted that, based on the reference image, the adaptive reflection interference elimination processing is performed on the pixel points in the initial indoor environment image, which can eliminate the reflection interference in the initial indoor environment image to some extent, thereby improving the quality of the initial indoor environment image, and facilitating subsequent collision-free object analysis.

[0101] As an example, the present step can include the following steps:

[0102] First, the cumulative sum of the gray values corresponding to each pixel point in the preset convolution window corresponding to the pixel point is negatively correlated to obtain a first gray value.

[0103] The gray value corresponding to the pixel point can be the gray value of the pixel point.

[0104] In the second step, the product of the gray value corresponding to each pixel point in the preset convolution window corresponding to the pixel point and the first gray value is determined as the reference weight corresponding to each pixel point in the preset convolution window, and a reference weight set is obtained.

[0105] In the third step, the product of each reference weight in the reference weight set and the gray value corresponding to each pixel point in the preset sliding window corresponding to the pixel point is determined as a first index, and a first index set is obtained.

[0106] For example, the product of each reference weight and the gray value corresponding to each pixel point in the preset sliding window corresponding to the pixel point can be determined as a first index, a sub-first index set corresponding to each reference weight is obtained, and finally a first index set is obtained. The sub-first index set corresponding to the reference weight can include a plurality of first indexes determined by the product of the reference weight and the gray value corresponding to each pixel point in the preset sliding window corresponding to the pixel point. The first index set can include the sub-first index set corresponding to all reference weights in the reference weight set.

[0107] In the fourth step, the cumulative sum of each first index in the first index set is determined as the target gray value corresponding to the pixel point.

[0108] For example, the formula for determining the target gray value corresponding to the pixel point can be:

[0109]

[0110] wherein, is the target gray value corresponding to the kth pixel point in the ith initial indoor environment image. Q is the number of pixel points in the preset sliding window or the preset convolution window. If the size of the preset sliding window and the preset convolution window is 5x5, Q can be 25. is the gray value corresponding to the rth pixel point in the preset convolution window corresponding to the kth pixel point in the ith initial indoor environment image. is the gray value corresponding to the nth pixel point in the preset sliding window corresponding to the kth pixel point in the ith initial indoor environment image. γ1 is a factor greater than 0 set in advance, mainly to prevent the denominator from being 0, such as γ1 can be 0.01. i is the frame number of the initial indoor environment image. k is the serial number of the pixel point in the ith initial indoor environment image. r is the serial number of the pixel point in the preset convolution window. n is the serial number of the pixel point in the preset sliding window. is the first gray value corresponding to the kth pixel point in the ith initial indoor environment image. is a reference weight corresponding to an rth pixel point in a preset convolution window corresponding to a kth pixel point in an ith frame of initial indoor environment image. is a first index between an nth pixel point in a preset sliding window corresponding to the kth pixel point and an rth pixel point in the preset convolution window corresponding to the kth pixel point. a negative correlation mapping of can be realized.

[0111] It should be noted that if the objects in the two frames of initial indoor environment images are always in an area with stable light environment, the sparsity of the optical flow field will not change greatly, and if the objects in the two frames of initial indoor environment images are in a gray change area, the optical flow field will be more sparse. However, the existing method assumes that the gray values of all pixel points on the dynamic target object change linearly with the change of light, and the change coefficients of all pixel points need to be uniform in order to obtain the optical flow matching result by assuming that the gradient variables of adjacent pixel points are consistent. However, this assumption is often not true. The biggest disturbance problem of environmental light change is that objects often have different reflection effects on light due to different colors and materials, and the corresponding gray value changes of objects with different colors and materials are often different under the same light intensity change. For example, the gray value changes of the car window and the car body are often not equal when the light changes. Therefore, adaptive reflection interference elimination processing is performed on the initial indoor environment image, which can eliminate the reflection interference caused by different colors and materials of the same object to some extent, and can facilitate subsequent optical flow matching processing. Specifically, the is taken as a convolution weight to convolve the initial indoor environment image to obtain new pixel gray values. The convolved image (target environment image) can highlight the light environment again and weaken the difference in gray value changes caused by different colors and materials of the objects. If the pixel region is not affected by the change of the light environment, that is, the light environment is stable and unchanged, the convolution value is the average value, so it will not have an impact. Only the gray values of the environmental light change area are corrected.

[0112] In step S4, the gray value corresponding to each pixel point in each frame of initial indoor environment image is updated to the target gray value corresponding to the pixel point to obtain the target environment image corresponding to the initial indoor environment image.

[0113] In some embodiments, the gray value corresponding to each pixel point in each frame of initial indoor environment image can be updated to the target gray value corresponding to the pixel point to obtain the target environment image corresponding to the initial indoor environment image.

[0114] As an example, the gray value corresponding to each pixel point in the above initial indoor environment image can be replaced by the target gray value corresponding to the pixel point to obtain the target environment image corresponding to the initial indoor environment image.

[0115] Reference Figure 2 The application can further include the following steps:

[0116] Step S5, performing feature point detection processing on each obtained target environment image to obtain a target feature point set included in the target environment image.

[0117] In some embodiments, the feature point detection processing can be performed on each obtained target environment image to obtain a target feature point set included in the target environment image.

[0118] The target feature points in the target feature point set can be corner points in the target environment image.

[0119] It should be noted that the target environment image is subjected to feature point detection processing, which can facilitate subsequent optical flow matching processing of the two obtained target environment images based on the target feature point set.

[0120] As an example, the target environment image can be subjected to corner point detection by a corner point detection algorithm, and the corner points detected by the corner point detection are determined as target feature points. The corner point detection algorithm can be, but is not limited to, a Shi-Tomasi corner point detection algorithm, a Harris corner point detection algorithm, and an opencv corner point detection algorithm.

[0121] Step S6, performing optical flow matching processing on the two obtained target environment images based on the two obtained target feature point sets to obtain a matching optical flow field.

[0122] In some embodiments, the optical flow matching processing can be performed on the two obtained target environment images based on the two obtained target feature point sets to obtain a matching optical flow field.

[0123] It should be noted that the optical flow matching processing of the two target environment images based on the target feature point set can obtain a more accurate matching optical flow field, which can facilitate subsequent accurate analysis of the running trajectory of the collision-free object.

[0124] As an example, this step can include the following steps:

[0125] First, according to the acquisition time corresponding to the two target feature point sets, the two target feature point sets are respectively determined as a first target feature point set and a second target feature point set.

[0126] The acquisition time corresponding to the target feature point set can be the acquisition time of the vehicle environment image corresponding to that target feature point set. The vehicle environment image corresponding to the target feature point set can be the vehicle environment image corresponding to the target environment image containing that target feature point set. The vehicle environment image corresponding to the target environment image can be any vehicle environment image involved in determining that target environment image.

[0127] For example, the target feature points in the target feature point set with the earlier acquisition time can be identified as the first target feature point, thus obtaining the first target feature point set, and the target feature points in the target feature point set with the later acquisition time can be identified as the second target feature point, thus obtaining the second target feature point set.

[0128] The second step involves determining the target difference index between the first target feature point and the second target feature point for each first target feature point in the first target feature point set and each second target feature point in the second target feature point set, based on the preset window corresponding to the first target feature point and the preset window corresponding to the second target feature point.

[0129] For example, determining the target difference index between the first target feature point and the second target feature point based on the preset window corresponding to the first target feature point and the preset window corresponding to the second target feature point may include the following sub-steps:

[0130] The first sub-step involves determining the reference image corresponding to the target environment image where the first target feature point is located as the first reference image, and determining the reference image corresponding to the target environment image where the second target feature point is located as the second reference image.

[0131] The reference image corresponding to the target environment image can be a reference image used to determine the target environment image.

[0132] The second sub-step involves determining the pixels in the first reference image that are at the same position as the first target feature point as the first reference feature point, and determining the pixels in the second reference image that are at the same position as the second target feature point as the second reference feature point.

[0133] For example, such as Figure 3 As shown, the target environment image 302 can be the target environment image corresponding to the initial indoor environment image 301. The reference image 303 can be the reference image corresponding to the initial indoor environment image 301. The first reference feature point 305 can be a pixel in the reference image 303 at the same position as the first target feature point 304.

[0134] The third sub-step is to combine the gray values of the pixel points at the same positions in the preset window corresponding to the first target feature point and the first reference feature point into a first coordinate to obtain a first coordinate sequence, and combine the gray values of the pixel points at the same positions in the preset window corresponding to the second target feature point and the second reference feature point into a second coordinate to obtain a second coordinate sequence.

[0135] For example, first, the pixel points in the preset window corresponding to the first target feature point can be sorted in the order from left to right and from top to bottom, and the obtained pixel point sequence is taken as a first target pixel point sequence. Then, the pixel points in the preset window corresponding to the first reference feature point can be sorted in the order from left to right and from top to bottom, and the obtained pixel point sequence is taken as a second target pixel point sequence. Then, the gray values of the pixel points at the same positions in the first target pixel point sequence and the second target pixel point sequence are combined into a first coordinate to obtain a first coordinate sequence. The abscissa of the first coordinate can be the gray value corresponding to the first target pixel point. The ordinate of the first coordinate can be the gray value corresponding to the second target pixel point.

[0136] Further, the pixel points in the preset window corresponding to the second target feature point can be sorted in the order from left to right and from top to bottom, and the obtained pixel point sequence is taken as a third target pixel point sequence. Then, the pixel points in the preset window corresponding to the second reference feature point can be sorted in the order from left to right and from top to bottom, and the obtained pixel point sequence is taken as a fourth target pixel point sequence. Finally, the gray values of the pixel points at the same positions in the third target pixel point sequence and the fourth target pixel point sequence are combined into a second coordinate to obtain a second coordinate sequence. The abscissa of the second coordinate can be the gray value corresponding to the third target pixel point. The ordinate of the second coordinate can be the gray value corresponding to the fourth target pixel point.

[0137] The fourth sub-step is to determine the Euclidean distance between the coordinates at the same positions in the first coordinate sequence and the second coordinate sequence as a first distance to obtain a first distance sequence.

[0138] The fifth sub-step is to determine the mean value of all the first distances in the first distance sequence as a first difference index.

[0139] For example, the formula for determining the first difference index between the first target feature point and the second target feature point can be:

[0140]

[0141] wherein, MSE abis a first difference index between the ath first target feature point in the first target feature point set and the bth second target feature point in the second target feature point set. Q is the number of pixel points in the preset sliding window. If the size of the preset sliding window is 5x5, Q can be 25.g an is the gray value corresponding to the nth pixel point in the preset window corresponding to the ath first target feature point.G an is the gray value corresponding to the nth pixel point in the preset window corresponding to the first reference feature point corresponding to the ath first target feature point.g bn is the gray value corresponding to the nth pixel point in the preset window corresponding to the bth second target feature point.G bn is the gray value corresponding to the nth pixel point in the preset window corresponding to the second reference feature point corresponding to the bth second target feature point.(g an , G an is the nth first coordinate in the first coordinate sequence.g bn , G bn is the nth second coordinate in the second coordinate sequence. is the Euclidean distance between the nth first coordinate and the nth second coordinate, that is, the nth first distance in the first distance sequence. a is the serial number of the first target feature point in the first target feature point set. b is the serial number of the second target feature point in the second target feature point set. n is the serial number of the pixel point in the preset sliding window. Since the number of elements in the first coordinate sequence and the second coordinate sequence is equal to the number of pixel points in the preset sliding window, and the elements in the first coordinate sequence and the second coordinate sequence can be composed of the gray values corresponding to the pixel points at the corresponding positions in the corresponding preset sliding window, n can also be the serial number of the first coordinate in the first coordinate sequence.

[0142] It should be noted that, can represent the ambient light difference between the pixel points at the same position in the preset sliding window corresponding to the ath first target feature point and the bth second target feature point between the corresponding reference images, and the gray difference between the pixel points at the same position in the preset sliding window corresponding to the ath first target feature point and the bth second target feature point between the corresponding target environment images. can represent the ambient light difference between the ath first target feature point and the bth second target feature point between the corresponding reference images, and the gray difference between the ath first target feature point and the bth second target feature point between the corresponding target environment images. Therefore, when the first difference index is smaller, it is often that the ath first target feature point and the bth second target feature point are more likely to match each other, and it is often that the ath first target feature point and the bth second target feature point are more likely to be pixel points corresponding to the same object.

[0143] A sixth sub-step is to normalize the first difference index to obtain a target difference index between the first target feature point and the second target feature point.

[0144] A third step is to, for each first target feature point in the first target feature point set, select a second target feature point with the smallest target difference index between the first target feature point from the second target feature point set as a third target feature point corresponding to the first target feature point.

[0145] A fourth step is to, when the target difference index between a first target feature point in the first target feature point set and the third target feature point corresponding to the first target feature point is less than or equal to a difference threshold, determine the first target feature point and the third target feature point corresponding to the first target feature point as two matching feature points that match each other.

[0146] The difference threshold can be a maximum target difference index that is considered to be a target feature point corresponding to the same position of the same object of the first target feature point and the third target feature point. For example, the difference threshold can be 0.2.

[0147] For example, when the target difference index between a first target feature point in the first target feature point set and the third target feature point corresponding to the first target feature point is less than or equal to the difference threshold, the first target feature point can be determined as a matching feature point, and the third target feature point corresponding to the first target feature point can be determined as a matching feature point, and the two matching feature points obtained are matched with each other.

[0148] A fifth step is to combine the matching feature points in the first target feature point set as a matching feature point set, and combine the matching feature points in the second target feature point set as another matching feature point set, to obtain two matching feature point sets.

[0149] A sixth step is to determine the optical flow field between the two matching feature point sets obtained as a matching optical flow field.

[0150] For example, first, matching feature points that match each other can be selected from the two matching feature point sets obtained, and each two matching feature points that match each other selected can be combined as a matching feature point group to obtain a matching feature point group set. Then, the speed between the two matching feature points in each matching feature point group in the matching feature point group set can be determined as a velocity vector that constitutes the matching optical flow field. The matching feature point group in the matching feature point group set can correspond to the velocity vector that constitutes the matching optical flow field one by one. That is, each velocity vector in the matching optical flow field corresponds to two matching feature points.

[0151] For example, the manner of determining the speed between two matching feature points in the matching feature point group can be: moving the two matching feature points to a target coordinate system, determining the value corresponding to the Euclidean distance between the two matching feature points in the coordinate system as the speed between the two matching feature points, connecting the two matching feature points, and determining the direction of the connecting line from the matching feature point with the earlier collection time to the matching feature point with the later collection time as the direction of the speed between the two matching feature points. The collection time of the matching feature point can be the collection time of the vehicle environment image corresponding to the target environment image in which the matching feature point is located. The target coordinate system can be a coordinate system with the upper left corner of the target environment image as the origin, the direction in which the width of the target environment image is located as the horizontal axis, and the direction in which the height of the target environment image is located as the vertical axis.

[0152] It should be noted that when the motion speed is slow, the change of the ambient light in the images of the continuous frames is basically gradual, so the optical flow tracking is generally not interrupted, but when the moving object is in an area with a dramatic change in ambient light, the optical flow is likely to be extremely sparse and distorted, the optical flow tracking is interrupted or a large error occurs, and the optical flow field is chaotic. For the problem of sparse optical flow caused by the change of ambient light between adjacent frames of images, the optical flow vectors are concentrated in the same or similar background light, and the matching range is converged, which can make the optical flow field of the same background light part more dense and accurate. For adjacent frames of images, the matching result of the optical flow in the area with the same ambient light is more reliable, and the matching result of the optical flow in the area with a dramatic change in ambient light is often unreliable, so the density of the optical flow field of the former is improved, the matching of the pixel points in the similar background environment is formed, the matching of the dark areas and the matching of the bright areas are formed, and the optical flow vectors are concentrated in the same ambient light. When the optical flow matching is performed between the areas with a large difference between the bright and dark ambient light, the feature points are consumed and the accuracy of the optical flow matching result is poor. After the matching range is converged, although the optical flow becomes more sparse when the moving object is in an area with a dramatic change in light and dark, the optical flow in the area under the same ambient light becomes more dense and accurate, and accurate optical flow vectors can be obtained through the local optical flow field, and stable optical flow tracking is maintained. The matching result of the pixel points in the area with a dramatic change in light and dark is extremely unstable, the optical flow field is sparse and chaotic. The purpose of the present application is to concentrate the optical flow vectors in the same ambient light area, as shown in FIG. 1, so that the optical flow vectors in the areas with a dramatic change in light and dark in the two frames of images are almost very sparse, but the optical flow vectors in the same dark area (the area that is dark in both frames of images) and the same bright area (the area that is bright in both frames of images) become more dense and the optical flow matching accuracy is high. Figure 4

[0153] ​The existing optical flow method assumes that the brightness of the motion pixel points in the adjacent two frames is constant. However, the motion pixel points cannot maintain constant brightness in the dynamic change process, and in the indoor environment of the underground parking lot, the multiple direction car lights and the installed high height lighting light source often lead to large differences in brightness in different areas. In the image with poor light environment, the image details are lost more, so the obtained optical flow field is generally sparse optical flow (speed). When the environment light of the moving object changes, the optical flow vector (speed) is still relatively sparse and lacks accuracy. Therefore, the obtained two target environment images are subjected to optical flow matching processing, which can improve this problem to a certain extent.

[0154] Step S7, according to the matching optical flow field, grouping the pixel points in the obtained two target environment images to obtain a pixel point group set corresponding to each target environment image, and performing importance analysis processing on each pixel point group in the pixel point group set to obtain an importance level corresponding to the pixel point group.

[0155] In some embodiments, the pixel points in the obtained two target environment images can be grouped according to the matching optical flow field to obtain a pixel point group set corresponding to each target environment image, and importance analysis processing can be performed on each pixel point group in the pixel point group set to obtain an importance level corresponding to the pixel point group.

[0156] It should be noted that the importance analysis processing of each pixel point group in the target environment image can improve the accuracy of the importance level determination of the pixel point group, facilitate the subsequent adaptive coding of the pixel point group in the target environment image, avoid encoding important information as easily lost coding information, reduce the loss of important information, and thus improve the safety of traffic.

[0157] As an example, this step can include the following steps:

[0158] First, clustering each velocity vector in the matching optical flow field to obtain a cluster set.

[0159] For example, clustering each velocity vector in the matching optical flow field to obtain a cluster set can include the following sub-steps:

[0160] First sub-step, determining the target distance between any two velocity vectors in the matching optical flow field.

[0161] For example, the matching feature points corresponding to any two velocity vectors can be moved to a target coordinate system, and in the coordinate system, the Euclidean distance between the first matching feature point and the second matching feature point is determined as a first Euclidean distance, the Euclidean distance between the third matching feature point and the fourth matching feature point is determined as a second Euclidean distance, and the mean of the first Euclidean distance and the second Euclidean distance is determined as the target distance between the two velocity vectors. The first matching feature point and the third matching feature point can be two matching feature points corresponding to the same velocity vector. The second matching feature point and the fourth matching feature point can be two matching feature points corresponding to the same velocity vector. The first matching feature point and the second matching feature point can be pixel points in the same frame of the target environment image. The third matching feature point and the fourth matching feature point can be pixel points in the same frame of the target environment image.

[0162] In a second sub-step, based on the target distance between the velocity vectors, each velocity vector in the matching optical flow field is clustered to obtain a cluster set.

[0163] The cluster in the cluster set can be a cluster composed of velocity vectors corresponding to the same object.

[0164] For example, for any two velocity vectors in the matching optical flow field, if the included angle between the directions of the two velocity vectors is less than or equal to a pre-set included angle threshold, the absolute value of the difference between the rates corresponding to the two velocity vectors is less than or equal to a pre-set difference threshold, and the target distance between the two velocity vectors is less than or equal to a pre-set distance threshold, the two velocity vectors are clustered into the same cluster. The included angle threshold can be the maximum included angle between the directions of two velocity vectors when the matching feature points corresponding to the two velocity vectors are considered to be pixel points corresponding to the same object. For example, the included angle threshold can be 1°. The difference threshold can be the maximum absolute value of the difference between the rates of two velocity vectors when the matching feature points corresponding to the two velocity vectors are considered to be pixel points corresponding to the same object. For example, the difference threshold can be 0.05. The distance threshold can be the maximum target distance when the matching feature points corresponding to the two velocity vectors are considered to be pixel points corresponding to the same object. For example, the distance threshold can be 1.

[0165] In a second step, the matching feature points corresponding to each velocity vector in each cluster in the cluster set are determined as motion feature points to obtain a motion feature point group corresponding to the cluster.

[0166] The motion feature point group corresponding to the cluster can be composed of the matching feature points corresponding to each velocity vector in the cluster. The number of motion feature points in the motion feature point group corresponding to the cluster can be equal to twice the number of velocity vectors in the cluster.

[0167] Thirdly, grouping the motion feature points in the motion feature point group corresponding to each cluster to obtain the first motion feature point group and the second motion feature point group corresponding to the cluster.

[0168] The motion feature points in the first motion feature point group can be located in the first target environment image. The motion feature points in the second motion feature point group can be located in the second target environment image. The first target environment image and the second target environment image can be two target environment images obtained. The objects corresponding to the first motion feature point group and the second motion feature point group corresponding to the cluster can be the same.

[0169] Fourthly, combining the pixel points in the minimum circumscribed rectangle corresponding to each first motion feature point group in the first target environment image into a pixel point group, and combining the pixel points other than the pixel points in the minimum circumscribed rectangle corresponding to each first motion feature point group in the first target environment image into a pixel point group to obtain a pixel point group set corresponding to the first target environment image.

[0170] The minimum circumscribed rectangle corresponding to the motion feature point group can be the minimum circumscribed rectangle of all motion feature points in the motion feature point group. Each pixel point in the minimum circumscribed rectangle corresponding to the motion feature point group can be a pixel point corresponding to the same object.

[0171] Fifthly, combining the pixel points in the minimum circumscribed rectangle corresponding to each second motion feature point group in the second target environment image into a pixel point group, and combining the pixel points other than the pixel points in the minimum circumscribed rectangle corresponding to each second motion feature point group in the second target environment image into a pixel point group to obtain a pixel point group set corresponding to the second target environment image.

[0172] Sixthly, inputting the region where the pixel point group is located into the importance degree discrimination network trained in advance to obtain the importance level corresponding to the pixel point group.

[0173] The importance degree discrimination network can be used to judge the importance level of the image region. The importance degree discrimination network can be a convolutional neural network.

[0174] For example, the training process of the importance degree discrimination network can include the following sub-steps:

[0175] First sub-step, constructing the importance degree discrimination network.

[0176] For example, a convolutional neural network can be constructed as the importance degree discrimination network before training.

[0177] Second sub-step, obtaining the target region set and the importance level corresponding to each target region in the target region set.

[0178] The target regions in the target region set can be regions where objects are located in the indoor parking lot and are photographed by the camera installed on the vehicle. The objects can be, but are not limited to, vehicles, people, columns, and ground. The importance level corresponding to the target region can be an importance level set for the target region according to actual conditions.

[0179] For example, if the target region set includes a target region where a vehicle is located, a target region where a person is located, a target region where a column is located, and a target region where ground is located, the importance level corresponding to the target region where the person is located can be set to a first level, representing the highest level, the importance level corresponding to the target region where the vehicle is located can be set to a second level, representing the second highest level, the importance level corresponding to the target region where the column is located can be set to a third level, representing the third highest level, and the importance level corresponding to the target region where the ground is located can be set to a fourth level, representing the fourth highest level.

[0180] In a third sub-step, the importance degree discrimination network is trained according to the target region set and the importance level corresponding to each target region in the target region set, and a trained importance degree discrimination network is obtained.

[0181] For example, the target region set can be used as a training set of the importance degree discrimination network, and the importance level corresponding to the target region can be used as a training label of the importance degree discrimination network. The importance degree discrimination network is trained to obtain a trained importance degree discrimination network.

[0182] In step S8, the pixel point groups in the target environment image are adaptively encoded according to the importance level corresponding to each pixel point group in the pixel point group set corresponding to each target environment image, and a target coding set corresponding to the target environment image is obtained.

[0183] In some embodiments, the pixel point groups in the target environment image can be adaptively encoded according to the importance level corresponding to each pixel point group in the pixel point group set corresponding to each target environment image, and a target coding set corresponding to the target environment image is obtained.

[0184] The target coding can be the coding of the pixel point group.

[0185] It should be noted that adaptively encoding the pixel point groups in the target environment image based on the importance level corresponding to each pixel point group can avoid encoding important information into coding information that is easy to lose, can reduce the loss of important information, and thus can improve the safety of traffic.

[0186] As an example, the pixel group set corresponding to the target environment image can be divided into two categories, respectively as an effective information category and an inefficient information category, according to the importance levels of the pixel groups in the pixel group set.

[0187] For example, if the pixel group set includes 6 pixel groups, and the importance levels of the 6 pixel groups can be as shown in Table 1.

[0188] Table 1

[0189] Serial number 1 2 3 4 5 Importance level First level Second level First level Fourth level Third level

[0190] As can be seen from Table 1, the importance level of the first pixel group in the pixel group set is one level, the importance level of the second pixel group is two levels, the importance level of the third pixel group is one level, the importance level of the fourth pixel group is four levels, and the importance level of the fifth pixel group is three levels. The encoding process diagram when the pixel groups are Huffman encoded can be as shown in Figure 5 Figure 5 It can be seen that the target code corresponding to the first pixel group with an importance level of one can be 0, the target code corresponding to the second pixel group with an importance level of two can be 110, the target code corresponding to the third pixel group with an importance level of one can be 10, the target code corresponding to the fourth pixel group with an importance level of four can be 11110, and the target code corresponding to the fifth pixel group with an importance level of three can be 1110.

[0191] For another example, the Huffman encoding of each pixel group in the pixel group set corresponding to the target environment image according to the importance levels of the pixel groups in the pixel group set can include the following sub-steps:

[0192] In a first sub-step, the pixel groups in the pixel group set are divided into two categories, respectively as an effective information category and an inefficient information category, according to the importance levels of the pixel groups in the pixel group set corresponding to the target environment image.

[0193] Among them, the object corresponding to the pixel group in the effective information category can be a collision-free object. The object corresponding to the pixel group in the inefficient information category can be an object that does not need to be avoided by the vehicle, for example, the ground photographed in the image.

[0194] For example, the pixel group with the lowest importance level can be classified as an inefficient information category, and the pixel group with an importance level other than the lowest can be classified as an effective information category.

[0195] ​In a second sub-step, the transmission weight of 0.5 is collectively allocated to the pixel points in the effective information class according to the redundancy, and the transmission weight of 0.5 is collectively allocated to the pixel points in the low-efficiency information class according to the redundancy, and all the pixel points are sorted according to the transmission weight, and a Huffman tree is constructed from the two pixel types with the lowest transmission weight upwards to obtain the target code corresponding to the pixel point group. The target code corresponding to the pixel point group can include the codes corresponding to each pixel point in the pixel point group. Subsequently, the two target environment images obtained can be transmitted according to the target code set corresponding to the target environment image, and the two target environment images obtained are transmitted to the collision avoidance system for collision avoidance object analysis.

[0196] It should be noted that, since the number of pixel points in the effective information class is usually lower than that of the remaining pixel points, the pixel points in the effective information class can obtain a higher transmission weight according to the above allocation ratio, and the all the pixel points are sorted according to the transmission weight, and a Huffman tree is constructed from the two pixel types with the lowest transmission weight upwards to obtain the encoding result of each pixel point group, so that the higher the transmission weight, the shorter the code length, and vice versa.

[0197] In step S9, the two target environment images obtained are compressed according to the target code set corresponding to the target environment image.

[0198] In some embodiments, the two target environment images obtained can be compressed according to the target code set corresponding to the target environment image.

[0199] It should be noted that, based on the target code set, the two target environment images obtained can be compressed to reduce the loss of important information during compression. Therefore, the target environment images with high quality are obtained, and the pixel point groups in the target environment images are adaptively encoded, which can reduce the loss of important information, thereby improving the safety of traffic.

[0200] As an example, the target environment image can be compressed according to the target code set corresponding to the target environment image.

[0201] Based on the same inventive concept as the above method embodiment, the present application provides a collision avoidance system, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the steps of the indoor environment image processing method are implemented when the above computer program is executed by the processor.

[0202] It should be noted that a collision-free system can be a software system that can implement the steps of an indoor environment image processing method, which can specifically be: the method of step S1 can be used to obtain two initial indoor environment images corresponding to the target vehicle. The method of step S2 can be used to perform ambient light extraction processing on each initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image. The method of step S3 can be used to perform adaptive reflection interference elimination processing on each pixel point in the initial indoor environment image based on the reference image corresponding to each initial indoor environment image, and a preset sliding window and a preset convolution window corresponding to each pixel point in the initial indoor environment image, to obtain a target gray value corresponding to the pixel point. The method of step S4 can be used to update the gray value corresponding to each pixel point in each initial indoor environment image to a target gray value corresponding to the pixel point, to obtain a target environment image corresponding to the initial indoor environment image. The methods of steps S5 to S9 can be used to compress the two target environment images obtained, which can specifically include: the method of step S5 can be used to perform feature point detection processing on each target environment image obtained to obtain a target feature point set included in the target environment image. The method of step S6 can be used to perform optical flow matching processing on the two target environment images obtained according to the two target feature point sets obtained, to obtain a matching optical flow field. The method of step S7 can be used to group the pixel points in the two target environment images obtained according to the matching optical flow field, to obtain a pixel point group set corresponding to each target environment image, and to perform importance analysis processing on each pixel point group in the pixel point group set, to obtain an importance level corresponding to the pixel point group. The method of step S8 can be used to perform adaptive coding on the pixel point groups in the target environment image according to the importance level corresponding to each pixel point group in the pixel point group set corresponding to each target environment image, to obtain a target coding set corresponding to the target environment image. The method of step S9 can be used to compress the two target environment images obtained according to the target coding set corresponding to the target environment image. According to the two target environment images stored, target environment prompt information can be determined, and the target environment prompt information can be sent to a target terminal for collision-free prompting.

[0203] The target environment prompt information can be prompt information about the environment near the target vehicle obtained based on the two target environment images. For example, the target environment prompt information can be "10 meters away, there is a pedestrian, please drive carefully."

[0204] As an example, first, a region where each pixel group in the pixel group set corresponding to the target environment image is located can be input into a pre-trained target recognition network to determine the category corresponding to the pixel group. The category corresponding to the pixel group can be, but is not limited to, a vehicle, a person, a column, and the ground. The target recognition network can be used to identify the category of the image region. For example, the target recognition network can be a convolutional neural network. Then, the target pixel group can be selected from the pixel group set according to the category corresponding to each pixel group in the pixel group set corresponding to the target environment image. The target pixel group can be a group composed of pixel points corresponding to an object that needs to be avoided. For example, the target pixel group can be a pixel group of the category of a person. Finally, the distance between the target vehicle and the object corresponding to the target pixel group can be measured by the millimeter wave radar, and the target environment prompt information can be generated. The target environment prompt information can include the category corresponding to the target pixel group and the distance between the target vehicle and the object corresponding to the target pixel group.

[0205] Optionally, the training process of the target recognition network can include the following steps:

[0206] First, the target recognition network is constructed.

[0207] For example, a convolutional neural network can be constructed as the target recognition network before training.

[0208] Second, a target region set and the category corresponding to each target region in the target region set are obtained.

[0209] The target region in the target region set can be a region where an object photographed in an indoor parking lot is located. The object can be, but is not limited to, a vehicle, a person, a column, and the ground. The category corresponding to the target region can be the category of the object corresponding to the target region.

[0210] Third, the constructed target recognition network is trained according to the target region set and the category corresponding to each target region in the target region set, to obtain the trained target recognition network.

[0211] For example, the target region set can be used as the training set of the target recognition network, and the category corresponding to the target region can be used as the training label of the target recognition network, to train the constructed target recognition network to obtain the trained target recognition network.

[0212] Optionally, the speed of the object that needs to be avoided can be obtained for the two target environment images, the category of the object that needs to be avoided is identified, the distance between the target vehicle and the object that needs to be avoided can be obtained through the dot matrix data of the millimeter wave radar, and the driver is prompted, so that the driver can make the optimal collision avoidance scheme.

[0213] In actual situations, in order to more accurately judge the movement of the collision-free object, more than 2 frames of target environment images can also be compressed, so as to subsequently transmit the target environment images to a module that can analyze the collision-free object, so as to more accurately analyze the movement of the collision-free object.

[0214] In summary, first, since the initial indoor environment image often contains the indoor environment information near the target vehicle driving, obtaining the initial indoor environment image can facilitate subsequent collision-free object analysis of the driving environment near the target vehicle, and obtaining two initial indoor environment images corresponding to the target vehicle can facilitate subsequent analysis of the running track of the collision-free object and can facilitate subsequent collision-free prompting of the target vehicle. Then, the foreground removal processing is performed on each initial indoor environment image, so as to obtain a reference image representing the light intensity of the environment, which can facilitate subsequent adaptive reflection interference elimination processing of the initial indoor environment image to remove the reflection interference in the initial indoor environment image caused by the light intensity. Next, based on the reference image, the adaptive reflection interference elimination processing is performed on the initial indoor environment image, which can eliminate the reflection interference in the initial indoor environment image to a certain extent, thereby improving the quality of the initial indoor environment image and facilitating subsequent collision-free object analysis. Continuing, the feature point detection processing is performed on the target environment image, which can facilitate subsequent light flow matching processing of the two obtained target environment images based on the target feature point set. Then, based on the target feature point set, the light flow matching processing is performed on the two target environment images, which can obtain a more accurate matching light flow field and can facilitate subsequent accurate analysis of the running track of the collision-free object. Thereafter, the importance analysis processing is performed on each pixel point group in the target environment image, which can improve the accuracy of the importance level determination of the pixel point group and can facilitate subsequent adaptive coding of the pixel point group in the target environment image, can avoid encoding important information as easily lost coding information, and can reduce the loss of important information, thereby improving the safety of traffic. Furthermore, based on the importance level of each pixel point group, the adaptive coding is performed on the pixel point group in the target environment image, which can avoid encoding important information as easily lost coding information, can reduce the loss of important information, and thereby can improve the safety of traffic. Finally, based on the target coding set, the two obtained target environment images are compressed, which can reduce the loss of important information in the compression process. Therefore, the target environment image with high quality is obtained, and the adaptive coding is performed on the pixel point group in the target environment image, which can reduce the loss of important information, thereby improving the safety of traffic.

[0215] It should be noted that the collision-free system can be a hardware system, which can include a camera for collecting an initial indoor environment image and a data processing module, wherein the data processing module can be used to implement the steps of the image processing method for indoor environment. The camera for collecting the initial indoor environment image can be connected with the data processing module.

[0216] The above embodiments are only used to illustrate the technical solutions of the present application, but not limit it; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced by equivalents; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. An image processing method for an indoor environment, characterized by, The method comprises the following steps: obtaining two initial indoor environment images corresponding to a target vehicle; extracting ambient light from each initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image; performing adaptive reflection interference elimination processing on each pixel point in the initial indoor environment image based on the reference image corresponding to each initial indoor environment image and a preset sliding window and a preset convolution window corresponding to each pixel point in the initial indoor environment image to obtain a target gray value corresponding to the pixel point; updating the gray value corresponding to each pixel point in each initial indoor environment image to the target gray value corresponding to the pixel point to obtain a target environment image corresponding to the initial indoor environment image; the step of extracting ambient light from each initial indoor environment image to obtain a reference image corresponding to the initial indoor environment image comprises: performing high-frequency removal processing on the initial indoor environment image to obtain a low-frequency component image; smoothing the low-frequency component image to obtain the reference image corresponding to the initial indoor environment image; the step of performing adaptive reflection interference elimination processing on each pixel point in the initial indoor environment image to obtain a target gray value corresponding to the pixel point comprises: performing negative correlation mapping on the cumulative sum of the gray values of the pixel points in the preset convolution window corresponding to the pixel point to obtain a first gray value; determining the product of the gray value of each pixel point in the preset convolution window corresponding to the pixel point and the first gray value as the reference weight of each pixel point in the preset convolution window to obtain a reference weight set; determining the product of each reference weight in the reference weight set and the gray value of each pixel point in the preset sliding window corresponding to the pixel point as a first index to obtain a first index set; determining the cumulative sum of each first index in the first index set as the target gray value corresponding to the pixel point.

2. The image processing method for an indoor environment according to claim 1, wherein The method further comprises: performing feature point detection processing on each obtained target environment image to obtain a target feature point set included in the target environment image; performing optical flow matching processing on the two obtained target environment images according to the two obtained target feature point sets to obtain a matching optical flow field; grouping the pixel points in the two obtained target environment images according to the matching optical flow field to obtain a pixel point group set corresponding to each target environment image, and performing importance analysis processing on each pixel point group in the pixel point group set to obtain an importance level corresponding to the pixel point group; performing adaptive coding on the pixel point groups in the target environment image according to the importance level corresponding to each pixel point group in the pixel point group set corresponding to the target environment image to obtain a target coding set corresponding to the target environment image; compressing the two obtained target environment images according to the target coding set corresponding to the target environment image.

3. The image processing method for an indoor environment according to claim 2, wherein the step of performing optical flow matching processing on the two obtained target environment images according to the two obtained target feature point sets to obtain a matching optical flow field comprises: According to the collection time corresponding to the two target feature point sets, the two target feature point sets are determined as a first target feature point set and a second target feature point set respectively; For each first target feature point in the first target feature point set and each second target feature point in the second target feature point set, a target difference index between the first target feature point and the second target feature point is determined according to a preset window corresponding to the first target feature point and a preset window corresponding to the second target feature point; For each first target feature point in the first target feature point set, a second target feature point with a minimum target difference index between the first target feature point and the second target feature point is selected from the second target feature point set as a third target feature point corresponding to the first target feature point; When a target difference index between a first target feature point in the first target feature point set and a third target feature point corresponding to the first target feature point is less than or equal to a preset difference threshold, the first target feature point and the third target feature point are determined as two matching feature points matched with each other; Matching feature points in the first target feature point set are combined as a matching feature point set, and matching feature points in the second target feature point set are combined as another matching feature point set, to obtain two matching feature point sets; A light flow field between the two matching feature point sets is determined as a matching light flow field.

4. The image processing method for an indoor environment according to claim 3, wherein The target difference index between the first target feature point and the second target feature point is determined according to the preset window corresponding to the first target feature point and the preset window corresponding to the second target feature point, including: A reference image corresponding to a target environment image in which the first target feature point is located is determined as a first reference image, and a reference image corresponding to a target environment image in which the second target feature point is located is determined as a second reference image; A pixel point at a same position as the first target feature point in the first reference image is determined as a first reference feature point, and a pixel point at a same position as the second target feature point in the second reference image is determined as a second reference feature point; Gray values of pixel points at same positions in preset windows corresponding to the first target feature point and the first reference feature point are combined as a first coordinate to obtain a first coordinate sequence, and gray values of pixel points at same positions in preset windows corresponding to the second target feature point and the second reference feature point are combined as a second coordinate to obtain a second coordinate sequence; Euclidean distances between coordinates at same positions in the first coordinate sequence and the second coordinate sequence are determined as first distances to obtain a first distance sequence; An average value of all first distances in the first distance sequence is determined as a first difference index. The first difference index is normalized to obtain the target difference index between the first target feature point and the second target feature point.

5. The image processing method for an indoor environment according to claim 3, wherein The pixel points in the obtained two target environment images are grouped according to the matched optical flow field, and a pixel point group set corresponding to each target environment image is obtained, including: Each velocity vector in the matched optical flow field is clustered to obtain a cluster set; Each velocity vector in each cluster in the cluster set is determined as a motion feature point to obtain a motion feature point group corresponding to the cluster; The motion feature points in the motion feature point group corresponding to each cluster are grouped to obtain a first motion feature point group and a second motion feature point group corresponding to the cluster, wherein the motion feature points in the first motion feature point group are located in the first target environment image, the motion feature points in the second motion feature point group are located in the second target environment image, and the first target environment image and the second target environment image are the two obtained target environment images; The pixel points in the minimum circumscribed rectangle corresponding to each first motion feature point group in the first target environment image are combined into a pixel point group, and the pixel points in the first target environment image except the pixel points in the minimum circumscribed rectangle corresponding to each first motion feature point group are combined into a pixel point group, to obtain a pixel point group set corresponding to the first target environment image; The pixel points in the minimum circumscribed rectangle corresponding to each second motion feature point group in the second target environment image are combined into a pixel point group, and the pixel points in the second target environment image except the pixel points in the minimum circumscribed rectangle corresponding to each second motion feature point group are combined into a pixel point group, to obtain a pixel point group set corresponding to the second target environment image.

6. The image processing method for an indoor environment according to claim 2, wherein The importance analysis processing of each pixel point group in the pixel point group set is performed to obtain an importance level corresponding to the pixel point group, including: The region where the pixel point group is located is input into a pre-trained importance degree discrimination network to obtain an importance level corresponding to the pixel point group; The training process of the importance degree discrimination network includes: Constructing an importance degree discrimination network; Obtaining a target region set and an importance level corresponding to each target region in the target region set; According to the target region set and the importance level corresponding to each target region in the target region set, the constructed importance degree discrimination network is trained to obtain a trained importance degree discrimination network.

7. The image processing method for an indoor environment according to claim 2, wherein The importance level corresponding to each pixel point group in the pixel point group set corresponding to each target environment image is used to perform adaptive coding on the pixel point groups in the target environment image, including: According to the importance level corresponding to each pixel point group in the pixel point group set corresponding to the target environment image, each pixel point group in the pixel point group set is subjected to Huffman coding to obtain a target code corresponding to the pixel point group.

8. An image processing system for an indoor environment, characterized by comprising: A processor and a memory are included, and the processor is used to process instructions stored in the memory to implement the image processing method for indoor environment in any one of claims 1-7.

Images