The invention relates to a traffic scene participant identification method based on a vehicle-mounted binocular camera The method overcomes the defects of the prior art According to the technical scheme, data are acquired through a binocular system The method is characterized by comprising the steps that 1, the vehicle-mounted binocular camera acquires a current road image and conducts precise image pretreatment and binocular parallax matching cost computation; 2, longitudinal road surface parameter estimation is conducted, and horizon line and road surface range information is acquired; 3, the position and dimension information of a potential obstacle are determined by means of parallax matching cost; 4, the characteristic information of the obstacle is described by means of a deep convolutional neural network, an obstacle detection area is verified based on output of a characteristic chromatogram by means of a softmax classifier, and the category of the obstacle is determined Besides profile and texture information, parallax information can also be utilized, so that identification accuracy is far higher than that of a superficial zone identification method

Traffic scene participant identification method based on vehicle-mounted binocular camera

The invention provides a binocular camera disparity map based road obstacle detection system and method. The system mainly comprises a binocular stereo rectification module, a binocular stereo matching module, a Gaussian filter module, an image cropping module and a disparity map scanning module. The binocular stereo rectification module transforms one image according to positional relation (R, T) between two cameras to enable corresponding polar lines of left and right camera images to be parallel. The binocular stereo matching module performs binocular stereo matching by adopting a map partitioning algorithm, and calculates disparity between image points corresponding to the left and right cameras to obtain a disparity map of a scene. The Gaussian filter module performs Gaussian filter on the original disparity map to obtain a fuzzy disparity map. The image cropping module introduces prior information to crop the disparity map to a certain degree. The disparity map scanning module scans the disparity map line by line to determine the position of an obstacle on the disparity map, then performs three-dimensional reconstruction on an extracted area, and determines the position of the obstacle so as to provide decision-making bases for an on-board visual system.

Binocular camera disparity map based road obstacle detection system and method

The invention discloses a non-contact object size and distance image measuring instrument, which comprises modules including a zoom lens, a CCD (charge coupled device) image sensor, a laser light source, a DSP (digital signal processing) image processing board card, a miniature touch displayer, a power supply module, an instrument panel, a portable instrument case and the like. The measuring instrument firstly obtains images of objects to be measured, then, the images to be processed are subjected to denoising enhancing and interpolation super-resolution processing, and then, light spot detection and pixel calibration are carried out; after a user selects measuring objects, a system segments the measuring objects, and then, the size of the objects and the distance between the objects are calculated; the instrument can simultaneously measure the distance between any two points; a measuring result is finally sent to a display to be displayed in a marked way. The measuring instrument has the advantages that the configuration of the laser light source is flexible, the pixel calibration is simple, the calculation on the sizes and the distance of the objects is accurate, the manufacturing difficulty is low, the portability is realized, the measurement is accurate, the measurement efficiency is high, and the application range is wide.

Non-contact object size and distance image measuring instrument

A system and method for binocular stereo vision is disclosed. The method includes acquiring a pair of images including a first image and a second image, the pair of images being captured by one or more cameras; obtaining a training model; determining a plurality of feature images according to the first image; determining one or more features of an object in the first image according to the training model and the plurality of feature images; determining a first area in the first image, the first area including at least one of the determined one or more features; obtaining depth information of the determined first area based on the first and second images; and determining a second area in the first image based on the determined first area and depth information, the second area including at least one of the determined first area.

Methods and systems for binocular stereo vision

The invention discloses an auxiliary driving obstacle detection method based on binocular stereoscopic vision, and belongs to the technical field of environmental perception. The method mainly comprises the steps of synchronously collecting two frame images output by two CCD cameras respectively, wherein the two CCD cameras are arranged right ahead of a vehicle; carrying out distortion correction on the collected images; calling a stereo matching sub-process to carry out stereo matching on the two corrected images to obtain the corresponding relations of all pixel points; respectively calculating the projection matrixes of the two CCD cameras according to camera parameters; finding all corresponding pixel point pairs in the two corrected images; carrying out scene three-dimensional reconstruction on the images collected in real time according to the projection matrixes to obtain the space coordinate of an obstacle scene, screening obstacle areas, and outputting an obstacle position information image. The method achieves the function of automatically detecting road obstacles in the static or advancing process of the vehicle, and has the advantages of being capable of achieving the unattended operation and high in detection result precision.

Auxiliary driving obstacle detection method based on binocular stereoscopic vision

The invention discloses an intelligent vehicle obstacle avoiding method based on binocular vision and belongs to the field of computer vision technology and fuzzy recognition. The intelligent vehicle obstacle avoiding method includes that a left video camera and a right video camera are calibrated in a binocular mode, a relation between a video camera image coordinate system and a scene coordinate system is obtained, and the left video camera and the right video camera are used for collecting a scene in front of a vehicle; distortion and epipolar rectification are conducted on two images, feature points of the images are extracted, and feature vector indication is conducted on the feature points of the images; stereo matching is conducted according to the feature vector, an initial matching dot pair is obtained according to the feature vector, false matching dot pairs are removed, and a correct matching dot corresponding relation is built; according to the correct matching dot pairs, a disparity map is obtained through calculation, and three-dimensional real information of an obstacle is recovered according to the disparity map; and a fuzzy controller is built according to the three-dimensional information of the obstacle, and the vehicle is guided to avoid the obstacle automatically. By means of the intelligent vehicle obstacle avoiding method, the shape of the obstacle and the distance between the obstacle and the vehicle can be detected correctly, and experiments show that the method is good in real-time performance and high in obstacle avoiding rate.

Li Dan

Papers

Intelligent vehicle obstacle avoiding method based on binocular vision

Vision detecting system based on DSP