Kuburitswa kwekombuta yekuona raibhurari OpenCV 4.2

zvakaitika free library library OpenCV 4.2 (Open Source Computer Vision Library), iyo inopa maturusi ekugadzirisa uye kuongorora zvirimo zvemufananidzo. OpenCV inopa anopfuura zviuru zviviri nemazana mashanu algorithms, ese echinyakare uye anoratidza kufambira mberi kwechiono chekombuta uye masisitimu ekudzidza muchina. Iyo raibhurari kodhi yakanyorwa muC ++ uye inoparadzirwa ne pasi peBSD rezinesi. Zvisungo zvakagadzirirwa mitauro yakasiyana-siyana yekuronga, kusanganisira Python, MATLAB uye Java.

Raibhurari inogona kushandiswa kuona zvinhu zviri mumifananidzo nemavhidhiyo (semuenzaniso, kuzivikanwa kwezviso uye manhamba evanhu, zvinyorwa, nezvimwewo), kutevedzera mafambiro ezvinhu nemakamera, kurongedza zviito muvhidhiyo, kushandura mifananidzo, kubvisa 3D modhi, kugadzira 3D nzvimbo kubva kumifananidzo kubva kumakamera estereo, kugadzira mifananidzo yemhando yepamusoro nekubatanidza mifananidzo yakaderera, kutsvaga zvinhu zviri mumufananidzo zvakafanana neakaunzwa seti yezvinhu, kushandisa nzira dzekudzidza muchina, kuisa mamaki, kuona zvinhu zvakajairika mune zvakasiyana. mifananidzo, inobvisa otomatiki zvirema zvakaita seziso dzvuku .

В nyowani kusunungura:

• Kumashure kwekushandisa CUDA kwawedzerwa kuDNN (Deep Neural Network) module nekuitwa kwemashini ekudzidza algorithms yakavakirwa neural network uye yekuyedza API rutsigiro yaitwa. nGraph OpenVINO;
• Kushandisa SIMD mirairo, kushanda kwekodhi kwakagadziridzwa kuburitsa stereo (StereoBM/StereoSGBM), resizing, masking, kutenderera, kuverenga kwemavara asipo uye mamwe akawanda maoperation;
• Yakawedzerwa multi-threaded kuitiswa kwebasa racho pyrDown;
• Yakawedzera kugona kuburitsa vhidhiyo hova kubva mumidziyo midhiya (demuxing) uchishandisa iyo videoio backend yakavakirwa paFFmpeg;
• Yakawedzera algorithm yekukurumidza frequency-selective kuvakwazve kwemifananidzo yakakuvadzwa FSR (Frequency Selective Reconstruction);
• Yakawedzerwa nzira Ric kupindirana kwenzvimbo dzakajairwa dzisina kuzadzwa;
• Akawedzera kutsauka normalization nzira Logos;
• Iyo G-API module (opencv_gapi), inoshanda seinjini yekugadzira mifananidzo inoshanda uchishandisa graph-based algorithms, inotsigira yakanyanya kuoma kunzwisisa komputa yekuona uye yakadzika muchina kudzidza algorithms. Tsigiro yeIntel Inference Engine backend inopihwa. Yakawedzera rutsigiro rwekugadzirisa vhidhiyo hova kune eexecution modhi;
• Yakabviswa vulnerabilities (CVE-2019-5063, CVE-2019-5064), iyo inogona kutungamira kune anorwisa kodhi kuuraya kana uchigadzira isina kusimbiswa data muXML, YAML uye JSON mafomati. Kana munhu ane null kodhi akasangana panguva yeJSON parsing, kukosha kwese kunokopwa kune buffer, asi pasina kunyatso tarisa kuti inodarika miganhu yenzvimbo yakagoverwa yekurangarira.

Source: opennet.ru