Yakadzora Kuparadza Kwemberi - Kupera uye Kudzokera - Kupera Intelligent Algorithm mumatauriri ekuongorora

1. Front - End Algorithm Kubatsira

The theFront - KuperaAlgorithms inoshanda zvakananga mukati mekamera kamera, kazhinji inokweretesamupendero comp computingkugona. Aya algorithms anovavariraprocess raw sensor data munharaunda, nekudaro kuderedza bandwidth uye server mutoro nekuita mabasa ekutanga pane kamera kamera. Ngationgororei zvinhu zvikuru:

a. Kamera hardware uye sensor yekubatanidza

Makamera emazuva ano ekuongorora anosanganisira akawanda marudzi eSensors:

• Mufananidzo Sensors (CMOS, CCD): Contract Consaal data (mifananidzo uye mavhidhiyo) pasi pemamiriro ezvinhu akasiyana-siyana.
• Infrared (IR) sensors: Ita kuti kamera ive yekutakura vhidhiyo mune yakaderera mwenje kana rima rakazara.
• Lidar uye yakadzika sensors: Kuyera madaro uye kuona zvinhu mu 3D nzvimbo, inobatsira pakusiyanisa pakati pezvinhu uye kumashure mune chiitiko.
• Microphones: Dzimwe nguva zvakabatanidzwa kune Audio - based analytics.

Aya ma sensors anotumira data rakabikwa kune iyo yekugadzirisa unit, ipo algorithms likeMufananidzo Pre - Kugadzirisavanoiswa.

b. Mufananidzo Pre - Kugadzirisa uye Kuderedzwa Kwemuruki

Usati waongorora chero kuongororwa kwakaoma,Mufananidzo Pre - Kugadzirisainokosheswa kusimudzira kunaka kwetsoka, kunyanya mumamiriro ezvinhu asina kunaka emwenje kana nzvimbo dzine ruzha:

• Kuratidzira algorithms: Bvisa sensor ruzha, kazhinji uchishandisa mafirita sengeGaussian Blur or Kwete - Yenzvimbo Inoreva Kuratidzira.
• Kusiyanisa uye kupenya kugadzirisa: Algorithms likeAdaptive Histograph kuenzaniswaRongedza kupenya uye kusiyana nekuwedzera kuoneka.
• Kuongororwa kwekugadziriswa: Kuongororwa kwekugadzwa (i.e.,Sobel Operator, Canny Edrest Kuonekwa) Inogona kubatsira kutsanangura zvinokonzerwa nechinhu, icho chakakosha kuchitevera.

c. Kuongororwa kwekufamba uye kumashure kwekutanga

Kufamba kwekuonandeimwe yemabasa akakosha anoitwa kumberi - End Algorithms. Iyo inowanzoenderana neiyo nheyo yekufananidza mafuremu anoteuki kuti aone zvinhu zvinofamba.

• Background Kugadziriswa. Chero shanduko yakakosha yakafemerwa sekufamba.
• Tsvina kuwirirana.
• Optical kuyerera.Kalman mafiritazvekutevera.

d. Chinhu chinowanikwa uye kuteedzera

Pamberi - Kupera, Chinhu Chekuona uye Kutsvaga kunoitwa munharaunda kuti uzive uye track zvinhu (i.e., vanhu, mota, mhuka). Maitiro makuru anosanganisira:

• Yolo (iwe unongotarisa kamwechete): A State - of - the - the - Art Algorithm inogona kuona zvinhu zvakawanda zviri muchokwadi - nguva. Yolo inokamura mufananidzo mune grid uye inofanotaura kusunga mabhokisi kune chimwe chinhu chimwe nechimwe muGrid.
• Haar Cascade Classifiers.
• Kalman Filter: Inoshandiswakuteverakufambisa zvinhu pamapuranga. Iyo inofungidzira mamiriro echinhu chinofamba (chinzvimbo, velocity) uye vanofanotaura nezvechinzvimbo chamangwana.

e. Anomaly Kuwanikwa uye Chiitiko Zvinokonzeresa

Anomaly Kuwanikwa kumberi - Kupera Kazhinji kunotarisa pakuzivisa zviitiko zvisina kujairika mumavhidhiyo chikafu:

• Kamwe kamwe kufamba: Kuonekwa kwekukurumidza kana kusingafungidziriki mafambiro, senge mumwe munhu anomhanya kana kamwe kamwe kuumbwa.
• Cross - Kuongororwa mutsara: Inoshandisa Virtual Tripwires kana mitsara inokonzeresa zvambiro kana chinhu chichiyambuka.
• Nharaunda kupinda: Kuteeredzwa kana chinhu chikapinda kana chinobuda nzvimbo yakatsanangurwa mukati mechimiro.

Aya algorithms anogona kukonzeresa chaiyo - nguva yambiro yeBack - Kuperasystem kana tumira zviziviso zvekuchengetedzeka kune vashandi vekuchengetedza.

2. Back - End Algorithm Kubatsira

The theBack - KuperaSisitimu ine mhosva yekukasira kusimudza, kubata kuomarara data anaye analytics uye kuchengeta hombe mavhoriyamu evhidhiyo data. Inoshanda nekugamuchira vhidhiyo mahova kana metadata kubva kumberi - makamera ekupedzisira uye anoita kuongorora Heano kuputsa kweKey MabasaKuitwa Nekudzoka - Kupera Algorithms:

a. Vhidhiyo rwizi uye kutapurirana kwedata

• Kuunganidzwa kwedata: Makamera anotenderera vhidhiyo data kumusana - Kupera kungave kuburikidza neakananga internet kubatana, nzvimbo yenzvimbo network (mafuro), kana masevhisi ezvemakore.
• Kumanikidzwa: Kuti uderedze bandwidth kushandiswa, hova dzemavhidhiyo dzinowanzo kumanikidzwa uchishandisa zviyero sengeH.264 or H.265, iyo inochengetedza mhando yemhando iyo ichidzikisa faira saizi.

b. Vhidhiyo ongororo uye kudzidza zvakadzama

• Chinhu chinowanikwa: Iyo yekumashure - Kupera inoshandisa zvakadzika mwero mwero sengeYolo, Faster R - CNN, kanaSSD. Mhando idzi dzakadzidziswa pamatanho makuru kuti uzive zvinhu zvakasiyana siyana zvakaita sevanhu, mota, mhuka, nezvimwe.

• Kuzivikanwa kwechiso.Facenet or Deepface. Aya mhando dzinoenzaniswa nezviso muVhidhiyo Tsoka kune Database yeVanhu Vanozivikanwa.

• Kuzivikanwa kwechiito: Pamusoro pekuona zvinhu zvekuona, kumashure - magumo anogona zvakare kusarudzira zviito kana maitiro mukati mevhidhiyo. Semuenzaniso, kuona kurwa, kufamba kwekufungidzira, kana mamwe maitiro akafungidzirwa achishandisaRnns (Recurrent neural network) or 3d cnns.

• Chiitiko Chiitiko.

c. Metadata tagging uye kusarongeka

• Tagging: Chimiro chega chega kana chevhidhiyo chikamu chakamiswa neakakodzera metadata (i.e., nguva, nzvimbo, inoonekwa zvinhu, zviitiko).
• Indexing: Vhidhiyo uye Chiitiko data ine indexed kubvumira kuti isatsvakwa kutsvaga. Uchishandisa technologies sengeElasticsearch, zvinova nyore kutsvaga kuburikidza nehuwandu hwakawanda hwe data vhidhiyo zvichibva pamatiki kana metadata.

Semuenzaniso, iwe unogona kutsvaga "vanhu vanoonekwa munzvimbo yakadzorwa kubva ku2 PM kusvika 3 PM."

d. Kuongororwa Kwemaitiro uye Anomaly Kuwanikwa

• Kuziva Kwekutanga. Iyo modhi saka mireza mireza kubva pane zvakajairwa.

• Chiitiko Correlation: Back - End Systems inogona kugadzirisa zviitiko zvakawanda kana hova dze data (i.e., kusanganisakufamba kwekuonanaKuzivikanwa kwechiso). Kana chiitiko chisina kujairika chakaonekwa, hurongwa hunogona kuburitsa zvekuzivisa.

• Kureba - Kuongorora Kwese.

e. Gore kubatanidza uye kusajaira

• Cloud.

• Cloud AI kugadziriswa. Gore rinogona zvakare kushandiswa kudzidzisa mhando pamatanho makuru.

3. Chiitiko Chekushandisa

Nekukakavara kwepamberi kwepamberi - kupera uye kumashure - kuguma kwehungwaru Algorithms, manhamba ekuongorora masisitimu ari kushandiswa mune dzakasiyana siyana application:

a. Urban kuongororwa mumaguta akangwara

• Traffic trafficing: Makamera anogona kutarisa kuyerera kwemigwagwa, tarisa tsaona, uye track trackcles yekutyora sekumhanya kana kumhanya marara matsvuku.

• Boka revanhu: Makamera akavezwa nevanhu vachiverenga uye vanoongorora maitiro algorithms vanobatsira kubata mwero maungira, kuvimba nekuchengetedza munzvimbo yeruzhinji.

• Kuchengetedzwa kweveruzhinji: Makamera anogona kuona zvisina kujairika maitiro (i.e., kurwa kana kuseta) uye nekukurumidza ziviso zviremera.

b. Retail Kuongorora kwekutengesa kwekutengesa uye vatengi vanonzwisisika

• Kubata kwekutengesa.

• Vatengi Vanoongorora.

c. Healthcare uye kuchengetedzeka kwechipatara

• Murwere Kutarisa: Muzvipatara, makamera anehungwaru ekuongorora anogona kuongorora mafambiro emurwere kuti awonekwe, kuwanikwa kusingabvumirwe kunzvimbo dzinonzwisisika, kana varwere vari mukutambudzika.

• Dziviriro Spage.

d. Kudzivirira Kanyaya Zvivakwa

• Yakakwira - nzvimbo dzekuchengetedza.

e. Kuchengetedzwa kwemusha

• Intruder Kuwanikwa: Muchengetedzeka kumba, makamera nechiso uye chekucherekedza algorithms inogona kuziva mapenzi, anozivisa varidzi vemba, uye trigger alarms.

• Package yekubira kudzivirira: Makamera anokwanisa kuona zviitiko zvinofungidzirwa zvine chekuita nekuba kwepakeji uye zivisa varidzi vemba.

Mhedziso

Kubatanidzwa kweakangwara algorithmspane zvese zviri zviviriFront - KuperauyeBack - Kuperainochinja munda wekuongorora. Kubva pane yekutanga data yekuwana uye chiitiko chekutanga pane kamera padanho rekutanga kuongorora uye muchina wokudzidza kuSeva - Side Sezvo kudzidza kweAi uye muchina kuenderera mberi nekuitika, masisitimu aya anowedzera kuchengetedzeka, manejimendi ari nani, uye kugona kufungidzira kunogona kudzivirira kungangodzivisha vasati vakunda.