近日,澳大利亞科學(xué)家用方程式表達(dá)出了與蒼蠅視力相關(guān)的大腦細(xì)胞活性,。他們通過(guò)這些方程式,,發(fā)現(xiàn)了非常簡(jiǎn)單有效的方法,可以從原始數(shù)據(jù)中處理運(yùn)動(dòng)模式,,這種運(yùn)動(dòng)模式指的是一個(gè)物體,、表面、邊緣在一個(gè)視角下由一個(gè)觀察者(比如眼睛,、攝像頭等)和背景之間形成的明顯移動(dòng),,并用于小型無(wú)人飛行機(jī)器人遙感導(dǎo)航系統(tǒng)。
據(jù)報(bào)道,,澳洲科學(xué)家建立的這個(gè)系統(tǒng)在將來(lái)可能用來(lái)為小型無(wú)人駕駛飛機(jī),、搜索和救援機(jī)器人,、汽車導(dǎo)航系統(tǒng)和其他系統(tǒng)的視覺(jué)系統(tǒng)編程。大衛(wèi)·歐·卡洛(David O’Carroll)是澳大利亞阿德萊德大學(xué)的研究昆蟲(chóng)視覺(jué)的計(jì)算神經(jīng)科學(xué)家,,他說(shuō):“我們從生物學(xué)中獲取靈感,,制作出了這樣一個(gè)非線性系統(tǒng)。這個(gè)系統(tǒng)涉及的計(jì)算量非常少,,而且,,這個(gè)系統(tǒng)得出計(jì)算結(jié)果所需要的浮點(diǎn)運(yùn)算次數(shù)比傳統(tǒng)方法少成千上萬(wàn)倍。”
為了制造出小型化的飛行機(jī)器人,,研究人員需要更簡(jiǎn)單的方式來(lái)處理運(yùn)動(dòng)過(guò)程?,F(xiàn)在,,研究人員已經(jīng)從小小的蒼蠅身上找到了靈感,,因?yàn)樯n蠅僅用相對(duì)少的神經(jīng)元就可以非常靈巧的飛翔。在10年前,,歐·卡洛和其它研究者煞費(fèi)苦心的開(kāi)展了蒼蠅飛行研究,,并測(cè)量出飛行過(guò)程中大腦細(xì)胞的活性,同時(shí),,進(jìn)一步將這些結(jié)果轉(zhuǎn)化為一套計(jì)算規(guī)則,。
11月13日,歐·卡洛和他的同事生物學(xué)家羅素·布林克沃思(Russell Brinkworth) 在《公共圖書(shū)館·計(jì)算生物學(xué)》上發(fā)表了一篇文章中稱,,他們測(cè)試了這套系統(tǒng),。歐·卡洛說(shuō):“筆記本電腦的功率達(dá)幾十瓦,而我們的系統(tǒng)功率消耗不足毫瓦,。”
研究者的算法由5個(gè)方程組成,,通過(guò)這5個(gè)方程,可以計(jì)算從攝像機(jī)獲得的數(shù)據(jù),。每個(gè)方程表示了飛行中所用的技巧,,通過(guò)這個(gè)方程可以掌控強(qiáng)度、對(duì)比度和移動(dòng)的變化量,,并且方程的參數(shù)隨著輸入的不同不斷改變,。這套算法不會(huì)比較兩幀圖像的像素變化,只是強(qiáng)調(diào)大范圍的運(yùn)動(dòng)變化模式,。從這種意義上說(shuō),,它的工作原理有點(diǎn)像視頻壓縮系統(tǒng)。
為了測(cè)試這套算法,,歐·卡洛和布林克沃思分析了動(dòng)畫(huà)片中的高分辨率圖像,。當(dāng)他們將輸入和輸出進(jìn)行比較后, 他們發(fā)現(xiàn)這個(gè)算法可以工作在大量自然光環(huán)境中以及工作在移動(dòng)探測(cè)器都感到困難的地方,。肖恩·亨伯特(Sean Humbert )說(shuō):“這真是一個(gè)令人驚異的工作,。”亨伯特是馬里蘭大學(xué)的一名航天工程師,,他制造了小型無(wú)人飛行機(jī)器人。亨伯特說(shuō):“傳統(tǒng)的遙感導(dǎo)航系統(tǒng)需要大量的設(shè)備來(lái)計(jì)算,。但是裝在這些機(jī)器人身上的設(shè)備非常小,。”
歐·卡洛說(shuō):“我們的工作從昆蟲(chóng)的視覺(jué)獲得靈感,并制造出一個(gè)現(xiàn)實(shí)世界中可用的模型,。但是,,在這個(gè)過(guò)程中,我們已經(jīng)制作出和昆蟲(chóng)一樣復(fù)雜的系統(tǒng),。這是一個(gè)有趣的事情,,這個(gè)事情不是讓我們?nèi)ネ耆私膺@個(gè)系統(tǒng)是如何工作的,它只是讓我們了解大自然是正確的,。”(生物谷Bioon.com)
生物谷推薦原始出處:
PLoS Comput Biol 5(11): e1000555. doi:10.1371/journal.pcbi.1000555
Robust Models for Optic Flow Coding in Natural Scenes Inspired by Insect Biology
Russell S. A. Brinkworth*, David C. O'Carroll
Discipline of Physiology, School of Molecular and Biomedical Science, The University of Adelaide, South Australia, Australia
The extraction of accurate self-motion information from the visual world is a difficult problem that has been solved very efficiently by biological organisms utilizing non-linear processing. Previous bio-inspired models for motion detection based on a correlation mechanism have been dogged by issues that arise from their sensitivity to undesired properties of the image, such as contrast, which vary widely between images. Here we present a model with multiple levels of non-linear dynamic adaptive components based directly on the known or suspected responses of neurons within the visual motion pathway of the fly brain. By testing the model under realistic high-dynamic range conditions we show that the addition of these elements makes the motion detection model robust across a large variety of images, velocities and accelerations. Furthermore the performance of the entire system is more than the incremental improvements offered by the individual components, indicating beneficial non-linear interactions between processing stages. The algorithms underlying the model can be implemented in either digital or analog hardware, including neuromorphic analog VLSI, but defy an analytical solution due to their dynamic non-linear operation. The successful application of this algorithm has applications in the development of miniature autonomous systems in defense and civilian roles, including robotics, miniature unmanned aerial vehicles and collision avoidance sensors.