Optimum detectors for artificial object recognition

目前的数码相机具有线性响应，并且只适合在受控光线条件下进行简单的图像成像。事实上，与人类视觉相比，这些相机有两个关键的和令人衰弱的限制，即有限的动态范围和对象颜色的可变性。当场景的动态范围大于摄影机的动态范围时，第一个限制会导致饱和。第二个限制更微妙，这意味着与人类视觉不同，在人工目标识别系统中很难使用来自线性摄像机的颜色信息。使用具有相对窄的光谱响应的像素和与检测到的光电流的对数成比例的输出，可以克服这两个限制。我们打算通过开发一种基于有机光电二极管和对数响应的新相机技术来克服这些限制。

Current digital cameras have a linear response and are only really suitable for simple picture imaging under controlled light conditions. In fact compared to human vision these cameras have two critical and debilitating limitations, a limited dynamic range and variability in object colour. The first of these limitations causes saturation when the dynamic range of a scene is larger than that of the camera. The second limitation is subtler and means that unlike human vision it is difficult to use colour information from linear cameras in artificial object recognition systems. Both these limitations can be overcome using pixels with a relatively narrow spectral response and an output that is proportional to the logarithm of the detected photocurrent. We intend to overcome these limitations by developing a new camera technology based upon organic photodiodes and a logarithmic response.