Evolutionary learning of local descriptor operators for object recognition

Abstract

Nowadays, object recognition is widely studied under the paradigm of matching local features. This work describes a genetic programming methodology that synthesizes mathematical expressions that are used to improve a well known local descriptor algorithm. It follows the idea that object recognition in the cerebral cortex of primates makes use of features of intermediate complexity that are largely invariant to change in scale, location, and illumination. These local features have been previously designed by human experts using traditional representations that have a clear, preferably mathematically, well-founded definition. However, it is not clear that these same representations are implemented by the natural system with the same structure. Hence, the possibility to design novel operators through genetic programming represents an open research avenue where the combinatorial search of evolutionary algorithms can largely exceed the ability of human experts. This paper provides evidence that genetic programming is able to design new features that enhance the overall performance of the best available local descriptor. Experimental results confirm the validity of the proposed approach using a widely accept testbed and an object recognition application.

Evolutionary-computer-assisted design of image operators that detect interest points using genetic programming

“…It is possible to identify three types of GP-based approaches: (1) those that employ GP to detect low-level features which have been predefined by human experts, such as corners or edges [21,44,[60][61][62]67] and recently one regarding vegetation indices used on remote sensing [46,47];…”

“…Indeed, EC is based on the core principles of biological evolution, a natural process that exhibits an adaptive power that by far outstrips that of any humanengineered system [54]. Currently, a large amount of experimental evidence exists that confirms the ability of EC to outperform manmade solutions in many domains, such as antenna design, mathematical proofs, and even CV [4,25,26,44,56,62]. In many cases, the stochastic nature of EC allows it to sample large portions of the search space, and sometimes produce solutions that might not be evident to a human expert.…”

Multiple Face Recognition Using Local Features and Swarm Intelligence

“…In many recent computer vision applications, distinctive and representative regions of images are identified using interest points, which have been mainly applied for object recognition and related tasks [6], [32], [33]. Object recognition may be successful only if it is possible to find some distinctive image features among many alternative objects in cluttered real scenes [34].…”

The estimation of hölderian regularity using genetic programming

“…If we take this view, we can observe a close relation with the works in [22,23,25], for example, where GP was used to synthesize image operators that determine the saliency of each image pixel. Another example is the work of [17], where GP is given the task of finding operators that optimize the description of local image regions, or the work in [26] that uses GP to detect edge points. Regarding the application of GP to mathematics, several interesting proposals have also been developed.…”